Exploiting Earnings Volatility:

An Innovative New Approach to Evaluating,

Optimizing, and Trading Option Strategies to Profit
from Earnings Announcements

Brian Johnson
 Copyright 2015 Trading Insights, LLC

All rights reserved. Except as permitted under the U.S. Copyright Act of
1976, no part of this publication may be reproduced, distributed, or
transmitted in any form or by any means, without the prior written permission
of the publisher. This e-book is licensed for your personal use only. This e-
book may not be resold or transferred.

Disclaimer: Information in this e-book and in the accompanying

spreadsheets is provided solely for informational and general educational
purposes and should not be construed as an offer to sell or the solicitation of
an offer to buy securities or to provide investment advice. Option trading has
large potential rewards, but also large potential risk. You must be aware of
the risks and be willing to accept them in order to invest in the options
markets. Do not trade with money you cannot afford to lose.
 Dedication

To my former derivative students, whose desire to learn and seemingly

limitless supply of insightful questions fueled my enduring fascination with
options and inspired many new research ideas.
 Table of Contents
Copyright 2015 Trading Insights, LLC
Dedication
Introduction
Tips on Viewing Images
1 - Volatility Review
2 – The Aggregate IV Formula
3 – The Basic Spreadsheet
4 – True Greeks
5 – UA Earnings Strategy
6 – Integrated: Input & Import
7 – Integrated: Volatility Model & Simulation
8 – Integrated: Optimization & Analysis
9 – CREE Earnings Strategy
10 – Practical Considerations
About the Author
Resources
 Introduction

Brian Johnson, a professional investment manager with many years of trading

and teaching experience, wrote his first book, Option Strategy Risk / Return
Ratios: A Revolutionary New Approach to Optimizing, Adjusting, and
Trading Any Option Income Strategy in 2014. It has been one of the top-
selling option books and it has received outstanding reviews by the option
trading community.
His new book, Exploiting Earnings Volatility, introduces an innovative
new framework for evaluating, optimizing, and trading option strategies to
profit from earnings-related pricing anomalies. Leveraging his extensive
background in option-pricing and decades of experience in investment
management and trading, Brian Johnson developed this inventive approach
specifically to design and manage option earnings strategies.
These revolutionary new tools can be applied to any option earnings
strategy on any underlying security. In an Active Trader article titled
“Modeling Implied Volatility,” Mr. Johnson introduced a formula for
aggregating discrete volatility measures into a single metric that can be used
with conventional option pricing formulas to accurately model implied
volatility before and after earnings announcements. The practical application
of this formula has profound implications for option trading and strategy
development.
Exploiting Earnings Volatility is written in a clear, understandable fashion
and explains how to use this revolutionary approach to 1) solve for the
expected level of earnings volatility implicitly priced in an option matrix, 2)
calculate historical levels of realized and implied earnings volatility, 3)
develop strategies to exploit divergences between the two, and 4) calculate
expected future levels of implied volatility before and after earnings
announcements.
Furthermore, Exploiting Earnings Volatility also includes two Excel
spreadsheets. The Basic spreadsheet employs minimal input data to estimate
current and historical earnings volatility and utilizes those estimates to
forecast future levels of implied volatility around earnings announcements.
The Integrated spreadsheet includes a comprehensive volatility model that
simultaneously integrates and quantifies every component of real-world
implied volatility, including earnings volatility. This powerful tool allows the
reader to identify the precise level of over or undervaluation of every option
in the matrix and to accurately forecast future option prices and option
strategy profits and losses before and after earnings announcements.
The Integrated spreadsheet even includes an optimization tool designed to
identify the option strategy with the highest level of return per unit of risk,
based on the user’s specific assumptions.
Written specifically for investors who have familiarity with options, this
practical guide begins with a detailed review of volatility, the single most
important (and often misunderstood) component of option pricing. The
aggregate implied volatility formula is fully explained in Chapter 2, which
also includes graphical examples to communicate the mathematical
relationships visually and intuitively.
Chapter 3 includes a detailed step-by-step guide to using the Basic
spreadsheet to calculate historical and implied levels of earnings volatility
and to forecast future levels of implied volatility before and after earnings
announcements.
The next chapter provides a conceptual and mathematical explanation of
“True Greeks,” accurate measures of risk and return sensitivity that reflect the
real-world behavior of options. New option Greeks that are specific to
earnings announcements are also introduced in this chapter. The Integrated
spreadsheet calculates “True Greek” values for individual options and for
option strategies.
Building on the above foundation, Chapter 5 includes a trade example that
uses actual market data and analytical results from both spreadsheets to
design a unique option strategy to exploit earnings-related pricing and
volatility anomalies.
After illustrating the capabilities of the spreadsheets in a real-world trade
example, Chapters 6 through 8 explain every module of the Integrated
spreadsheet tool and how to best use this spreadsheet in practice: how to
import the data, enter user specifications, solve for volatility parameters, and
optimize and evaluate option strategies. The data for a single underlying
security and its option matrix with actual prices are used in both spreadsheets
and throughout these chapters to demonstrate the process traders would
employ in a real-world environment.
The next chapter includes another real-world trading example with actual
market data, which illustrates how the analytical tools in this book can be
used in different market environments. The final chapter examines practical
considerations and prospective applications of these innovative new tools.
This book introduces a new analytical framework that may sound
complicated at first, but is really quite intuitive. Formulas are provided to
ensure an accurate mathematical description of the analytical framework. The
formulas presented in the book are limited to basic high-school algebra, so
they should be accessible to most readers. However, for those readers who
still have nightmares about high-school algebra, important mathematical
relationships are also explained intuitively and depicted graphically.
Most important, you will not need to perform any of these calculations
manually. Exploiting Earnings Volatility includes a link to Excel
spreadsheets that perform all of the calculations described in the book. In
addition, the Integrated spreadsheet functions are all automated and
accessible via
“push-button” macros.
All of the formulas in the book are presented in Excel format to make it
easier for readers who would like to experiment with these tools in their own
Excel spreadsheets.
The unique price and volatility behavior of options before and after
discrete earnings announcements is an enigma to most option traders, even to
many professional option traders. The aggregate volatility formula is
relatively simple, but it has profound implications. When integrated with a
real-world volatility model, it offers unparalleled insights into earnings
volatility, price behavior, option strategy construction, and prospective value-
added opportunities.
 Tips on Viewing Images

The ability to examine the graphs and tables in this book is essential to
understanding the material. Depending on the screen size of your E-reader, it
may be difficult to see all of the detail on the images. Many of the images
have a landscape orientation, but most users prefer to use the portrait
orientation when reading. As a result, many of the images will be compressed
to fit the narrower width of the screen.
The larger images all appear at the top of a new page. The easiest way to
enlarge these images is to simply rotate your E-reader 90 degrees from
portrait to landscape orientation. However, the screen orientation on your E-
reader must first be unlocked.
If you would like to examine the graphs and tables in even greater detail,
first select the image (for example, tap twice on the image with the Kindle
Fire). This would bring up the graphic on a separate screen. You would then
be able to zoom in on the graph or table. Close the image screen when you
are finished and continue reading.
 1 - Volatility Review

Option pricing before and after earnings announcements is highly inefficient

and the mechanics of calculating the corresponding implied volatility values
is a mystery to most market participants. As a result, for traders with the
proper tools and an understanding of earnings volatility, option earnings
strategies represent one of the best and most reliable sources of value-added
opportunities in the option market. Even better, these opportunities occur
every quarter and there are hundreds of individual stocks that have options
with sufficient liquidity to exploit these opportunities.
Until now, there has never been a consistent, objective framework for
evaluating and constructing option earnings strategies that simultaneously
models every aspect of volatility, including earnings effects. This book will
provide the tools you need to objectively evaluate and develop option
earnings strategies on any underlying stock, in any market environment.
However, before I introduce these new tools, we need to examine volatility
and how volatility affects option values. The key to understanding option
valuation is the asymmetrical payoff function. Portions of the following
crucial section also appeared in my first book: Option Strategy Risk / Return
Ratios.

Asymmetrical Payoff Functions

Call and put options have asymmetrical payoff functions. This sounds
complicated, but is actually relatively straightforward. A call option gives the
owner or buyer the right, but not the obligation, to purchase the underlying
asset at the strike price on or before the expiration date. For now, let’s keep
things simple and ignore the fact that many options can be exercised prior to
expiration. This will allow us to focus our attention on what happens at
option expiration.
The following example (depicted in Figure 1.1 below) should help
illustrate the concept of asymmetry. If we purchased a one-year call option on
IBM with a strike price of $100, we would only choose to exercise the call
option if the price of IBM was above the $100 strike price (in the money) on
the expiration date.
If the price of IBM was $110 on the expiration date, the payoff would be
$10. The payoff is also called the intrinsic value and represents the value of
exercising an in-the-money option at expiration. In this case, we could
purchase IBM for the $100 strike price and immediately sell it at the market
price of $110 for a payoff of $10. If the market price of IBM was $120 on the
expiration date, the payoff would be $20. For every dollar the price of IBM
rose above the strike price of $100, the payoff of the call option would
increase by $1. As a result, the slope of the payoff function above the strike
price is positive 1.0 (one dollar increase in payoff for every one dollar
increase in the price of IBM).
If the price of IBM was below the $100 strike price (out of the money) on
the expiration date, we would choose not to exercise the option and it would
expire worthless. In that scenario, the payoff would be zero – although we
would incur a loss on the trade.
It is important not to confuse payoffs and profits. Payoff functions (not
profit and loss functions) should be used to determine the value of options. It
would not matter how much the price of IBM dropped below $100; the call
option would expire worthless and the payoff would still be zero. All options
that are out of the money on the expiration date expire worthless and
therefore have an intrinsic value and payoff of zero. As a result, the slope of
the payoff function below the strike price is zero (zero change in the payoff
function for a one dollar increase in the price of the underlying security).
 Figure 1.1: Call Option Payoff Diagram

Note the discrete change in the slope of the payoff function that occurs at
the strike price. The slope of the yellow payoff function is zero when the
price of the underlying stock (IBM) is below the strike price ($100) and the
slope of the payoff function line is plus 1.0 when the price of the stock is
above the strike price. This payoff function is asymmetric and this
asymmetry creates value for the call option.

Volatility, Asymmetry, & Option Values

The value of an option represents the present value of its probability-

weighted future payoffs. What does that mean? Let’s use the IBM call option
payoff function from Figure 1.1 again to work through a simple example.
First, let’s assume that the stock price of IBM today is $100 and that one year
from today, there were only two possible states of the world: the price of
IBM would either increase by 10% (+ $10) or the price of IBM would
decrease by 10% (- $10). In other words, one year from today there would be
a 50% probability of IBM closing at $90 and a 50% probability of IBM
closing at $110.
The payoffs in those two hypothetical scenarios would be $0 and $10,
respectively (see yellow boxes in Figure 1.1 above). Given that the two
possible payoffs of $0 and $10 both had a 50% probability of occurring, the
average payoff one year from today would be $5 = [(50% x $0) + (50% x
$10)]. To determine the value of the call option today, we would technically
need to discount the probability-weighted payoff of $5 back to the present
using a proxy for the risk-free interest rate.
However, as I write this, short-term interest rates are approximately zero
and have been near zero for several years. Even more important, discounting
the expected future payoff would complicate our example unnecessarily and
shift our focus away from our primary objective: understanding the effects of
asymmetric payoff functions and volatility on option values.
To summarize the hypothetical option valuation example, IBM is
currently trading at $100. If there was a 50% probability of IBM increasing
or decreasing by 10% in one year, the value of a one-year call option with a
strike price of $100 would be $5 (ignoring discounting).
 Now let’s calculate the value of the same call option holding all values
constant, except for the expected level of volatility. Let’s assume the
expected level of volatility for the next year increases from 10% to 20%. The
stock price of IBM today would still be $100 and one year from today, there
would only be two possible states of the world: the price of IBM would either
increase by 20% (+ $20) or the price of IBM would decrease by 20% (- $20).
One year from today there would now be a 50% probability of IBM closing
at $80 and a 50% probability of IBM closing at $120.
The payoffs in those two hypothetical scenarios would now be $0 and
$20, respectively. Given that the two possible payoffs of $0 and $20 both had
a 50% probability of occurring, the average payoff one year from today
would now be $10 = [(50% x $0) + (50% x $20)].
As is evident in this simple example, option prices are a direct function of
the expected level of volatility. As the expected level of volatility increases,
the values of both call and put options also increase. The reason is
asymmetry. Increasing volatility magnifies the value of asymmetry.
Decreasing volatility diminishes the value of asymmetry.

Volatility Assumptions & Conventions

Volatility is a measure of price dispersion and is typically expressed as a

one standard deviation (SD) annualized percentage price change in the
underlying security based on a log-normal return distribution. Figure 1.2
depicts the normal and log-normal price distributions of a $100 non-dividend
paying stock, one year into the future. The assumed risk free interest rate was
0.25% and the annual standard deviation was 25%. The future stock price is
shown on the independent x-axis and the probability is shown on the
dependent y-axis. The percentages were calculated in increments of $1 and
Microsoft Excel connected the probabilities using a smooth curve. The
normal distribution is represented by the dashed purple line and the log-
normal distribution is depicted by the solid blue line.
First, note that the normal distribution is symmetric and is centered
(approximately) at the current stock price of $100. The symmetry of a normal
distribution leads to a very significant problem when valuing options:
negative prices for the underlying security. In practice, there is no upper
bound on stock prices (provided there are investors who are willing to pay
more for the stock), but stock prices cannot drop below zero. As a result, it is
not practical to use normal distributions when pricing options.
The Black Scholes Option Pricing Model (BSOPM) assumes that the
returns of the underlying security are distributed log-normally, which means
that the natural log of the ending price divided by the beginning price (ending
price/beginning price) are distributed normally. Prices derived from a log-
normal distribution cannot drop below zero, but are not limited in how much
they can increase. The resulting log-normal distribution in Figure 1.2 appears
to be skewed to the right: the left tail of the distribution looks truncated
relative to the right tail of the distribution.
 This would seem to imply that the expected future price of the underlying
security is higher for the log-normal distribution than for the normal
distribution, but that is not the case. Notice that the center of the log-normal
distribution is shifted slightly to the left – just enough to ensure the expected
price change of the two distributions are identical.
That brings up an interesting question that few option traders fully
appreciate: what is the expected return of the underlying security under the
BSOPM assumptions? The expected return of the underlying security is equal
to the rate of return earned by investing at the risk-free interest rate, which is
the same rate used to discount future payoffs when valuing call and put
options. As a result, the expected returns from buying call and put options are
also equal to the rate of return earned by investing at the risk-free interest
rate. This implies that investors are risk-neutral and that expected returns are
not a function of risk.
As I write this, one-year US T-bill rates are only 0.15%, which indicates
that expected returns for underlying securities (and calls and puts) under
BSOPM assumptions are effectively zero in the current environment. That
means that option prices implicitly assume an expected return of
approximately zero for all stocks on earnings announcement dates. This
assumption will become very important when we begin designing directional
option strategies in later chapters.
The above return distributions were calculated for a one-year holding
period. However, there are obviously many different options, all with varying
amounts of time remaining until expiration. The industry convention is to
express volatility in annual terms, regardless of the time remaining until
expiration for a given option.
This yields two important benefits. First, the annualized volatility
estimate can be used directly in the BSOPM and binomial option models.
Even more important, the resulting annualized volatilities are directly
comparable across all variables: time, underlying security, strike price, time
to expiration, option type, etc. This further elevates the importance of
volatility relative to other option metrics.

Volatility versus Time

Just because volatility is expressed in annual terms does not mean that it
cannot be used to calculate expected price changes for other periods.
Volatility is a function of the square root of time. The following formula can
be used to calculate the expected future price of an underlying security under
the simplifying assumptions of a 0% risk-free rate and a 0% dividend yield:

ST = S0 x EXP(N * σ * (TD/252)^0.5)

ST = Stock Price T years into the future

S0 = Stock Price today (Zero years into the future)
EXP represents the Excel function e raised to a power, e =
2.718281828
N = Number of standard deviations (can be positive or negative)
σ = Annualized volatility
TD = Number of trade days remaining
252 = Approximate number of trading days in calendar year

You will note that the number of trade days (TD) was used in the above
formula, instead of the number of calendar days. All of the formulas in this
book use trade days, as do all of the spreadsheets. The improved precision of
trade days is required to accurately quantify the impact of earnings volatility.
Now let’s use the above formula to forecast a hypothetical negative one
standard deviation price for a $100 stock, 42 trading days in the future,
assuming an annual volatility of 25%.

ST = S0 x EXP(N * σ * (TD/252)^0.5)
ST = 100 x EXP(-1 * 0.25 * (42/252)^0.5)
ST = 100 x EXP(-1 * 0.25 * (0.408248))
ST = 100 x EXP(-0.102062) = 90.29

The resulting price of $90.29 represents a 9.71% decline from the initial
stock price of $100. If we repeat this calculation for a range of time periods
(measured in trade days), we can construct the graph in Figure 1.3, which
depicts the positive and negative one standard deviation percentage price
changes as a function of the number of trading days. The initial price of the
non-dividend paying stock was $100; the assumed risk free interest rate was
0.00% and the annual standard deviation was 25%.
 The number of trading days is shown on the independent x-axis and the
percentage price change is shown on the dependent y-axis. The negative one
SD percentage price changes are represented by the dashed red line and the
positive one SD percentage price changes are depicted by the solid green line.
The red circle on the dashed red line represents the percentage price change
for the hypothetical negative one SD move over 42 trade days (approximately
two months).
 The ability to forecast expected price changes for a given level of
annualized volatility is a critical element of option trading and is especially
crucial for option earning strategies. It should be one of the main tools in
your option trading toolbox. I use this formula on a daily basis.

Constant and Continuous

The BSOPM assumes that volatility is constant. Unfortunately, this is not

true in practice. In reality, market participants typically assume that volatility
will increase or decrease over time and these assumptions are reflected in the
term structure of volatility, which is also called the horizontal skew. Please
refer to Chapter 7 for a more comprehensive discussion of modeling the
effects of the horizontal skew.
Even for options with the same expiration date, volatility is not constant.
Instead, the expected level of volatility varies as a function of the strike price
of the option. The relationship between volatility and strike prices is called
the vertical skew. Please refer to Chapter 7 for a more comprehensive
discussion of modeling the effects of the vertical skew. The vertical skew
exists because the log-normal distribution does not accurately reflect the
market’s probability expectations with respect to the underlying security’s
future price changes.
The BSOPM also assumes that price changes are continuous, which
means that we should never observe large discrete changes in the price of the
underlying security. This assumption is obviously violated in practice,
particularly when quarterly earnings are announced.
Despite all of the invalid assumptions, the BSOPM can be used in
practice, but volatility modeling is required to quantify and correct the effects
of every invalid assumption. The effects of the vertical and horizontal skews
are modeled in the Integrated spreadsheet that accompanies this book. The
analytical framework for integrating discrete price changes from earnings
announcements into the BSOPM’s volatility assumptions will be introduced
in the next chapter and are also modeled in both the Basic and Integrated
spreadsheets that are included with this book.

Implied Volatility (IV)

 Until now, I have ignored the fact that the term “volatility” can be used to
describe several different concepts, all of which are relevant to this book. The
first is implied volatility, which is a measure of expected or forward-looking
volatility.
Option values are determined by only a handful of variables: the stock
price, the strike price, the time to expiration, the risk-free interest rate, and
the expected level of volatility (and dividends, which we will ignore for
now). What is unique about the expected level of volatility relative to the
other variables?
Do we know an option’s current stock price with certainty? Yes, it is
observable on our broker platforms. What about the strike price and time to
expiration? Certainly, both are used to describe the option. Do we know the
risk-free interest rate that corresponds to the time to expiration of the option?
Well, this confuses some traders, but US Treasury Bill yields are typically
used as a proxy for the risk-free interest rate. US T-Bills are highly liquid and
their yields are also quoted on broker platforms in real time.
Do we know the expected level of volatility? Unfortunately, we do not –
at least not directly. However, since we know the values of the other
variables that are required to calculate the value of an option and we also
know the market price of the option, we can use an option valuation formula,
such as the Black Scholes Option Pricing Model (BSOPM), to solve for the
implied volatility (IV).
Do you see why this is called implied volatility? Implied volatility is the
expected level of volatility implied by or embedded in the price of the option.
In other words, IV represents the market’s estimate of the future level of
volatility of the underlying security, as determined by the price of the option.
If we know the option price, we can solve for the corresponding implied
volatility. Similarly, if we know the expected level of volatility, we can solve
for the price of the option. The relationship between the price of an option
and the expected level of volatility of the underlying security is completely
deterministic, but it does require an option valuation model, which relies on a
given set of assumptions.
To keep things simple, we will focus on the BSOPM in this book, but we
will include dividend adjustments for practical purposes. We will also discuss
the binomial model, which is a discrete version of the BSOPM; properly
constructed, the binomial model’s option values converge to the BSOPM’s
option values as the number of intervals used in the binomial model
increases. The spreadsheets that accompany this book also use the BSOPM
and binomial models, but are not limited by their unrealistic and simplistic
volatility assumptions.
Now, let’s find the implied volatility of the hypothetical one-year IBM
call option that we explored earlier in this chapter (depicted in Figure 1.1).
You will recall that the stock and strike prices were both $100, the time to
expiration was one year, and the assumed risk-free interest rate was 0%. If
the value of the call option was $5, what would be the implied volatility? You
will recall from the earlier valuation example that an expected level of future
volatility of 10% resulted in a call option value of $5 = [(50% x $0) + (50% x
$10)]. As a result, an option price of $5 would imply an expected level of
volatility of 10%.
If the value of the call option was $10, what would be the implied
volatility? In the earlier example, we found that an expected level of future
volatility of 20% resulted in a call option value of $10 = [(50% x $0) + (50%
x $20)]. Therefore, an option price of $10 would imply an expected level of
volatility of 20%.
Remember that we need to implicitly or explicitly use an option valuation
model to calculate implied volatility. In this example, you should note that we
did not use BSOPM to solve for implied volatility. Instead, we used the very
simple symmetric binomial (two-state) model described earlier that had only
two possible future stock prices, both of which were a function of the current
stock price and the expected future level of volatility. If we had used the
BSOPM, we would have arrived at a different value for implied volatility.
Implied volatility is arguably the most important option pricing variable;
it is synonymous with the price of an option, but unlike price, it is directly
comparable across time, underlying securities, option types, strike prices, and
expiration dates. It is impossible to evaluate option prices without modeling
implied volatility.
However, as explained earlier, using the BSOPM to calculate implied
volatility assumes that volatility is constant and that price changes in the
underlying security are continuous. Instead, due to the periodic release of
important quarterly earnings data and forward guidance, stocks experience
large discrete price changes after every quarterly earnings announcement.
Modeling the expected price dispersion due to earnings announcements
requires a new type of implied volatility called implied earnings volatility.
Instead of using option prices to solve for implied volatility, the analytical
framework presented in the following chapter will formally document the
mathematical relationship between implied volatility, “normal” (or non-
earnings) volatility, and earnings volatility.
This formula will allow us to use option prices and the corresponding
implied volatilities to solve for each option’s normal implied volatility and
the overall level of implied earnings volatility for all options on a given
stock. This will provide the foundation for an intuitive new approach for
designing option strategies to exploit earnings volatility.

Historical Volatility

The second volatility concept that we will need is historical volatility,

which is a measure of past or backward-looking volatility. Practitioners use
past levels of price volatility as a means of forecasting future levels of price
volatility, which are then used to forecast option prices. To be useful as an
estimate of implied volatility, historical volatility must be expressed in the
same units as implied volatility. In other words, historical volatility must
represent an annualized volatility of the historical log returns of the
underlying security.
To calculate the annualized historical volatility of a return series,
annualize the root mean squared daily log returns of the underlying security.
To calculate the annualized historical volatility:
1. Calculate the ratio of the daily closing prices (closing price/previous
closing price).
2. Calculate the natural log of the daily price ratios (from step 1).
3. Square the log returns (from step 2).
4. Calculate the average of the squared log returns (from step 3).
5. Calculate the square root of the average of the squared log returns (from
step 4).
6. Annualize the root mean squared (RMS) log returns (from step 5).
While I did use the term standard deviation before, you should note that
we did not and should not use the formula for standard deviation to calculate
historical volatility. Standard deviation measures the deviation around the
mean or average value of a data series, while the root mean squared return
measures the deviation around zero.
You will recall that the expected daily return for all assets is
approximately zero under the BSOPM assumptions, which is consistent with
the root mean squared (RMS) calculation. The standard deviation calculation
implicitly assumes that the expected daily return equals the average daily
return, which can significantly understate volatility, especially for small data
sets.
The hypothetical example in Figure 1.4 illustrates the potential magnitude
of the problem of using standard deviation as a measure of historical
volatility. The twelve daily returns are listed in the first column, all of which
equal 2.02%. The natural log of one plus the daily returns is provided in the
next column, followed by the squared log returns in column three.
Each of the log returns equals 2.00%, which means that the mean or
arithmetic average of the log returns also equals 2.00%. Therefore, the daily
deviations from the mean must all equal zero, as does the standard deviation
(SD) around the mean. Using the standard deviation formula, we would have
calculated a historical volatility of 0.00% for a security that was moving in
price by 2.02% per day.
 Unlike the standard deviation formula, the root mean squared (RMS)
formula provides an accurate and representative historical volatility measure
and is not affected by the average level of volatility. The daily RMS for the
hypothetical example was 2.00% (steps 1-6). You will recall that volatility is
a function of the square root of time. We can use this fact to calculate an
annualized historical volatility from the daily historical volatility.
If we assume there are 252 trade days per calendar year, we multiply the
daily volatility value of 2% by the square root of 252 to arrive at the
annualized historical volatility (see formula below).

Annual Volatility = Periodic Volatility * ((252/TD) ^ 0.5)

TD = Number of trade days used to calculate the periodic log return
series

Annual Volatility = Periodic Volatility * ((252/TD) ^ 0.5)

Annual Volatility = 2.00% * ((252/1) ^ 0.5) = 31.75%

If we determined that the resulting historical volatility of 31.75% was a

fair and unbiased estimate of future volatility, we could use the value of
31.75% as a forecast of the future level of annualized volatility in the
BSOPM to estimate the value of an option on the underlying security.
As we saw with implied volatility, there is also a corresponding measure
of historical volatility due to earnings announcements, which is called
historical earnings volatility. Instead of performing the RMS calculation on a
series of sequential daily returns, we use the RMS formula on the daily
returns history from past earnings announcements.
Specifically, we calculate the historical daily earnings returns using the
closing price for the first trade day immediately following the earnings
announcement and the closing price for the previous trade day, which will
immediately precede the earnings announcement. In other words, we use the
closing prices immediately before and after the earnings announcement to
calculate historical returns due to earnings. The resulting historical earnings
returns will still be daily returns, because they will always be measured using
closing prices over a span of one trading day.

Realized volatility

Using the price of an option in conjunction with an options valuation

model (such as the BSOPM), we can calculate the implied volatility of the
option, which represents the market’s estimate of future price volatility. We
can also calculate historical volatility, which can also be used to estimate
future price volatility. Why would we need two estimates of volatility?
Because the market’s estimate of future volatility is frequently wrong.
If we could come up with a more accurate estimate of future volatility
than the market’s implied volatility estimate, then we could design option
strategies to exploit our insight and earn excess returns. We will look at
actual strategies later, but for now, let’s assume that the market’s annualized
implied volatility estimate was only 50%, but the annualized historical
volatility was 100%.
 In this hypothetical example, it would appear that the market’s implied
volatility estimate is too low. Since volatility is synonymous with price,
option prices would also be too low. In other words, if we believe our
annualized future volatility estimate of 100% is more accurate than the
market’s implied volatility estimate of 50%, then options must be cheap or
undervalued and should be purchased.
How could we profit from such a trade? Let’s ignore directional price
changes and focus on volatility. If we purchased undervalued options at the
market’s implied volatility of 50% and the market’s implied volatility
estimate of 50% converged to our future volatility estimate of 100%, then the
prices of our options would increase and we would earn an excess return.
What if the market’s implied volatility estimate (50% annualized) never
changed, but our expected future level of volatility (100% annualized) proved
to be correct? What would this mean and how would it affect our strategy? It
would mean that we were ultimately correct and the realized volatility
experienced over the holding period (the time we held the strategy) was equal
to our future volatility forecast, which was based on historical volatility.
In this scenario, we would expect to earn an excess return because the
realized volatility experienced over the holding period was greater than the
market’s implied volatility, despite the fact that the market’s estimate of
implied volatility did not change. Why did we earn an excess return? Because
option payoff functions are asymmetric and the undervalued options we
purchased benefited from a higher than expected level of volatility.
Since our annualized volatility estimates were derived from log returns, if
we want to compare the realized volatility to the annualized levels of
historical or implied volatility, we need to calculate the log of the realized
volatility as well. Specifically, we need to calculate the natural log of the
realized ending price divided by the realized beginning price (realized ending
price/realized beginning price). However, the resulting realized volatility
would not be annualized; it would be unique to the length of the holding
period. How do calculate the expected level of future volatility for a specific
holding period? We can use the following formula to convert an annualized
volatility to the expected volatility for a holding period of any length.

Holding Period Volatility = Annualized Volatility * ((HPD/252) ^ 0.5)

HPD = Number of trade days in holding period
 If we assume a holding period of ten trading days and an annualized
volatility of 100%, then the expected level of volatility for ten days would be:

Holding Period Volatility = Annualized Volatility * ((HPD/252) ^ 0.5)

Holding Period Volatility = 100% * ((10/252) ^ 0.5)
Holding Period Volatility = 100% * (0.1992) = 19.92%

If we assume a holding period of 10 trade days and an annualized

volatility of only 50%, then the expected level of volatility for 10 days would
be:

Holding Period Volatility = Annualized Volatility * ((HPD/252) ^ 0.5)

Holding Period Volatility = 50% * ((10/252) ^ 0.5)
Holding Period Volatility = 50% * (0.1992) = 9.96%

In the continuation of our earlier example above, the market was

implicitly expecting a price movement of 9.96% over the holding period,
which corresponds to an annualized implied volatility of 50%. We assumed
that the realized volatility would instead be consistent with the annualized
historical volatility of 100%, which would result in a realized price change of
19.92%, far exceeding the market’s estimate, which would have allowed us
to profit from buying options at too low a price.
To make this example more accessible, I avoided a discussion of the
Greeks, which will be explained more fully in a later chapter. However, for
those of you with a working knowledge of the Greeks, when implied
volatility is too low (50%) and we expect implied volatility to increase and/or
we expect realized volatility to exceed implied volatility, we would want to
construct a strategy with positive Vega to profit from the expected increase in
implied volatility and positive Gamma to benefit from a higher than expected
level of realized volatility.
Vega represents the change in the value of an option or option strategy for
an instantaneous 1% increase in implied volatility, holding all of the other
variables constant. Gamma reflects the curvature of the option price function.
More specifically, Gamma equals the change in Delta for an instantaneous $1
increase in the price of the underlying security, holding all of the other
variables constant. While that is the definition of Gamma, it is not very
intuitive, which is why I refer to the curvature or convexity of the option
price function instead.
All long option positions have positive Gamma, which means they benefit
from realized changes in the price of the underlying security. What does
“long” mean? It means the investor purchased the option, which gives the
buyer the right but not the obligation to exercise the option. Conversely, a
short position means the investor sold the option (without previously owning
it), which means the seller has the obligation to sell (call) or buy (put) the
underlying security at the specified strike price. Why do long option positions
benefit from realized volatility? Because of asymmetry and asymmetry is the
source of positive Gamma.
As was the case for the other volatility concepts presented earlier, there is
an earnings component to realized volatility as well, which we will call
realized earnings volatility. Since earnings are typically announced after the
market closes or before the market opens, we will calculate realized earnings
volatility for the one day holding period beginning at the close immediately
before the earnings announcement and ending at the close on the following
trade day. As was the case with historical earnings volatility, realized
earnings volatility will always be calculated over a period of one trading day.
Let’s repeat the realized volatility calculation above for a one-day holding
period associated with a single earnings announcement. If we had an
annualized historical earnings volatility of 100%, the corresponding level of
earnings volatility over a one-day holding period would be:

Holding Period Volatility = Annualized Volatility * ((HPD/252) ^ 0.5)

Holding Period Volatility = 100% * ((1/252) ^ 0.5)
Holding Period Volatility = 100% * (0.0630) = 6.30%

If we had an annualized implied earnings volatility of only 50%, the

corresponding level of earnings volatility over a one-day holding period
would be:

Holding Period Volatility = Annualized Volatility * ((HPD/252) ^ 0.5)

Holding Period Volatility = 50% * ((1/252) ^ 0.5)
Holding Period Volatility = 50% * (0.0630) = 3.15%

As you can see from this example, it does not matter whether we are
talking about volatility or earnings volatility. For every volatility concept,
there is a corresponding earnings volatility counterpart.
However, to exploit anomalies in earnings volatility, we first need the
tools to calculate implied, historical, and realized earnings volatility. The
following chapter will explain the mathematical relationship between implied
volatility, normal volatility, and earnings volatility. This intuitive formula
will lay the groundwork for innovative new spreadsheet tools that will offer
unparalleled insights into option earnings strategies that can be optimized to
provide the highest level of expected returns per unit of risk.
 2 – The Aggregate IV Formula

For many years, I avoided trading options on individual stocks and focused
almost exclusively on index and futures options. There were two principal
reasons for my reluctance to trade stock options: 1) I did not feel that I had
any advantage in implementing option strategies on individual stocks and 2) I
did not have the requisite tools to accurately evaluate and model stock option
strategies around earnings announcements.
It frustrated me to know that I was missing out on the profit opportunities
from an entire class of options, so several years ago I researched the subject
further and eventually derived the mathematical formula that precisely
quantifies the behavior of implied volatility (IV) before and after earnings
announcements. This breakthrough provided the market edge that I was
lacking in individual stocks and formed the foundation for the toolset needed
to calculate option pricing and true-risk metrics for stock options that
incorporated earnings effects.
I first introduced the formula in my June 2011 Active Trader article titled
“Modeling Implied Volatility.” The article explained that implied volatility
on individual stocks can be broken down into two components: earnings
volatility and “normal” or non-earnings volatility. The aggregate implied
volatility formula provides a means of correctly combining both components
into a single, precise aggregate implied volatility value that can be used in
conventional option valuation models. Unfortunately, while the aggregate
implied volatility formulas are accurate, the approach I outlined in the Active
Trader article for modeling “normal” volatility was not precise enough to use
for strategy construction purposes.
Over the past few years, I designed a new comprehensive volatility model
that can be used in conjunction with the aggregate implied volatility formula
to simultaneously model both normal and earnings volatility for every option
in a matrix. The comprehensive volatility model and attendant strategy
optimization and analysis tools are included in the Integrated spreadsheet that
accompanies this book.
The aggregate implied volatility formula will be explained fully in this
chapter using hypothetical examples. In addition, several graphs will be used
to help you visualize the mathematical relationships and to provide a more
intuitive understanding of the behavior or implied volatility around earnings
announcements. The remainder of the book will explain how to apply the
formula in practice using real-world strategies with actual market data.
The aggregate implied volatility formula that appears in this chapter is
mathematically equivalent to the original aggregate implied volatility formula
that appeared in the 2011 Active Trader article, but I rearranged a few of the
terms to make the formula more intuitive. In this chapter, I will present three
algebraically equivalent versions of the formula that solve for aggregate
implied volatility, earnings volatility, and normal volatility, respectively.
Finally, I will include numerical examples, for those of you who would like
to verify that you are applying the formula correctly.
For this example, we will assume that the annualized earnings volatility
equals 180% and the annualized normal volatility equals 30%. Why the huge
disparity between earnings volatility and normal volatility? As I explained
earlier, the BSOPM assumes that price changes are continuous, which
implies that the distribution of information that affects stock prices is also
released continuously. This is obviously not the case and one of the most
glaring exceptions is periodic earnings announcements.
Public companies report earnings quarterly. When reporting earnings,
senior corporate executives also conduct conference calls where they disclose
new information about the future prospects of the company in the form of
forward guidance. The release of new financial data and the new insights
gained from forward guidance immediately alter investors’ expectations
regarding future earnings, growth rates, and ultimately value, which can
create large discrete changes in the price of the stock.
In the following calculation examples, we will also assume there is
exactly one (1) earnings trading day and eleven (11) normal trading days
remaining until expiration of the option. Therefore, the option would have a
total of twelve (12) trading days remaining until expiration. We will use the
first and most intuitive version of the formula to solve for the aggregate
implied volatility.
I use the term aggregate implied volatility because it aggregates earnings
volatility and normal volatility into a single implied volatility value.
Aggregate implied volatility is synonymous with implied volatility, which
means that it is directly comparable to the implied volatility calculated by
your broker or research platform and it can be used in conventional option
pricing and risk models, including the BSOPM.
Intuitively, the formula simply asserts that the aggregate implied volatility
is a weighted average of earnings and normal volatility and the respective
weights are proportional to the number of earnings and normal trading days
remaining until expiration of the option. If we know any two of the three
volatility terms, we can solve for the third.
The first section of the formula below calculates the weights and the
second section uses those weights to solve for the aggregate implied
volatility.

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WE = Earnings Volatility Weight
WN = Normal Volatility Weight

WE = NTDE/( NTDE + NTDN)

WE = 1/( 1 + 11) = 8.3333%

WN = NTDN/( NTDE + NTDN)

WN = 11/( 1 + 11) = 91.6667%

By definition, the weights will sum to 100%. The aggregate implied

volatility is completely explained by earnings and normal volatility, and the
sum of the earnings and normal trading days remaining will always equal the
total number of trading days remaining until expiration.
As a result, the weight assigned to earnings volatility equals the number
of earnings trading days (1) divided by the total number of trading days
remaining until expiration (12). Similarly, the weight assigned to normal
volatility equals the number of normal or non-earnings trading days (11)
divided by the total number of trading days (12).

IV = Aggregate Implied Volatility (Annualized)

VE = Earnings Volatility (Annualized)
VN = Normal Volatility (Annualized)

IV = ((WE*(VE^2))+(WN)*(VN^2))^0.5
IV = ((8.3333%*(180%^2))+(91.6667%)*(30%^2))^0.5
 IV = (35.25%)^0.5
IV = 59.37%

I stated above that aggregate implied volatility is a weighted average of

earnings and normal volatility, but that was an oversimplification. As you can
see from the above formula, the squared aggregate implied volatility is a
weighted average of the squared earnings volatility and the squared normal
volatility.
To find the squared aggregate implied volatility, we square the earnings
volatility (180%^2) and multiply it by the earnings weight (8.3333%). We
then square the normal volatility (30%^2) and multiply it by the normal
weight (91.6667%) and add both volatility terms together. Finally, we
calculate the square root of the squared aggregate implied volatility (35.25%)
to find the aggregate implied volatility (59.37%).
So, what does this mean? It means that if the annualized earnings
volatility was 180% and annualized normal volatility was 30%, the aggregate
implied volatility would equal 59.37%. It is important to note that the
aggregate implied volatility formula is not an approximation. If we knew the
earnings and normal volatility with certainty (as we do the number of
earnings and normal trading days) and the earnings and normal volatilities
conformed to the assumed BSOPM log-normal distribution, the aggregate
implied volatility formula would yield the exact value of implied volatility
derived from option pricing theory (for European options).
Technically, changes in the optimal path-dependent early exercise of
American options would not be precisely captured by the aggregate implied
volatility formula, but in practice, these negligible deviations would be
swamped by the magnitude of earnings events. The BSOPM does not even
consider the value of optimal early exercise of American options.
I concede that we do not know earnings and normal volatility with
certainty, but I will explain practical ways to estimate those values in later
chapters. For now, let’s focus on the implications of this formula. We can
observe the price of any option, which means that we can also calculate the
real-time implied volatility. If we can estimate the normal or earnings
volatility, we can solve for the remaining volatility term.
We can then use the above formula to determine how the implied volatility
of any option will change as time passes. In other words, the formula allows
us to use our estimates of earnings and normal volatility to calculate the
future values of implied volatility for any option as the option ages and the
number of trading days remaining until expiration declines.
Before we calculate future implied volatility values, we need to review
the alternative versions of the aggregate implied volatility formula that allow
us to solve for earnings volatility and normal volatility. The earnings and
normal volatility weight calculations are the same for all versions of the
formula, so the weight formulas and calculations were not repeated below.

IV = Aggregate Implied Volatility (Annualized)

VE = Earnings Volatility (Annualized)
VN = Normal Volatility (Annualized)

VE = (((IV^2)-(WN*(VN^2)))/WE)^0.5
VE = (((59.37%^2)-(91.6667%*(30%^2)))/8.3333%)^0.5
VE = (324%)^0.5
VE = 180%

VN = (((IV^2)-(WE*(VE^2)))/WN)^0.5
VN = (((59.37%^2)-(8.3333%*(180%^2)))/91.6667%)^0.5
VN = (9%)^0.5
VN = 30%

To find the squared earnings volatility, we first square the implied

volatility (59.37%^2) and subtract the squared normal volatility (30%^2)
multiplied by the normal weight (91.6667%). We divide the difference by the
earnings weight (8.3333%) to find the squared earnings volatility (324%).
Finally, we calculate the square root of the squared earnings volatility (324%)
to find the earnings volatility (180%).
The process for using the implied volatility and earnings volatility to
calculate the normal volatility is the same. If you know any two volatility
terms, you can use the formula to solve for the third. The numerical values in
the preceding formulas were rounded for display purposes, but I used the
actual values in the calculation to avoid loss of precision.

Volatility Weights vs. Time (Single Option)

 Now that we have applied the aggregate implied volatility formula to an
option on a single date, let’s calculate how the volatility weights change as
they approach and pass the date of the earnings announcement. To do so, we
will use the same assumptions from the previous static example: 180%
earnings volatility, 30% normal volatility, one earnings trading day, and
eleven normal trading days. We will assume the values of the volatility
components remain constant, but the number of trading days will decrease as
time passes.
We will also need to know when the earnings announcement will occur.
In this example, earnings will be announced after the market closes, when
two trading days remain until option expiration. The first of those trading
days immediately following the announcement will be the earnings trading
day by definition. The first normal trading session following earnings is when
the new earnings information is processed by the market participants. Their
resulting sales and purchases stimulate the forces of supply and demand
throughout the trading day to arrive at a new post-earnings equilibrium price
for the underlying stock.
The second of the two post-earnings trading days will be a normal trading
day. Let’s use this hypothetical example to evaluate how the earnings and
normal volatility weights would change over time.

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WEN = Earnings Volatility Weight (N Trading Days before Earnings)

WE = NTDE/( NTDE + NTDN)

WE10 = 1/( 1 + 11) = 8.3% (as calculated earlier)
WE9 = 1/( 1 + 10) = 9.1%
WE8 = 1/( 1 + 9) = 10.0%
WE7 = 1/( 1 + 8) = 11.1%
WE6 = 1/( 1 + 7) = 12.5%
WE5 = 1/( 1 + 6) = 14.3%
WE4 = 1/( 1 + 5) = 16.7%
WE3 = 1/( 1 + 4) = 20.0%
 WE2 = 1/( 1 + 3) = 25.0%
WE1 = 1/( 1 + 2) = 33.3%
WE0 = 1/( 1 + 1) = 50.0%
WE-1 = 0/( 0 + 1) = 0.0%

We could use the analogous weight formula to calculate the

corresponding normal volatility weights, but we know the earnings and
normal volatility at a given point in time must add to 100%, so we can simply
subtract the earnings weights from 100%.
WE10 represents the earnings volatility weight ten trading days before
earnings. As explained above, two additional trading days will occur
following earnings and then the option will expire. As a result, ten days
before earnings the option will have a total of twelve trading days remaining
until expiration: one earnings trading day and eleven normal trading days.
We calculated the value of WE10 in the initial static example and that value is
rounded above to one decimal place.
Every day before the release of earnings is a normal or non-earnings
trading day. Therefore, for every day that passes before the release of
earnings, we reduce the number of normal trading days by one day. This also
reduces the number of total trading days by one day. During this period
approaching earnings, the numerator (the number of earnings trading days)
will remain constant and the denominator (the total number of trading days)
will decline by one day for each day that elapses.
This means that the relative weight of earnings volatility increases and the
relative weight of normal volatility decreases. The earnings volatility weight
will increase from the initial value of 8.3% (ten days before earnings) to 50%
(immediately before earnings). During the same period, the normal volatility
weight will decrease from 91.7% to 50%.
The day after earnings are announced, the number of earnings trading
days will decline by one day and the number of normal trading days will
remain constant. As a result, the numerator will decline from a value of one
trading day to a value of zero trading days, which explains why the earnings
volatility weight will fall from 50% to 0% one day after earnings are
announced. The normal volatility weight will obviously increase from 50% to
100%. The earnings and normal volatility weights for this example are
provided in table format in Figure 2.1.
 The same earnings and normal volatility weights are depicted graphically
in Figure 2.2. The independent x-axis represents the number of trading days
before earnings. The dependent y-axis depicts the volatility weights. The
earnings volatility weights are shown by the solid purple line; the dashed blue
line illustrates the normal volatility weights. Remember, in the aggregate
implied volatility formula, the volatility weights are applied to the squared
volatility values, not to the volatility values.
As you can see from the volatility values in Figure 2.2, the change in the
volatility weights over time is not linear. This is due to the daily change in
the denominator of the weight calculation.
 Volatility Weights vs. Time (Multiple Options)

On the date of the earnings announcement, the option in the preceding

example had two days remaining until expiration. We calculated the volatility
weights beginning ten trading days before earnings and ending one day after
earnings. Let’s repeat the same volatility weight example for options with the
following number of trading days remaining after the earnings
announcement: 1, 2, 7, 12, 17, 21, 42, 126, 252, and 504.
The number of trading days for the last three options is consistent with
option expiration dates occurring six months (126 trading days), one year
(252 trading days), and two years (504 trading days) in the future. How
would the weight calculation potentially be different for these three options?
Unlike the options with fewer than 63 trading days remaining until
expiration, the six month, one year, and two year options would all
experience multiple earnings trading days. Two for the six month option,
four for the one year option, and eight for the two year option.
To keep things simple at this stage, the hypothetical volatility weight and
implied volatility examples in this chapter assume one trading day for all
options, regardless of their expiration date. The Basic spreadsheet introduced
in the next chapter uses several simplifying assumptions to reduce the amount
of input data required from the user. As a result, the Basic spreadsheet also
assumes one earnings trading day for all options. The Integrated spreadsheet
requires more data, but is far more sophisticated and powerful than the Basic
spreadsheet. It does not use simplifying assumptions and it uses the actual
number of earnings trading days for every option in the matrix.
The earnings volatility weights for options with a range of trading days
remaining after earnings (1, 2, 7, 12, 17, 21, 42, 126, 252, and 504) are
shown in Figure 2.3. The shaded values in the second column represent the
earnings weights for the same option that we evaluated earlier (Figures 2.1
and 2.2). What conclusions can we draw from the data? The most obvious
result is that the effect of earnings volatility is much greater for shorter
options – those with fewer trading days remaining until expiration. This
makes sense if we think about the formula.

WE = NTDE/( NTDE + NTDN)

 The denominator of the earnings volatility weight formula would be
smaller for shorter-term options, which would increase the relative
importance and weight of earnings volatility. The earnings weight would be
higher and would increase at a more rapid rate for short-term options
relative to long-term options. Given that earnings dates are known in
advance, these weights can be calculated precisely.
As we computed earlier, the earnings volatility weight for the option with
two trading days remaining after earnings increased from 8.3% (ten trading
days before earnings) to 50% (on the announcement date). The corresponding
earnings volatility weight for an option with seven trading days remaining
after earnings would have only increased from 5.9% to 14.3%. This is a
dramatic reduction in the impact of earnings volatility for a very small (5
day) increase in the number of trading days. Given the assumption of one
earnings trading day for all options, the earnings volatility weight for all
options drops to 0% one day after earnings.
The earnings volatility weights for the ten different options described
earlier are presented in Figure 2.3, which assumes one earnings trading day
for all options. Even though we will not relax this assumption until we begin
using the Integrated spreadsheet in later chapters, I provided Figure 2.4,
which will allow you to see the effect of using the actual number of earnings
trading days on the earnings volatility weights for the same ten options.
Finally, the earnings volatility weights for all ten options are depicted
graphically in Figure 2.5, which assumes one earnings trading day for all
options.
In Figure 2.5, the independent x-axis represents the number of trading
days before earnings. The dependent y-axis depicts the earnings volatility
weights. The earnings volatility weights for each option are illustrated by a
different line. They earnings volatility lines appear in the same order as the
legend on the right-side of the graph.
 Aggregate IV vs. Time

In the above examples, we assumed that the earnings volatility and

normal volatility remained constant. This assumption allowed us to focus on
the relative magnitude and pattern of earnings volatility weights for a
selection of options with different expirations as they approached and passed
the earnings announcement date. If we assume earnings and normal volatility
remain constant, implied volatility will be completely determined by the
relative volatility weights.
In practice, implied earnings volatility and implied normal volatility do
not remain constant, but their day-to-day variability will be dominated by the
known pattern of volatility weights approaching earnings, at least for short-
term options.
While the relative volatility weights are a key factor in option earnings
behavior and are also quite intuitive, we are ultimately interested in implied
volatility, which we will need to solve for the value of an option.
As you will recall from the aggregate implied volatility formula (repeated
below), the squared aggregate implied volatility is a weighted average of the
squared earnings volatility and the squared normal volatility. The implied
volatility equals the square root of the squared implied volatility.

IV = Aggregate Implied Volatility (Annualized)

VE = Earnings Volatility (Annualized)
VN = Normal Volatility (Annualized)

IV = ((WE*(VE^2))+(WN)*(VN^2))^0.5

The aggregate implied volatility formula was used with the earnings
weights from Figure 2.3 to calculate the aggregate implied volatilities for the
same options from our earlier example (1, 2, 7, 12, 17, 21, 42, 126, 252, and
504 trading days). The results are shown in tabular format in Figure 2.6 and
in graphical format in Figure 2.7. The weights used in the implied volatility
calculations assume one earnings trading day for all options, regardless of the
time remaining until expiration.
In Figure 2.7, the independent x-axis represents the number of trading
days before earnings. The dependent y-axis depicts the aggregate implied
volatilities. The implied volatilities for each option are illustrated by a
different line. The implied volatility lines appear in the same order as the
legend on the right-side of the graph.
It should be no surprise that we see the same pattern in implied volatility
that we observed in the earnings volatility weights approaching the earnings
announcement. The implied volatility of the option with one trading day
remaining after earnings increases from 61.3% ten trading days before
earnings to 180% on the announcement date (Figure 2.6). The increase in
implied volatility can be completely explained by the relative increase in the
earnings volatility weight relative to the normal volatility weight. Why does
the IV equal 180% on the earnings announcement date? Because the earnings
volatility equals 180% and the earnings volatility weight equals 100%. On
the earnings date, there is only one trading day remaining and it is an
earnings trading day by definition. Therefore the implied volatility must be
equal to the earnings volatility for that option.
The implied volatility of the option with two trading days remaining after
earnings increases from 59.4% ten trading days before earnings to 129.0% on
the earnings announcement date. The corresponding implied volatilities for
an option with seven trading days remaining after earnings are only 52.5%%
and 73.5%. The increase in implied volatility over the same ten days is much
more muted for longer-term options. The implied volatility for a two-month
(42 trading days) option increases by only 1.8% (40.6% - 38.8%).
Given the assumption of one earnings trading day for all options, the
earnings volatility weight for all options drops to 0% one day after earnings,
which has a predictable crushing effect on implied volatility. The implied
volatility of the option with two trading days remaining until expiration drops
from 129% on the earnings date to 30% one day later. Why 30%? Because
normal volatility equals 30% and the normal volatility weight equals 100%.
As we discovered with earnings volatility weights, the effect of earnings
volatility on the implied volatility of short-term options is much greater than
the effect on long-term options.
You might find the graph of the implied volatilities in Figure 2.7 even
more intuitive than the table in Figure 2.6. If you are an experienced option
trader, you will recognize this pattern of implied volatility that we see
repeated before and after every quarterly earnings announcement.
 The aggregate implied volatility formula can calculate the aggregate
implied volatility of any option, on any underlying security, regardless of the
number of trading days remaining until expiration. It is an extremely
powerful and versatile framework, but we need interactive tools to implement
this approach in practice.
 3 – The Basic Spreadsheet

It is now time to move from theory to practice. This chapter provides a step-
by-step explanation of how to use the first of two spreadsheets that
accompany this book: the Basic spreadsheet. I designed the Basic spreadsheet
to be accessible by a wide range of users. To minimize potential
compatibility issues, I did not use macros in this spreadsheet. In addition, the
Basic spreadsheet uses some simplifying assumptions to reduce the amount
of input data required from the user, but still estimates historical and implied
earnings volatility and uses those estimates to forecast future levels of
implied volatility around earnings announcements.
Please be aware that both spreadsheet tools have important limitations.
Unlike $1000 per year option analytical platforms that control and supply all
input data, the user is responsible for entering and verifying all of the input
data in both spreadsheets. While I included some data-validation rules to
prevent obvious input errors, the algorithms in these spreadsheets are
complex and highly sensitive to erroneous data. As a result, inaccurate or
invalid input data will generate misleading volatility estimates and possible
spreadsheet errors.
To eliminate any additional costs for the reader, I designed both
spreadsheets to use the free version of Solver that is included with Microsoft
Excel. Unfortunately, the free version of Solver has significant size and speed
limitations and its solution algorithms are not as sophisticated as those in
commercial optimization packages. I set up the spreadsheet optimization
problems to work as well as practicable with the free version of Solver, but
you may still experience varying degrees of success with Solver on certain
problems. However, you may be able to improve upon your initial results by
rerunning Solver and/or by slightly modifying Solver’s solution parameters.
The purpose of this chapter is to explain how to use the Basic spreadsheet,
not how to construct the Basic spreadsheet or how to perform every
intermediate calculation. If you have an interest in the intermediate
calculations, please review the formulas in the spreadsheet cells. Any
formulas or simplifying assumptions that materially affect volatility estimates
or prospective investment decisions will be explained in detail.
The Basic spreadsheet contains several individual tabs. The tabs will be
referenced by name. References to a cell or groups of cells on a given tab will
use the following column-row format: (A1:A20), which indicates the cells in
Column A from row 1 to row 20, inclusive.
There will be two tabs in the Basic spreadsheet that are very
straightforward and will not be discussed in detail. The first is the “Holidays”
tab, where you will enter the dates of exchange holidays (B3:B24) that will
be used by the day-count algorithms to calculate the number of trading days
throughout the Basic spreadsheet. The other tab that will not be discussed
extensively is the “Saved” tab. It is not used directly in any calculations;
instead, it is provided as a convenient location to copy and paste data from
other tabs that you might want to use again in the future.
As you proceed through this chapter, keep in mind that our goal in using
the Basic spreadsheet is to estimate historical and implied earnings volatility
and to use those estimates to forecast future levels of implied and realized
earnings volatility.

Historical Earnings Volatility

Just as we use historical volatility to evaluate implied volatility and to

forecast future realized volatility, it makes sense to use historical earnings
volatility to analyze implied earnings volatility and to predict future earnings
realized volatility. As I explained in the 2011 Active Trader article, “The
earnings volatility level is specific to each company. Factors that make
estimating earnings more difficult, such as earnings variability, rapid earnings
growth, and cyclicality, all tend to increase earnings volatility. Also, high P/E
ratios magnify the price response to earnings surprises, which also increases
earnings volatility.”
Furthermore, many factors that affect a company’s earnings tend to
persist over time: management, marketing, capital structure, product lines,
competitors, cyclical earnings patterns, and earnings growth rates. In
addition, the same institutional analysts tend to cover the same companies
from quarter to quarter and from year to year. While there is obviously some
turnover, many shareholders own the same stocks for extended periods and
traders also tend to focus on the same stocks over time. As a result, we have
the same investors, traders, and analysts forecasting the same company’s
earnings every quarter. It should be no surprise that historical earnings return
data is one of our best sources of information about future earnings returns.
We will use actual market data in this chapter and throughout the
remainder of the book. The Basic and Integrated spreadsheets use data for
Under Armor (UA) that was available as of the close on July 23, 2014. The
purpose of this chapter is to explain how to use the Basic spreadsheet, not
how to evaluate a specific trade. However, as you proceed through this
chapter, consider the implications of the UA data for constructing an option
earnings strategy. In Chapter 5, we will use Basic and Integrated spreadsheets
to design the optimal option earnings strategy for UA immediately preceding
its July 24, 2014 pre-open earnings announcement.
For both spreadsheets, user inputs are limited to cells with white text and
dark shaded backgrounds. All of the other cells contain formulas, which
should not be modified in any way. Editing formula cells would compromise
the functionality of the spreadsheets and invalidate the resulting calculations.
All tables and graphs in this section are located on the “HistoricalEV” tab of
the Basic spreadsheet.
Figure 3.1 includes a table of the historical earnings data and summary
statistics for UA. Figure 3.2 depicts the same historical earnings data for UA
graphically. The historical rates of return (calculated from the close
immediately preceding earnings to the close on the next trading day) for the
twelve previous UA earnings announcements are entered in cells B3:B14.
The Basic spreadsheet uses these returns to calculate the summary statistics
and to generate the earnings return graph.
The (natural) log returns for the twelve historical earnings events are
calculated in cells C3:C14. As explained earlier, option models generally
assume a log-normal return distribution and log returns must be used in the
calculation of earnings volatility to preserve the integrity of the aggregate
implied volatility formula. The next two columns contain intermediate
calculations, so let’s focus on the summary statistics at the bottom of the
table in Figure 3.1.
To profit from earnings announcements, we must be able to forecast
directional price changes due to earnings, identify overstated or understated
levels of implied earnings volatility, or both. The summary statistics are
designed to provide insight into both directional price changes and the
expected level of earnings volatility.
Rows 15 through 20 contain summary statistics related to the directional
return forecasts. Row 16 reports the results of a linear regression that uses the
twelve previous historical earnings returns to forecast the expected log-return
for the next earnings announcement. The return forecast for the July 24, 2014
earnings release is 6.24% (B16), which corresponds to the intercept term at
time zero. The slope of 0.53% (D16) indicates that UA earnings returns have
been gradually increasing over time. The R squared (E16) of 0.04 reveals that
the explanatory power of the regression is quite low.
Additional directional statistics for UA are reported in rows 18 to 20 and
are based on the linear regression estimate (6.24% in Cell C18), the median
log-return (3.11% in Cell C19), and the mean log-return (2.41% in Cell C20).
The next column calculates the ratio of these return values to the root-mean-
squared error (D18:D20). The resulting calculation yields a Z-score, which
can be used to forecast the probability of a positive earnings return on the
next earnings date. The linear regression estimate, the median, and the mean
historical earnings returns all suggest that the next earnings return will be
positive for UA with corresponding probabilities of 77% (E18), 64% (E19),
and 61% (E20), respectively. We will use this directional information when
we design an option earnings strategy for UA in Chapter 5.
In addition to a directional return forecast, we also need to calculate the
annualized historical earnings volatility for UA. This will help us forecast the
future realized level of earnings volatility as well as give us some insight into
the appropriate level of implied earnings volatility.
The daily root-mean-squared-error (RMSE) of the historical log returns
for UA was 8.62% (C22). This equates to an annualized earnings volatility of
136.76% (D22). We will be able to use the annualized historical earnings
volatility to estimate the expected future earnings volatility for UA. In
addition, we will also be able to use our annualized earnings volatility
forecast in the aggregate IV formula. The daily and annualized standard
deviation (SD) around the mean earnings return is also provided (C23:D23),
but as explained earlier, the RMSE provides a superior estimate of volatility
versus the standard deviation.
 I always like to study relationships visually as well as numerically.
Human beings excel at pattern recognition and it is often possible to identify
relationships visually that are not otherwise apparent. Figure 3.2 is a graph of
the historical log-returns for the twelve previous UA earnings events. The
independent x-axis represents the earnings quarter and the vertical y-axis
depicts the one-day earnings return. The red-line is a trendline that Excel
generates automatically based on the historical log-return data.
Do you notice anything in the graph that you did not see in the numerical
data? The one item that jumped out at me was that the variability of the
earnings returns had increased recently. In other words, it had become
increasingly difficult for investors and analysts to forecast the earnings for
UA in recent quarters. If that trend continued, the historical earnings
volatility estimates calculated by the spreadsheet would be too low. Hold that
thought. We will revisit the volatility forecast again in Chapter 5.
 Before we explore how to use the Basic spreadsheet to calculate implied
earnings volatility, we need to revisit the source of the historical earnings
return data in Figures 3.1 and 3.2. As you would imagine, it would be quite
cumbersome to personally track the earnings dates for every stock and to
manually calculate the quarterly return data. I recognized this immediately
when I began to use this approach to design and trade option earnings
strategies.
As a result, I subscribed to OptionSlam.com, one of the premier providers
of historical earnings data and filtering tools. While writing this book, I
advised OptionSlam.com on their filtering methodology and I requested
several site enhancements, including the ability to access historical earnings
data for backtesting purposes.
After publication, I plan to work closely with the team at
OptionSlam.com to integrate some of the analytical concepts from this book,
which will further enhance their filtering tools and volatility statistics. Please
see the resource section at the end of this book for additional information on
OptionSlam.com, including a discounted subscription offer for Exploiting
Earnings Volatility readers.
Figure 3.3 is an OptionSlam.com screenshot of the historical earnings
return data for UA as of July 1, 2014. This page illustrates a recent
OptionSlam upgrade that allows traders to screen for stock candidates and
examine each candidate’s earnings data as of a user-specified date in the
past. In other words, the historical earnings screen and the historical earnings
statistics for each company show the results we would have seen had we run
the screener on a specified analysis date in the past.
We can use this feature to backtest our earnings strategies, without being
influenced by seeing the actual earnings data that occurred after the historical
analysis date. This tool allows us to evaluate our investment process using
actual market data before we commit capital to our strategies. I never trade
any strategy without first testing and evaluating its performance historically.
Without this feature, it would have been impossible to evaluate historical
earnings candidates objectively.
There are three main sections on the OptionSlam.com screenshot in
Figure 3.3. Working from left to right, the first section provides the earnings
date, the pre-earnings close, and the after-hours percentage price move (close
to open).
The middle section reports the open, high, low, close, and one-day
percentage change for the first trading session following earnings. Both the
one-day closing and maximum percentage returns are provided. For purposes
of calculating the earnings volatility, we are interested in the closing
percentage return, which is the return on the underlying stock from the close
immediately preceding the earnings announcement to the close on the next
trading day. The return data entered in the Basic spreadsheet (B3:B14) were
taken directly from the “Closing%” column in this section.
The third section at the right-side of the page shows the mean and median
absolute and raw returns for the twelve previous earnings events, which were
calculated using one-day closing percentage returns. The user has the option
of displaying the mean and median values for the one-day closing percentage
returns or the one-day maximum percentage returns.
The mean and median values of the absolute values of the one-day
closing percentage returns are useful in evaluating historical earnings
volatility. The mean reduces the impact of all outliers relative to the RMSE
calculation and the median further reduces the effect of outliers on the
volatility estimate. The resulting mean and median values can be used as a
proxy for earnings volatility, but the annualized RMSE of the log-returns that
is calculated in the Basic spreadsheet should always be used when estimating
earnings volatility for use in the aggregate IV formula. The RMSE of the log-
returns is the only metric that is consistent with option pricing theory and the
calculation of aggregate implied volatility.
The third section also provides the mean and median raw one-day closing
returns for the twelve previous earnings events. Unlike the absolute values,
the raw returns preserve the directional movement of past earnings events,
which can be useful in identifying a systematic directional earnings return
bias for individual stocks. The positive raw returns are shown in green text
and the negative raw returns are displayed with red text to make it easier to
identify directional trends. The mean and median values provided in the
Basic spreadsheet represent the mean and median of the raw log-returns,
which is why they differ slightly from the raw mean and median statistics
calculated by OptionSlam.com.
 Implied Earnings Volatility (Dynamic)

In addition to calculating historical earnings volatility, we are also

interested in calculating the implied earnings volatility that immediately
preceded previous earnings dates. If implied volatility is the expected level of
volatility implied by or embedded in the price of the option, then implied
earnings volatility represents the expected level of earnings volatility implied
by or embedded in the price of the option.
While we can calculate the implied volatility directly from the price of an
option, we have to be more creative in solving for the implied earnings
volatility (IEV). You will recall that the aggregate implied volatility
represents a weighted average of two distinct sources of volatility: earnings
volatility and normal or non-earnings volatility. The “ImpliedEV” tab of the
Basic spreadsheet uses the aggregate IV formula and implied volatility data
immediately before and after earnings to solve for the implied earnings
volatility (IEV). Before examining how the spreadsheet solves for the IEV,
let’s review the steps required to use this tab and examine the results.
To use the “ImpliedEV” tab to solve for the implied earnings volatility
preceding a past earnings date:
1. Enter the analysis date (C2), which will always be the date immediately
preceding the earnings announcement.
2. Enter an initial “seed” estimate of the annualized implied earnings
volatility (C3).
3. Enter the expiration dates for the at-the-money (ATM) options that you
plan to use to estimate the IEV (B5:B15).
4. Enter the implied volatility for each ATM option immediately
preceding earnings (F5:F15).
5. Enter the implied volatility for each ATM option at the close of the
next trading day (J5:J15).
6. Use Solver to determine the IEV that best explains the changes in the
ATM implied volatilities.
It is important to note that the ATM IVs in columns F and J will probably
be derived from options with different strike prices. The expiration dates will
be the same, but we are interested in the at-the-money IV and the price of the
underlying security will undoubtedly change when earnings are released. The
IV of the option with the strike price closest to the closing price of the
underlying security should always be used for this calculation. Using the
ATM IVs minimizes the effect of the vertical skew on estimation of the IEV.
Figure 3.4 includes a table of the input data and all required calculations
to estimate the IEV for UA on April 23, 2014. The table includes data for
eleven different option expiration dates ranging from 4/26/2014 to 1/15/2016
(B5:B15). As explained above, the input data for each option expiration date
are entered in columns B, F, and J.
In addition to determining the number of earnings trading days (one by
definition) and normal trading days (E5:E15), the spreadsheet calculates the
estimated post-earnings ATM IV for each option expiration date (I5:I15) as
well as the estimated change in the ATM IV for each option expiration date
(K5:K15). The actual change in the ATM IV for each option is also provided
(L5:L15).
Solver uses its solution algorithm to find the value of IEV that best
explains the change in ATM IV across all options. Solver found that an
annualized IEV of 143.0% (C3) best explained the observed changes in the
ATM IVs due to the April 23, 2014 UA earnings event. The root-mean-
squared-error was 1.73% (M3), which is a measure of the average difference
between the estimated changes in the ATM IVs and the actual changes in the
ATM IVs. The estimation errors for each option expiration date are provided
in cells M5:M15.
 If you have never used Solver before, then you will first need to enable
the Solver Add-in in your version of Microsoft Excel. This exact procedure
will depend on your version of Excel. In MS Office 2010, you would use the
following button sequence:

File => Options => Add-Ins => Manage Excel Add-ins => Go

This will bring up the following screen (Figure 3.5), which you can use to
use to enable the Solver Add-in. You may need to browse to find the Solver
Add-in. Please refer to your Excel documentation for further instructions on
managing Add-ins.

The exact button sequence to run Solver may also depend on your version
of Excel. In MS Office 2010, the following sequence will bring up the Solver
problem description associated with the current spreadsheet tab.

Data => Solver

On the “ImpliedEV” tab, this should bring up the Solver problem

specification depicted in Figure 3.6. If Solver does not retrieve the problem
specification, then you will need to set the objective cell ($M$3) to “Min”
and set the changing variable cells to $C$3. Use the Add button to enter the
single constraint in the image below. Finally, the “Make Unconstrained
Variables Non-Negative” should be checked and the Solving method should
be “GRG Nonlinear.” To run Solver, press the “Solve” button.
 A comprehensive discussion of optimization theory and tools is beyond
the scope of this book, but here is a very brief explanation of what Solver is
doing in this problem. Solver is using its nonlinear solution algorithm to find
the value of the IEV (the variable cell: $C$3) that minimizes the value of the
RMSE (the objective cell: $M$3), subject to satisfying the constraint that the
IEV ($C$3) must be greater than or equal to zero. In other words, Solver is
solving for the annualized implied earnings volatility that best explains the
observed decline in implied volatilities after earnings.
 Now that I have explained how to use the “ImpliedEV” tab of the Basic
spreadsheet, how does it work? It should be no surprise that the aggregate IV
formula is the key. Specifically, the “ImpliedEV” tab uses the version of the
aggregate IV formula that allows us to solve for the normal IV, given the
implied IV and the earnings volatility.
The following example reproduces the calculation of the estimated post-
earnings ATM IV (H5) for the option with two trading days remaining until
expiration (Row 5). The spreadsheet repeats the sample calculations below
for every option in the spreadsheet.

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WE = Earnings Volatility Weight
WN = Normal Volatility Weight

WE = NTDE/( NTDE + NTDN)

WE = 1/( 2) = 50%

WN = NTDN/( NTDE + NTDN)

WN = 1/( 2) = 50%

VE = Earnings Volatility = 142.9894% (C3) (optimal Solver value)

IV = Implied Volatility = 105% (F5) (IV immediately before
earnings)
VN = Normal Volatility

VN = (((IV^2)-(WE*(VE^2)))/WN)^0.5
VN = (((105.0%^2)-(50%*(143%^2)))/50%)^0.5
VN = (16.04%)^0.5
VN = 40.05% (I5)

Remember, the Basic spreadsheet relies on the simplifying assumption

that all options have only one earnings trading day. As a result, after earnings
are released, all remaining days are assumed to be normal trading days. After
earnings, the assumed earnings weight will drop to 0% and the normal
volatility will equal the observed implied volatility.
If we assume that the normal volatility remains unchanged on the first
trading day following the earnings release, we can use the aggregate implied
volatility formula and Solver to solve for the implied earnings volatility that
best explains the observed changes in ATM IVs. The resulting IEV estimate
of 143% (C3) results in an estimated normal volatility of 40.05% (I5). That
value is an estimate of the observed implied volatility value the day after
earnings, which equals 39.8% (J5): a difference of less than 0.3% (M5).
To recap, we entered the analysis date, the expiration dates, and the at-
the-money implied volatilities before and after earnings. Solver did the rest,
solving for the implied earnings volatility that was embedded in the at-the-
money option prices immediately before the earnings announcement.

Implied Earnings Volatility (Static)

This chapter has demonstrated how to use the “HistoricalEV” and

“ImpliedEV” tabs of the Basic spreadsheet to solve for the historical earnings
volatility from past earnings events and how to estimate the implied earnings
volatility immediately (IEV) preceding past earnings announcements.
However, to identify and exploit earnings volatility opportunities, we also
need to estimate the implied earnings volatility (IEV) embedded in current
option prices.
The “EVvsTime” tab of the Basic spreadsheet performs the required
calculations that allow us to use Solver to estimate the IEV based on current
implied volatility levels. It also uses the resulting IEV estimate to calculate
future implied volatilities approaching the earnings announcement.
To use the “EVvsTime” tab to solve for the implied earnings volatility at
any point in time:
1. Enter the beginning analysis date (C2), which will always be before the
next earnings date.
2. Enter the ending analysis date (D2), the final forecast date for future
implied volatilities.
3. Enter the next earnings date (C3), which will always be the date of the
close immediately preceding the earnings announcement.
4. Enter an initial “seed” estimate of the annualized implied earnings
volatility (C4).
5. Enter the expiration dates for the at-the-money (ATM) options that you
plan to use to estimate the IEV (C6:C15).
6. Enter the implied volatility for each ATM option on the analysis date
(G6:G15).
7. Use Solver to determine the IEV that best explains the current ATM
implied volatilities.
Figure 3.7 includes a table of the input data and all required calculations
to estimate the IEV for UA on July 14, 2014. I selected an arbitrary analysis
date approximately ten days before the earnings announcement that occurred
on July 24, 2014. We will use Solver to solve for the IEV on the beginning
analysis date (July 14, 2014) and we will use the resulting IEV estimate to
calculate the future equilibrium implied volatilities from the beginning to the
ending analysis dates. The table in Figure 3.7 includes data for nine different
options expiring from 7/26/2014 to 1/15/2016 (C6:C14). Since we are
interested in estimating the IEV, all of the options should expire after the
earnings announcement.
As explained above, the input data for each option are entered in columns
C and G. In addition to determining the number of earnings trading days (one
by definition) and the number of normal trading days (F6:F15), the
spreadsheet calculates the implied normal ATM IV for each option (I6:I15)
based on the assumed IEV (C4). The deviations from the average normal
ATM IV for each option are shown in column J (J6:J15).
Solver uses its solution algorithm to find the value of IEV that minimizes
the RMSE, which is a proxy for the average difference between the implied
normal ATM IV across all options. Solver found that an annualized IEV of
137.4% (C4) minimized the average difference between the normal ATM IVs
across all options on July 14, 2014. The root-mean-squared-error was 1.13%
(J4), which is a measure of the average difference between calculated normal
ATM IVs and the average normal ATM IV (I4). At-the-money options are
used to eliminate the effects of the vertical skew and to allow an “apples-to-
apples” comparison. The estimation errors for each option are provided in
cells J6:J15.
 The exact button sequence to run Solver may depend on your version of
Microsoft Excel. In MS Office 2010, the following sequence will bring up
the Solver problem description associated with the current spreadsheet tab.
 Data => Solver

On the “ImpliedEV” tab, this should bring up the Solver problem

specification depicted in Figure 3.8. If Solver does not retrieve the problem
specification, then you will need to set the objective cell ($J$4) to “Min” and
set the changing variable cells to $C$4. Use the Add button to enter the single
constraint in the image below. Finally, the “Make Unconstrained Variables
Non-Negative” box should be checked and the Solving method should be
“GRG Nonlinear.” To run Solver, press the “Solve” button.
 In this case, Solver is solving for the annualized implied earnings
volatility on July 14, 2014 that minimizes the average difference between the
normal ATM IVs and the average normal ATM IV across all options.
As was the case before, the aggregate IV formula is the key to these
calculations. The “ImpliedEV” tab also uses the version of the aggregate IV
formula that allows us to solve for the normal IV, given the implied IV and
the earnings volatility on a given analysis date.
The following example reproduces the calculation of the estimated post-
earnings ATM IV (I6) for the option expiring on 7/26/2014. On July 14,
2014, that option had nine trading days remaining until expiration (Row 6).
The spreadsheet repeats the sample calculations below for every option in the
spreadsheet.

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WE = Earnings Volatility Weight
WN = Normal Volatility Weight

WE = NTDE/( NTDE + NTDN)

WE = 1/( 9) = 11.1111%

WN = NTDN/( NTDE + NTDN)

WN = 8/( 9) = 88.8889%

VE = 137.3808% (C4) (optimal Solver value)

IV = Implied Volatility on Analysis Date = 55.4% (G6)
VN = Normal Volatility

VN = (((IV^2)-(WE*(VE^2)))/WN)^0.5
VN = (((55.4%^2)-(11.1111%*(137.3808%^2)))/88.8889%)^0.5
VN = (10.9362%)^0.5
VN = 33.07% (I6)

All of the tabs in the Basic spreadsheet rely on the simplifying assumption
that all options have only one earnings trading day. The Basic spreadsheet
could be modified to calculate the actual number of earnings days, but I
wanted to keep this spreadsheet as simple as possible. The “EVvsTime” tab
of the Basic spreadsheet uses the aggregate implied volatility formula and
Solver to solve for the implied earnings volatility that minimizes the average
deviation between calculated normal ATM IVs and the average normal ATM
IV. Without knowing the precise shape of the horizontal normal volatility
skew, this is an effective technique for estimating the implied IEV.
The Integrated spreadsheet does calculate the actual number of earnings
days and simultaneously estimates the vertical skew, the horizontal skew, and
the implied earnings volatility. This will eliminate all simplifying
assumptions, which will improve the accuracy of the estimates and greatly
expand our ability to design and optimize strategies. However, this
functionality will come at a cost. The Integrated spreadsheet is significantly
more complex and requires more input data. That is why I provided both the
Basic and Integrated spreadsheets with this book, to ensure the analytical
framework is as accessible and practicable as possible for readers with
varying backgrounds and experience levels.
To summarize, we entered the analysis dates, the earnings date, the
expiration dates, and the at-the-money implied volatilities. Solver did the rest,
solving for the implied earnings volatility on the analysis date, which allowed
us to calculate the implied normal volatilities for each ATM option.
The resulting IEV estimate of 137.4% (C4) represents the market’s
forecast of the UA earnings volatility on the next earnings date (July 23,
2014). In other words, 137.4% was the expected level of implied earnings
volatility embedded in UA’s ATM option prices on July 14, 2014.

Forecasting Implied Volatility

We now know the implied normal volatilities for each ATM option and
the implied earnings volatility, which applies to all of the ATM options that
expire after earnings. If we assume that the earnings and normal volatilities
remain constant, we can use the aggregate IV formula to forecast the implied
volatilities for all of the ATM options as they approach the earnings
announcement. To forecast the implied volatilities, the only remaining values
that we need to calculate are the daily volatility weights, which are solely
determined by the number of normal and earnings trading days remaining on
each date. These calculations were explained in detail in Chapter 2.
In the section above, we calculated the volatility weights and the implied
normal volatility (33.07%) for the ATM UA option expiring on 7/26/2014,
which had nine trading days remaining on the analysis date of July 14, 2014.
As calculated earlier, the initial earnings volatility weight was 11.1111% and
the normal volatility weight was 88.8889%. At the close of the last trading
day before the earnings release on July 23, 2014, the earnings and normal
volatility weights were both 50%. On that date, there was exactly one normal
and one earnings day remaining until expiration. As a result, the forecasted
implied volatility was 99.92% (see calculation below).

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WE = Earnings Volatility Weight
WN = Normal Volatility Weight

WE = NTDE/( NTDE + NTDN)

WE = 1/( 2) = 50.0%

WN = NTDN/( NTDE + NTDN)

WN = 1/( 2) = 50.0%

IV = Aggregate Implied Volatility (Annualized)

VE = Earnings Volatility (Annualized)
VN = Normal Volatility (Annualized)

IV = ((WE*(VE^2))+(WN)*(VN^2))^0.5
IV = ((50%*(137.3808%^2))+(50%)*(33.07%^2))^0.5
IV = (99.8355%)^0.5
IV = 99.92%

We can repeat these calculations for every one of the ATM options, on
every trading day leading up to the earnings announcement. The resulting
implied volatilities for all of the UA ATM options are shown in Figure 3.9.
The independent x-axis shows the analysis date.
Analysis date “1” represents the initial analysis date of July 14, 2014.
Analysis date “8” represents the last trading day before the earnings release:
July 23, 2014. The dependent y-axis corresponds to the annualized implied
volatility. The daily forecasted implied volatilities are shown with separate
lines for all nine ATM UA options. The lines are shown in order of the
shortest-term option to the longest-term option (top to bottom).
As explained previously, the earnings volatility will have the greatest
effect on options that expire shortly after the earnings date. This is due to the
greater weight of the earnings volatility relative to the normal volatility,
which magnifies the impact of the outsized earnings volatility on implied
volatility. If you attempt to replicate the implied volatility calculations,
always remember that the volatility weights are multiplied by the squared
annualized volatilities, not the actual annualized volatilities.
 Experienced option traders will immediately recognize the shape of the
implied volatility curves in Figure 3.9. We have all seen the familiar pattern
of implied volatility spikes and crashes that occur like clockwork every
quarter, but we can now forecast those quarterly implied volatility changes in
advance. But how accurate are those forecasts?
The Basic spreadsheet uses some simplifying assumptions and is a crude
tool, but it is still very effective at forecasting the future pattern of implied
volatilities. Figure 3.10 includes a table with the forecasted ATM IVs, the
actual ATM IVs, and the forecast errors for the nine ATM UA options from
the previous example. The ATM UA IV forecasts in the table are for the
close on the last trading day before the release of earnings: July 23, 2014.
The ATM IV forecasts were made using the ATM IV data on the analysis
date (July 14, 2014), seven trading days before the earnings announcement.
Let’s look at an example. The IV forecast for the ATM UA option
expiring on 7/26/14 was 99.9% and the actual ATM IV was 97.6%. The
resulting forecast error was only 2.3%. Given that this forecast was made
seven trading days earlier, when the ATM IV was only 55.4%, this is
remarkably accurate. In fact, this was the largest of the forecast errors. The
next largest error was 1.4%. The root-mean-squared-error (RMSE) was 1.1%,
and the average error was only 0.8%.
 The above example was not “cherry-picked” to identify an unusually
accurate forecast. As I explained earlier, the aggregate IV formula is not an
approximation. While the implied earnings volatility and the implied normal
volatilities can obviously change from day to day, the aggregate implied
volatility formula is consistent with option pricing theory and will always
accurately aggregate the two sources of volatility into the correct value of
implied volatility.

Conclusion

Using the three analysis tabs in the Basic spreadsheet (HistoricalEV,

ImpliedEV, and EVvsTime), we can now calculate historical earnings
volatility, past levels of implied earnings volatility, and the current levels of
implied earnings and normal volatility. We can even forecast future levels of
implied volatility with a reasonable degree of accuracy.
These new capabilities will allow us to make informed decisions
regarding earnings volatility, normal volatility, and implied volatility for all
of our option earnings strategies. In addition, our improved understanding of
these volatility relationships will help us identify new strategies and better
manage our existing strategies. However, before we can explore this new
frontier in strategy development, we need to explore the implications for the
Greeks, specifically the “True Greeks.”
 4 – True Greeks

The Basic spreadsheet introduced in the last chapter helps us understand and
quantify the effects of earnings volatility more accurately than ever before,
but the Integrated spreadsheet goes much further. The Integrated spreadsheet
includes a comprehensive volatility model that simultaneously integrates and
quantifies every component of real-world volatility, including earnings
volatility. That is what will allow us to calculate the True Greeks.
The Black Scholes Option Pricing Model (BSOPM) was revolutionary
when it was introduced and it is still a remarkable tool, but every one of its
underlying assumptions is violated in practice. To mention a few, price
changes are not continuous, returns are not distributed log-normally, and
volatility is not constant. Furthermore, the formulas for the traditional Greeks
(Delta, Gamma, Vega, Theta, and Rho), which were developed using
stochastic calculus, are derivatives of the BSOPM formula. As a result, the
traditional values of Delta, Gamma, Vega, Theta, and Rho rely on the same
invalid assumptions as the BSOPM, which means that values calculated with
the conventional Greek formulas are biased and are inconsistent with the
actual price behavior of options.
In addition, the conventional Greeks assume that when one variable (like
the price of the underlying stock) changes, all of the other variables (such as
implied volatility) remain constant. That is definitely not the case. Every
change in the price of an underlying security results in a corresponding
change in implied volatility, and that change in implied volatility can be
modeled and predicted with a reasonable degree of accuracy.
The comprehensive volatility model in the Integrated spreadsheet captures
each component of real-world volatility and simultaneously estimates every
volatility parameter required to model the volatility structure of the entire
option matrix. It then uses the aggregate IV formula to accurately combine all
volatility components into a single implied volatility value, which is used to
model and simulate the true behavior of option prices in real-world market
environments. This functionality allows us to discard all of the unrealistic and
limiting BSOPM assumptions and to calculate True Greeks that reflect the
actual behavior of option prices.
For readers who devote the time to understanding and mastering the
Integrated spreadsheet, the payoffs will be significant. The Integrated
spreadsheet will allow you to calculate the relative value of every option in a
matrix and to quantify every component of volatility, including earnings
volatility. The Integrated spreadsheet will combine this information with your
specific volatility and price forecasts to simulate accurate scenario values and
to identify the optimal options earnings strategy. Finally, the Integrated
spreadsheet will provide the True Greeks and expected profit distribution for
your strategy and display the results graphically.
We will begin to explore how to use the Integrated spreadsheet in Chapter
6, but we need to examine the True Greeks now, because we will use them to
evaluate our first option earnings strategy example in the next chapter.

What are the Greeks?

Before we investigate the True Greeks in detail, let’s take a brief step
back and look at the Greeks in general. What are the Greeks and how are they
used? The calculations and derivations of traditional Greeks are based on
stochastic calculus, which is a specialized field of mathematics that is beyond
the grasp of most traders. The resulting Greek formulas are complex and are
not very intuitive, which explains why many option traders have never
attained a working knowledge of the Greeks.
While the traditional Greek formulas and their derivations are complex,
calculation of the True Greeks is much more intuitive. Furthermore,
reviewing the True Greek formulas and the logic behind them should help
many traders better understand how to apply the Greeks in practice.
Interpreting the Greeks and the True Greeks is quite straightforward.
Greeks are nothing more than price sensitivities. The odd names intimidate
many option traders, but the Greeks simply measure the change in one
variable (typically the price of an option), resulting from a specified change
in one of the explanatory variables (such as the price of the underlying
security). Using this working definition, Delta would represent the change in
the price of an option for a one dollar increase in the price of the underlying
security, holding all of the other explanatory variables constant.
So what is an explanatory variable? It is a variable that has a direct causal
effect on the price or value of an option, which means that it must be one of
the input variables in the BSOPM: stock price, strike price, time to
expiration, annualized volatility, and the level of interest rates (if we ignore
the dividend yield).
 While the strike price of an option is one of the input variables to the
BSOPM, there are no Greeks associated with a change in the strike price.
Why? Because, the strike price of an existing option does not change.
However, the remaining input variables do change and will affect the value of
an option.
By calculating how much the value of an option will change in response
to a change in each of the input variables, we can systematically quantify the
sources and magnitudes of potential returns, which are also the sources of
risk. This is not a coincidence. Risk and return are always connected; you
cannot have one without the other - except very briefly in rare examples of
pure arbitrage.
The remainder of this chapter is divided into six sections, one dedicated to
each of the following True Greeks: True Delta, True Gamma, True Earnings
Vega, True Normal Vega, True Theta, and True Rho. You will notice that the
traditional Greek Vega has been split into True Earnings Vega and True
Normal Vega. For the first time, the aggregate implied volatility formula
introduced in this book makes it feasible to calculate these new True Greek
values, giving us much more insight into volatility exposure.
Each section in this chapter will briefly define the specific Greek metric
and will explain how and why the True Greek values are calculated? If you
would like a more detailed description of the traditional Greeks (including
graphical examples), please review Chapter 2 of my first book: Option
Strategy Risk / Return Ratios.

True Delta

To illustrate how True Greeks are different from traditional Greeks, we

will begin this section with a discussion of the traditional BSOPM measure of
Delta, which I alluded to in the brief example above. To reiterate, Delta
represents the instantaneous change in the price of an option for a one dollar
increase in the price of the underlying security, holding all of the other
explanatory variables constant. Instantaneous means just that, there is no
change in the time to expiration. The change in price is immediate and no
time elapses.
While Delta is interpreted as the change in the value of an option for a
one dollar increase in the price of the underlying security, this is misleading
and I will attempt to expand on this definition to avoid any confusion. The
value of Delta corresponds to (or is scaled to) a one dollar increase in the
price of the underlying security, but it is actually measured over an
infinitesimally small change in the price of the stock and is evaluated at the
current price of the underlying stock.
For those of you who are mathematically inclined and are familiar with
calculus, Delta is the first derivative of the option price with respect to a
change in the price of the underlying stock. Said differently, it represents the
slope or sensitivity of the option price function with respect to the stock
price. More specifically, it is the slope of the line tangent to the option price
function at the current stock price. This is easier to see graphically. Even if
you are not comfortable with mathematics, you should find that the graphical
examples are intuitive.
Figure 4.1 graphically depicts the traditional BSOPM Delta for an actual
call option based on actual market data. The option used in all of the Greek
examples in this chapter (except for True Normal Vega), was the Under
Armor (UA) $60 8/2/2014 call option at the close on July 23, 2014, the last
trading day before the earnings announcement. At the time, the price of UA
was $60.63, so the $60 strike price meant the call option was slightly in the
money. On July 23rd, the option had one earnings trading day and seven
normal trading days remaining until expiration.
The independent x-axis in Figure 4.1 represents the instantaneous change
in the price of the underlying stock (UA). As a result, the value of $0.00
corresponds to the closing stock price on July 23, 2014: $60.63.
The dependent y-axis shows the instantaneous change in the price of the
UA $60 8/2/2014 call option. The curved, dashed blue line denotes the
BSOPM option price change and the straight, solid purple line is the line
tangent to the UA call option price function at the current stock price of
$60.63. The traditional measure of Delta equals the slope of that tangent line
(0.5622).
In other words, the traditional measure of Delta is a linear estimate of the
change in the value of the UA $60 8/2/2014 call option per one dollar
increase in the value of UA. It is a linear measure of the price sensitivity of
the UA call option. It indicates that the value of the call option will increase
by roughly $0.56 (per share) for an instantaneous $1.00 increase in the price
of UA.
For a one dollar change in the price of UA, the BSOPM value of the UA
call option will actually increase by an amount greater than Delta and decline
by an amount less than Delta. This is evident by the fact that the dashed blue
line is above the solid purple line.
Delta is still the appropriate linear estimate of price sensitivity, with
symmetric pricing errors due to the curvature of the option price function.
Gamma will allow us to quantify the curvature of the option price function,
which is a separate issue. We will concentrate on Gamma in the next section
of this chapter.
 For now, let’s review the characteristics of the traditional Delta statistic:
1) It is measured over an infinitesimally small price change.
2) It represents the change in the value of an option according to the
BSOPM.
3) It assumes that all of the other explanatory variables remain constant.
Now, let’s examine how True Delta differs from the traditional measure
of Delta and why those differences are important. Our calculation of True
Delta also represents the slope of the option price function, but there are
important differences. First, we do not measure the change in the price of the
option over an infinitesimally small change in the price of the underlying
security. This means that the True Delta calculation is not a derivative, which
eliminates the need to use stochastic calculus. Instead, we calculate the slope
of the option pricing function that passes through two discrete price levels for
the underlying stock: minus one dollar and plus one dollar.
In addition, using discrete points to calculate the price sensitivity or slope
function allows us to use our choice of option pricing models. We are not
limited to pricing models that are continuous and differentiable, which means
that we are no longer forced to use the BSOPM and all of its invalid
assumptions to calculate the Greeks. This permits us to incorporate much
more accurate, real-world pricing and volatility assumptions into our option
pricing model. We will use this simple two-point slope methodology to
calculate all of the True Greeks.
Figure 4.2 graphically depicts the True Delta for the above UA call
option. The independent x-axis in Figure 4.2 represents the instantaneous
change in the price of the underlying stock (UA). The dependent y-axis
shows the instantaneous change in the price of the UA $60 8/2/2014 call
option.
The curved, dashed blue line denotes the expected change in the price of
the UA call option as determined by a comprehensive volatility model, not by
the BSOPM. The straight, solid purple line is no longer tangent to the UA
call price function. Instead, it passes through two discrete points that
correspond to plus and minus one dollar price changes in the underlying
stock (UA). The slope of the line passing through these two points gives us
the value of True Delta (0.5558) – See Figure 4.2.
 True Delta is still a linear estimate of the change in the value of the UA
$60 8/2/2014 call option per one dollar increase in the price of UA. However,
it is not a derivative of the BSOPM function. Instead, we calculate the slope
from two discrete points of the option price function, which is determined by
a comprehensive volatility model that includes all volatility components,
which are estimated from the full UA option matrix.
Let’s review the characteristics of True Delta:
1) It is measured over discrete price changes in the underlying security.
2) It represents the change in the value of an option according to a
comprehensive volatility model.
3) It does not assume that all of the other explanatory variables remain
constant.
The difference between the values of traditional Delta and True Delta in
this example is small, but the errors in the traditional Greek calculations can
be quite large and introduce an undesirable systematic bias. The critical issue
is that the True Greeks, including True Delta, capture how option prices
actually behave and are not constrained by artificial BSOPM pricing
assumptions.
The True Delta formula below will demonstrate how all of the True Greek
statistics are calculated.

DT = True Delta
ΔPO+ = Change in Price of Option for + $1 change in Price of Stock
ΔPO- = Change in Price of Option for - $1 change in Price of Stock
ΔPS+ = Positive Change in Price of Stock (+1)
ΔPS- = Negative Change in Price of Stock (-1)

DT = (ΔPO+ - ΔPO-)/(ΔPS+ - ΔPS-)

DT = (0.5927 - (-0.5190))/( 1 - (-1))
DT = (1.1117)/( 2) = 0.5558

According to the comprehensive volatility model, the price of the UA call

option in our example would increase by $0.5927 for a one dollar increase in
the price of UA. This includes the expected changes in implied volatility
associated with the price increase in UA and also incorporates the expected
level of earnings volatility.
 We also use the comprehensive volatility model to calculate the price
change in the UA call option due to a one dollar decrease in the price of UA.
According to the model, the price of the call option would decline by
$0.5190, which includes the modeled changes in implied volatility due to the
UA price decline.
True Delta represents the average price sensitivity of the option. It is
calculated over the price range in the underlying security of minus one dollar
to plus one dollar. The numerator in the True Delta formula represents the
change in the price of the option, and the denominator reflects the change in
the price of the underlying stock. True Delta represents the average change in
the price of the option per one dollar increase in the price of the underlying
stock.

True Gamma

As explained earlier, Gamma measures the curvature of the option price

function. As a result, it is the only Greek that we will examine that does not
measure the price sensitivity of an option. Instead, it represents the change in
Delta, per one dollar increase in the price of the underlying security. In our
case, True Gamma will represent the change in True Delta, per one dollar
increase in the price of the underlying security.
The traditional Gamma statistic suffers from the same flaws as the
traditional measure of Delta:
1) It is measured over an infinitesimally small price change.
2) It represents the change in the value of Delta according to the BSOPM.
3) It assumes that all of the other explanatory variables remain constant.
True Gamma offers the same benefits as True Delta:
1) It is measured over discrete price changes in the underlying security.
2) It represents the change in the value of True Delta according to a
comprehensive volatility model.
3) It does not assume that all of the other explanatory variables remain
constant.
True Gamma does not measure the change in True Delta over an
infinitesimally small change in the price of the underlying security. This
means that the True Gamma calculation is not a derivative, which eliminates
the need to use stochastic calculus. Instead, we calculate the change in True
Delta using two discrete price levels for the underlying security: minus one
dollar and plus one dollar.
Using discrete points to calculate the change in True Delta allows us to
use our choice of option pricing models, permitting us to incorporate much
more accurate, real-world pricing and volatility assumptions into our option
pricing model.
Figure 4.3 depicts the True Gamma for the above UA call option
graphically. The independent x-axis in Figure 4.3 represents the
instantaneous change in the price of the underlying stock (UA). The
dependent y-axis shows the instantaneous change in the price of the UA $60
8/2/2014 call option. The curved, dashed blue line denotes the expected
change in the price of the UA call option as determined by a comprehensive
volatility model, not by the BSOPM.
The straight, solid yellow line passes through two discrete points that
correspond to price changes in the underlying stock (UA) of minus two and
zero dollars. The slope passing through these two points gives us the value of
True Delta centered at a price change of minus one (0.4813).
The straight, solid green line passes through two discrete points that
correspond to price changes in the underlying stock (UA) of zero and plus
two dollars. The slope passing through these two points gives us the value of
True Delta centered at a price change of plus one (0.6274).
You will note that the True Delta at minus one dollar (0.4813) is less than
the initial value of True Delta (0.5558). Similarly, the value of True Delta at
plus one dollar (0.6274) is greater than the initial value of True Delta. In
other words, the True Delta of the UA call option increases as the price of
UA rises and decreases as the price of UA falls. Why? Because, asymmetric
payoffs benefit the owner of the option.
As the price of UA rises, the probability of exercising the call option
increases. As the price of UA falls, the probability of exercising the call
option declines. This forces the value of the call option to rise faster and fall
more slowly than suggested by the linear estimate of True Delta.
The True Gamma formula below is very similar to the True Delta
formula. However, instead of calculating the change in price per one dollar
change in the price of the underlying stock, True Gamma calculates the
change in True Delta per one dollar change in the price of the underlying
stock. Both formulas use the same discrete slope calculations.

GT = True Gamma
ΔDT+ = Change in True Delta for + $1 change in Price of Stock
(0.6274 - 0.5558)
ΔDT- = Change in True Delta for - $1 change in Price of Stock
(0.4813 – 0.5558)
ΔPS+ = Positive Change in Price of Stock (+1)
ΔPS- = Negative Change in Price of Stock (-1)

GT = (ΔDT+ - ΔDT-)/(ΔPS+ - ΔPS-)

GT = (0.0716 - (-0.0745))/( 1 - (-1))
GT = (0.1461)/( 2) = 0.0730

Because of cancelling terms, we could also use the actual values of True
Delta in the formula instead of the changes in True Delta. The Integrated
spreadsheet does all of these calculations for you, but I will include the
alternative version of the formula for readers who would like to experiment
further.
The numerator and denominator of the True Greek formulas can all be
modified to use the actual values instead of changes in the actual values. Note
that the values of the numerator and denominator in the alternative version of
the formula did not change, nor did the value of True Gamma.

GT = True Gamma
DT+ = True Delta $1 Above Current Stock Price (0.6274)
DT- = True Delta $1 Below Current Stock Price (0.4813)
PS+ = Stock Price $1 Above Current Stock Price (61.63)
PS- = Stock Price $1 Below Current Stock Price (59.63)
GT = (0.6274 - 0.4813)/(61.63 - 59.63)
GT = (0.1461)/( 2) = 0.0730
 The resulting calculation of True Gamma reflects the option price
behavior we observe in actual market environments. The realistic option price
performance is captured by True Delta, which reflects every component of
volatility, including the interaction of volatility with the price of the
underlying security.

True Earnings Vega

Using the True Greek discrete slope calculation framework in conjunction

with the aggregate implied volatility formula allows us to calculate the True
Earnings Vega. For the first time, we now have the tools to calculate the
average change in the value of any option per 1% increase in implied
earnings volatility. This new risk measure or Greek will provide unique
insights into constructing and evaluating option earnings strategies.
The calculation of the True Earnings Vega is very similar to the procedure
we used earlier to calculate True Delta. However, instead of calculating a
linear estimate of the change in the price of an option per one dollar change
in the price of the underlying stock, True Earnings Vega will calculate the
linear estimate of the change in the price of an option per 1% increase in the
annualized implied earnings volatility.
Note, we are only interested in the change in implied earnings volatility,
not implied volatility. In other words, we are assuming that normal volatility
remains unchanged. In this case, this is a realistic assumption. There should
be no cause and effect relationship between a change in implied earnings
volatility and a change in normal volatility.
Figure 4.4 graphically depicts the True Earnings Vega for the above UA
call option. The independent x-axis in Figure 4.4 represents the instantaneous
change in the annualized implied earnings volatility. The dependent y-axis
shows the instantaneous change in the price of the UA $60 8/2/2014 call
option.
The curved, dashed blue line denotes the expected change in the price of
the UA call option as determined by a comprehensive volatility model, not by
the BSOPM. The straight, solid purple line passes through two discrete points
that correspond to plus and minus one dollar price changes in the underlying
stock (UA). The slope of the line passing through these two points gives us
the value of True Earnings Vega (0.01335).
 You are probably asking yourself why you only see one line in Figure 4.4.
The reason is that the change in the value of the underlying stock for a 1%
change in the implied earnings volatility is almost linear: minus 0.0133
versus +0.0134. As a result, the solid purple True Earnings Vega line overlies
and obscures the blue dashed line of the option price function for the UA call
option.
You might also ask, why is True Earnings Vega so small: 0.01335? The
True Earnings Vega of 0.01335 indicates that the average change in the price
of the UA call option is only $ 0.01335 per share, per 1% change in the
annualized implied earnings volatility. The Greek examples in this chapter
are expressed on a per share basis, because it is possible to compare per share
Greeks directly with option prices. They are expressed in the same units and
most traders find per share Greeks more intuitive.
However, in reality, the change in the value of the option position per
contract (100 shares) would be $1.335, which is obviously still a small
number. Why so small? First, on July 23rd 2014, the UA call option in our
example had only one earnings trading day and seven normal trading days
remaining until expiration. As a result, the earnings volatility weight was
small relative to the normal volatility weight, which dampens the impact of
the 1% change in implied earnings volatility on implied volatility.
In addition, a 1% change in implied earnings volatility (IEV) represents a
very small change relative to what we actually observe in the options market.
As you use the spreadsheets to calculate IEV, you will discover that
annualized IEV typically exceeds 100% and can easily reach 200% to 300%
for stocks with a history of volatile responses to earnings.
As a result, it is not uncommon to see 20% to 30% swings in IEV. The
True Earnings Vega calculation is accurate and consistent with the traditional
Vega concept, but the volatility of the explanatory variable (IEV in this case)
must always be considered when applying the Greeks in practice.
The True Earnings Vega formula below is consistent with how all of the
True Greek statistics are calculated.

EVT = True Earnings Vega

ΔPO+ = Change in Price of Option for + 1% change in IEV
ΔPO- = Change in Price of Option for – 1% change in IEV
%ΔIEV+ = Positive Percentage Change in IEV (+1)
%ΔIEV- = Negative Percentage Change in IEV (-1)

EVT = (ΔPO+ - ΔPO-)/(%ΔIEV+ - %ΔIEV-)

EVT = (0.0134 - (-0.0.0133))/( 1 - (-1))
EVT = (0.02675)/( 2) = 0.01335

True Normal 30 Vega

In the previous section, we used the True Greek discrete slope calculation
framework in conjunction with the aggregate implied volatility formula to
calculate the True Earnings Vega. We can use the same approach to calculate
the True Normal 30 Vega. I have stated that implied volatility is a function of
earnings and normal volatility, which is true, but it is a slight
oversimplification. Even if we separate out earnings volatility, normal (non-
earnings) volatility is not constant across options in the matrix.
Due to the technical nature of the subject, I will not discuss volatility
modeling in detail here. For those of you who would like more information
on the comprehensive volatility model used in the Integrated spreadsheet, we
will review the volatility model more fully in Chapter 7.
Briefly, normal volatility is a function of the vertical and horizontal skews
and it also responds to changes in the price of the underlying stock. The
vertical skew captures the relationship between normal implied volatility and
option strike prices, which were typically listed vertically in financial
newspapers.
The horizontal skew captures the relationship between normal implied
volatility and option expiration dates, which were typically listed horizontally
in financial newspapers. The cause and effect relationship between the price
of the underlying security and the at-the-money (ATM) normal implied
volatility is also important and must be included in any volatility model.
The comprehensive volatility model in the Integrated spreadsheet uses the
entire option matrix to estimate all of the required volatility parameters and
applies them when calculating the True Greeks. The volatility model is
particularly important to the calculation of the True Normal 30 Vega.
Why is it called the True Normal 30 Vega (TN30 Vega)? At this point,
“True,” “Normal,” and “Vega” may be self-explanatory. It is a “True” Greek
because it quantifies the actual price behavior of options and is not limited by
the artificial BSOPM assumptions. “Normal Vega” means that it calculates
the sensitivity to a 1% change in normal implied volatility.
So what does the number 30 signify? The number 30 represents thirty
days or one month until expiration. The BSOPM assumes that changes in
implied volatility are constant across all options. We already know that this is
not true because of earnings volatility, but it is not even true for normal
volatility.
In practice, normal volatilities are mean-reverting. In other words,
extremely low and extremely high levels of normal volatility eventually
revert to a more typical or average level of normal volatility. This is true for
practically all securities across all markets. Due to mean reversion, the
normal implied volatilities of short-term options change more than the
normal implied volatilities of longer-term options. This is true regardless of
the initial shape of the horizontal skew curve: positively sloped, flat, or
inverted.
 As a result, instead of being constrained by the limitations of traditional
Vega, we do not need to assume that all normal implied volatilities change by
the same amount when calculating the True Normal 30 Vega. Instead, we can
assume a 1% change in the normal implied volatility for a one-month option,
and model the estimated changes in implied volatility for all other options as
a function of their time remaining until expiration.
I created such a model using daily implied volatility data for a range of
expiration dates, across a broad spectrum of diverse securities. The
underlying securities included stocks, ETFs, commodities, currencies, bonds,
and indices. The results from the model are depicted in Figure 4.5.
Figure 4.5 graphically depicts the normal at-the-money (ATM) implied
volatility (IV) multiplier versus the time remaining until option expiration.
The independent x-axis represents the time remaining until option expiration
(in years) and the dependent y-axis shows the ATM IV multiplier. The green
square, purple diamond, red triangle, and blue circle highlight the x and y
coordinates for option expirations of one week, one month, six months, and
one year, respectively. The blue line depicts the continuous, non-linear ATM
IV function for expiration dates ranging from one week to two years.
The key reference point in the diagram is the one-month option expiration
(0.0833. 1.0000). The 0.0833 represents 1/12th of a year and the 1.0 signifies
an ATM IV multiplier of 1.0. The multiplier tells us how much the ATM IV
of the option will change for a 1% change in the ATM IV of a 30 day or one-
month option. In that context, the multiplier for a one-month option must be
1.0 by definition.
The one-week option is represented by the green square with a time to
expiration of 0.0198 years (5 trading days/252 trading days). The multiplier
of 1.5697 indicates that for a 1% change in the ATM IV of the 30-day option,
the ATM IV of the one-week option will change by 1.5697%. As stated
earlier, the implied volatilities of short-term options are much more volatile
than the implied volatilities of longer-term options and that is evidenced by
the large ATM IV multiplier for a one-week option.
Based on that premise, we should expect the ATM IV multipliers of
options with expiration dates longer than one month to be less than 1.0 and
that is exactly the case. The ATM IV multipliers for the six-month and one-
year options are 0.2406 and 0.2194, respectively. For a 1% change in the
ATM IV of a 30-day option, the ATM IV of the six-month and one-year
option will only change by 0.2406% and 0.2194%, respectively.
The formula for the ATM IV multiplier is provided below in Excel
format.

Multiplier = ATM IV Multiplier

TTE = Time to Option Expiration in Years
EXP represents the Excel function e raised to a power, e =
2.718281828

Multiplier = 1.0+(-0.780854818)*(1-EXP(-8.629216*(TTE-1/12)))

If I was providing an option analytical platform and all of the data, I

would probably estimate the ATM IV multiplier function for each underlying
security individually, instead of estimating a single multiplier function that
would be applied to all securities. However, that would not be practical in the
spreadsheets that accompany the book.
So, how accurate is the above function for the broad range of securities
used in the estimation? The above ATM IV multiplier formula explained
96.6% of the variation in the ATM IV regression coefficients across all of the
stocks, ETFs, commodities, currencies, bonds, and indices used in the
estimation process. The fact that volatility is universally mean reverting
results in a remarkably similar pattern of ATM IV sensitivities as a function
of time to option expiration, regardless of the type of market or the specific
underlying security.
While there will still be some variation around the above ATM IV
multiplier estimates, keep in mind that the BSOPM essentially assumes an
ATM IV multiplier of 1.0 for all option expirations, which is obviously
incorrect. The above model represents a dramatic improvement over the
BSOPM. But how do we use the model to calculate the True Normal 30
Vega?
We use the same discrete slope calculation that we used earlier to
calculate True Delta and True Earnings Vega. However, instead of assuming
a 1% increase and decrease in volatility for all options, we will use the ATM
IV multiplier to determine the changes in the normal implied volatility for
each option, which will be a function of its time remaining until expiration.
We will use the specific changes in the normal implied volatility (positive
and negative) for each option to calculate the corresponding discrete changes
in the value of each option, which will be used in the slope calculation.
This should be easier to understand by reviewing the table in Figure 4.6.
Due to the complexity of these calculations, this is the only example in the
chapter that is hypothetical and does not use actual market data. Instead, it
uses four hypothetical options with times remaining until expiration of one
week, one month, six months, and one year.
All four options assume an underlying stock price of $100, a strike price
of $100, a normal implied volatility of 25%, and a risk-free interest rate of
0.25%. In addition, none of the options expire after an earnings event. Using
four at-the-money options that are not influenced by earnings greatly
simplifies the example by eliminating the effects of the vertical skew and
earnings volatility and should make the results more intuitive.
In Figure 4.6, the four options are listed in order of time to expiration,
from shortest to longest. The time to expiration in years is provided in the
first column, followed by the normal ATM IV multiplier, and the change in
the normal ATM IV that will be used to calculate changes in option values.
The next two columns report the change in option value for the stated
negative and positive changes in the normal ATM IV. The True Normal 30
Vega (TN30 Vega) and the traditional BSOPM Vega values are shown in the
next two columns. Finally, the BSOPM error, which represents the difference
between BSOPM Vega and the TN30 Vega, is reported in the last column.
Due to the ATM IV multiplier, the BSOPM Vega understates the TN30
Vega of the one-week option by $0.032 per share, which corresponds to
$3.20 per contract. This might not sound that significant, but it represents an
error of over 36%.
Similarly, the BSOPM Vega understates the TN30 Vega of the one-year
option by $0.309 per share, or $30.90 per contract. This is an error of 354%.
In other words, the BSOPM Vega is over 4.5 times as large as the TN30
Vega. The traditional Vega metric dramatically overstates the volatility
exposure of long-term options and understates the volatility exposure of
short-term options by naively assuming that implied volatility changes are
constant across the entire term structure of volatilities.
 The True Normal 30 Vega formula below is consistent with how all of the
True Greek statistics are calculated, with one exception. The instantaneous
option price changes are calculated for a 1% change in the 30-day normal at-
the-money implied volatility, which results in different and unique changes in
the normal volatility for each option in the matrix. The values for the
hypothetical one-year option in Figure 4.6 are used in the formula example
below:

N30VT = True Normal 30 Vega

ΔPO+ = Change in Price of Option for + 1% change in the 30-Day
ATM IV
ΔPO- = Change in Price of Option for - 1% change in the 30-Day
ATM IV
%ΔN30IV+ = Positive Percentage Change in 30-Day ATM IV (+1)
%ΔN30IV- = Negative Percentage Change in 30-Day ATM IV (-1)

N30VT = (ΔPO+ - ΔPO-)/(%ΔN30IV + - %ΔN30IV -)

N30VT = (0.087 - (-0.087))/( 1 - (-1))
N30VT = (0.174)/( 2) = 0.087

The calculation of True Normal 30 (TN30) Vega in the Integrated

spreadsheet accounts for the fact that normal volatility changes are not
constant across the term structure of volatilities. Instead, the TN30 Vega
models the changes in normal volatility as a function of the time to
expiration. It integrates these ATM IV changes with the vertical skew to
determine the change in normal volatility for each option. Finally, it uses the
aggregate IV formula to calculate the resulting change in aggregate implied
volatility, which includes the effects of earnings volatility. Finally, these
aggregate IV changes are used to solve for the resulting option prices and the
True Greeks.
It is important to understand that all components of volatility are used in
the calculation of the True Greeks, which ensures that the resulting values
reflect the actual price behavior of all options in the matrix. However, the
Integrated spreadsheet performs all of these calculations automatically and
you will not need to use any of the formulas in this book manually.

True Theta

The traditional Theta formula is a derivative of the continuous BSOPM.

As a result, the conventional value of Theta corresponds to (or is scaled to) a
one calendar day decrease in the time remaining until expiration of each
option, but it is actually measured over an infinitesimally small change in
time to expiration. Since the BSOPM function is continuous, the traditional
measure of Theta completely ignores the existence and effects of discrete
earnings announcements, which creates large systematic biases in the Theta
calculation.
True Theta eliminates these deficiencies and uses the aggregate IV
formula and the comprehensive volatility model to accurately quantify the
effect of the passage of time on the value of all options, even those expiring
after earnings announcements.
Unlike the other True Greeks, there is no need to calculate the slope of the
option price function based on positive and negative changes in the
explanatory variable. Time only moves in one direction. As a result, True
Theta calculates the change in the value of each option for a discrete one
trading day decrease in time remaining until expiration. Note that Theta is
the only Greek that corresponds to a decrease in the explanatory variable.
We will now examine the conventional and True Theta values for the
option example we used earlier in this chapter: the Under Armor (UA) $60
8/2/2014 call option at the close on July 23, 2014, the last trading day before
the earnings announcement. On July 23rd, the option had one earnings trading
day and seven normal trading days remaining until expiration.
Since the next UA earnings announcement occurred before the open on
July 24th, 2014, the passage of one trading day from the 23rd to the 24th
reduced the number of earnings trading days from one to zero. The number
of normal trading days remained constant at seven.
True Theta used the aggregate IV formula to automatically reduce the
earnings volatility weight to zero after the earnings announcement. True
Theta also used the comprehensive volatility model to adjust effects of the
horizontal skew curve due to the passage of time. The resulting True Theta
value for the UA call option at the close on July 23rd, 2014 was -$1.245 per
share.
Now, consider the traditional value of Theta for the same option on July
23, 2014. Given the continuous nature of the BSOPM, ask yourself the
following question: was the traditional value of Theta higher or lower than
the True Theta value of -$1.245?
The continuous BSOPM had no way of accounting for the discrete
earnings event that occurred before the open on July 24, 2014. As a result, the
traditional Theta formula incorrectly assumed that implied volatility would
remain constant until option expiration and that the option value would decay
continuously.
On July 23, 2014, the traditional BSOPM Theta value was -$0.115, which
understated the magnitude of the one-day time decay by $1.115 per share, or
a staggering $111.5 per contract. The value of True Theta was almost eleven
times the traditional BSOPM Theta value.
Due to the simplicity of the True Theta calculation, no graph is provided,
but here is the formula for True Theta:

TT = True Theta
PO0 = Original Price of Option
PO- = Price of Option for one trading day decrease in time to
expiration
ΔPO- = Change in Price of Option for one trading day decrease in
time to expiration

TT = ΔPO- = (PO- - PO0)

TT = -1.245
 The Integrated spreadsheet calculates the value of True Theta directly for
all options, so no intermediate values are provided in the above example.

True Rho

Using the True Greek discrete slope calculation framework allowed us to

calculate True Delta and True Earnings Vega. The calculation of the True
Rho is very similar to the procedure we used earlier to calculate True Delta.
However, instead of calculating a linear estimate of the change in the price of
an option per one dollar change in the price of the underlying stock, True Rho
will calculate the linear estimate of the change in the price of an option per
1% increase in the risk-free interest rate. US T-bill rates are typically used as
a proxy for the risk-free interest rate.
Since short-term US T-Bill rates are very near zero as I write this, the
discrete risk-free interest rate changes used in the slope calculations are
actually plus and minus 0.10%, not plus and minus 1.0%. This has no
material effect on the slope calculation or on the value of True Rho, but it is
important that the risk-free interest rate be greater than zero when attempting
to determine the value of an option. Given the correspondence to the earlier
graphical examples for True Delta, I did not include a graph for True Rho.
The True Rho formula below is consistent with how all of the True Greek
statistics are calculated, except for the smaller discrete changes in the risk-
free interest rate. True Rho is still interpreted as the change in the value of an
option for a 1% increase in the risk-free interest rate, regardless of the
magnitude of the discrete interest rate changes used in the formula below.

RT = True R
RF = Risk Free Interest Rate
ΔPO+ = Change in Price of Option for + 0.1% change in RF
ΔPO- = Change in Price of Option for – 0.1% change in RF
%ΔRF+ = Positive Percentage Change in IEV (+0.1)
%ΔRF- = Negative Percentage Change in IEV (-0.1)

RT = (ΔPO+ - ΔPO-)/(%ΔRF+ - %ΔRF-)

RT = (0.0008726 - (-0.0008726))/( 0.1 - (-0.1))
 RT = (0.0017452)/(0.2) = 0.008726

The value True Rho for the UA call option in our example is $0.008726
per share, or $0.8726 per contract. The value of True Rho is very small due to
the short time to expiration for the UA call option. The impact of a 1%
change in the annual risk-free interest rate would have a minimal impact on
an at-the-money option with only eight trading days remaining until
expiration. As is the case for all of the True Greeks, the True Rho calculation
incorporates the aggregate IV formula and the comprehensive volatility
model.

Conclusion

Even under normal conditions, the traditional BSOPM Greek values are
biased and are inconsistent with actual option price behavior. Earnings
announcements magnify these systematic biases and demonstrate the
importance of using the aggregate IV formula and the True Greeks to manage
risk and to design option strategies.
 5 – UA Earnings Strategy

In this chapter, we will look at the potential payoff from using the tools from
this book to identify the option earnings strategy that offers the highest
expected return per unit of risk. We will use actual market data for Under
Armor (UA) on July 23, 2014 to ensure the trade example is as realistic as
possible.
The UA strategy example will display results from both the Basic and
Integrated spreadsheets. We covered the Basic spreadsheet in Chapter 3 and
we will explore the Integrated spreadsheet in the next few chapters. As a
result, our discussion of the Integrated spreadsheet tools in this chapter will
be limited to explanations of the screenshots used to illustrate the UA trade.

Trading Edge

Our goal in designing an option earning strategy is to exploit option

pricing anomalies related to earnings, which creates an advantage or trading
edge. Our new spreadsheet tools now allow us to precisely quantify historical
earnings volatility as well as past and current levels of implied earnings
volatility (IEV). How could we use this information?
If we determined that the current IEV was too low relative to historical
IEV and historical earnings volatility, earnings volatility would be too cheap
and options that expired after the earnings announcement would be
underpriced. If this occurred several weeks prior to the earnings
announcement, we could design a strategy that would profit from the
expected increase in IEV.
If we identified the IEV disparity shortly before the earnings
announcement, we could design a strategy that would profit from a higher
than expected realized earnings volatility. In other words, we would bet that
realized volatility would exceed implied earnings volatility.
Instead, if IEV was too high, we could construct the opposite type of
strategy - one that would benefit from either a decline in IEV or from a lower
than expected level of realized earnings volatility.
Similarly, we can construct strategies that benefit from directional price
changes in the underlying stock due to earnings announcements. However,
before we can design option strategies to exploit directional biases and IEV
anomalies, we must first identify these opportunities. In Chapter 3, I
explained how to use the Basic spreadsheet to evaluate historical directional
price changes due to earnings announcements and how to calculate all of the
earnings volatility metrics.
I will introduce another directional resource in this chapter, but keep in
mind that modeling directional price changes due to earnings announcements
is not the main focus or purpose of this book. The simple directional
forecasting examples are only provided as inspiration for your own
directional modeling research.
The primary objective of this book is to introduce a theoretically sound
framework for quantifying the effects of earnings volatility on all options and
to provide tools that will help traders apply that framework in practice.
The framework not only helps us quantify and evaluate earnings
volatility, it can also be used to identify the relative value of every option in
the matrix. We will discuss volatility modeling in more detail when we
explore the Integrated spreadsheet in Chapter 7. For now, just remember that
the volatility model simultaneously estimates all of the volatility parameters,
including the implied earnings volatility.
The Integrated spreadsheet uses the resulting volatility model to calculate
the theoretical price of every option in the matrix, which it then compares to
the actual mid-market price to find the precise degree of overvaluation or
undervaluation of each option.
The Integrated spreadsheet then uses the volatility model (in conjunction
with our custom directional and volatility forecasts) to calculate realistic
option prices for a large number of probability-weighed scenarios. The
Integrated optimizer then attempts to find the strategy with the highest
expected profit per unit of risk. All of these tools help us create a trading
edge that we can use to exploit earnings anomalies.

Screening Tools

So where do we begin? There are literally thousands of stocks that we

could use to design an option earnings strategy, so first we need to narrow the
list to a more manageable number of candidates. I use OptionSlam.com’s
stock screener for this purpose. OptionSlam provides two versions of the
screener: the upcoming earnings stock screener for real-time analysis and the
historical stock screener for backtesting or historical analysis. Since we are
looking at historical dates, we will be using the historical stock screener in
this example. OptionSlam offers many different screening filters, but I will
limit my discussion to those used in this example.
Figure 5.1 is a screenshot of OptionSlam.com’s historical stock screener
with all of the filter settings that I used on June 30, 2014 to find prospective
candidates with earning announcements scheduled during the subsequent
three months. The “historical running date” (2014-06-30) and the “earnings
period from historical date” (Within Three Months) selections are both
shown in the top section of the screener.
In the second section of the form, I limited my selection to stocks with
weekly options. This served several purposes. First, stocks with weekly
options offer much more flexibility in constructing option earnings strategies.
As you already know, earnings volatility has a much greater effect on the
implied volatility of options expiring shortly after earnings, which makes it
more difficult for option traders to evaluate near-term weekly options. This
creates pricing anomalies and trading opportunities.
In addition, weekly options are typically available only on liquid stocks
with substantial option trading volume. Focusing exclusively on stocks with
weekly options should enhance liquidity. In this section, I also limited
prospects to stocks with an average daily volume of over one million shares.
Maximizing liquidity and reducing transaction costs are crucial to option
trading and both of these filters were included to help satisfy these objectives.
In the next section titled “Earnings Statistics,” I specified a minimum
EVR of 2.0 and a minimum stock price of $20 per share. EVR is
OptionSlam’s proprietary earnings volatility measure, which is based on the
most recent three years of quarterly earnings announcements, with the most
recent data weighted more heavily. EVR ranges between zero and ten, with
ten being the most volatile. Therefore, a minimum EVR of 2.0 eliminates the
stocks with the lowest level of historical earnings volatility. Why would we
do so?
The hypothesis is that stocks with higher levels of earnings volatility
would be more difficult for option traders to evaluate, which should increase
the frequency and magnitude of earnings-related pricing anomalies. Given
our ability to accurately quantify earnings volatility, this should give us an
edge that we can use to our advantage.
The minimum stock price of $20 increases the value of options relative to
transaction costs and commissions. The commissions are the same on $600
stocks and $6 stocks, but options on $600 stocks control a lot more market
value. As a result, setting a minimum stock price prevents us from wasting
our time evaluating options on low priced stocks, which would have
unacceptably high commissions. Keep in mind that there are no magic
numbers. You might prefer to use a minimum stock price of $15 and a
minimum daily volume of 500,000 shares. The choice is yours. Relaxing the
filters would increase the number of prospective candidates, which has both
costs and benefits.
The next section is titled “Tracking Price Change One Day After Previous
Earnings Release.” I did not use this section in my historical screen, but it
allows the user to apply volatility and directional filters to each of the
previous one, two, three, or four earnings events.
Instead of the above section, I prefer to screen on the mean and median
closing movements of the twelve previous earnings events. The first two
filters in this section limit the prospective candidates to stocks that
experienced a minimum mean and median absolute price change of 5% or
more over the past twelve earnings announcements. “Absolute” means the
absolute value of the one-day percentage price change experienced from the
close immediately preceding earnings to the close on the next trading day.
The purpose of this filter is the same as that of the EVR filter used earlier: to
limit candidates to stocks with higher levels of earnings volatility.
The last two filters in this section limit the prospective candidates to
stocks that experienced a minimum mean and median raw price change of
2% or more over the past twelve earnings announcements. The raw price
changes include both positive and negative returns. If the mean and median
raw returns were above 2%, that indicates a positive return bias due to
previous earnings events. In other words, the resulting candidates consistently
surprised analysts and traders to the upside, generating mean and median
returns in excess of 2% per quarter. We will attempt to use this historical bias
to our advantage.
The “Report Format Options” allow us to sort and display the resulting
candidates using several different metrics. I chose to sort the results based on
the “Average of Earnings Movements – Absolute,” which displayed the
candidates in descending order of historical earnings volatility. Figure 5.1 is a
screenshot of OptionSlam.com’s Historical Earnings Screener, with the filter
settings described above. Figure 5.2 is a screenshot of the historical
screener’s results on June 30, 2014.
 The filters I used returned 17 stocks, a manageable list of candidates.
However, if we look closer, we can immediately eliminate six more
candidates. It may be difficult to see in the screen-capture image, but the
mean and median closing returns for six of the candidates include a number
in parentheses with an asterisk (10*). This indicates that our specified twelve
earnings events were not available for these stocks. Given that we are using
the historical data to estimate earnings volatilities and directional biases, we
will eliminate candidates that do not have sufficient historical data. That
leaves us with only eleven prospective candidates for us to consider.
For each of the candidates in Figure 5.2, the symbol, next earnings date,
market, sector and EVR are provided on the left side of the table. The right
side of the table includes the mean and median absolute 1-day percentage
price changes over the twelve previous earnings events, followed by the
mean and median raw 1-day percentage price changes over the twelve
previous earnings events. Remember, the absolute percentages are a measure
of volatility and the raw percentages indicate directional bias.
A complete examination of all eleven candidates would not be practical
here, but I do want to restate our objective: identify a stock with implied
earnings volatility (IEV) and option pricing anomalies that we can exploit.
After examining the remaining eleven candidates, I selected Under Armor
(UA) and chose to implement the option earnings strategy on July 23, 2014,
the last trading day before the UA second fiscal quarter 2014 earnings
announcement.

UA Historical Earnings Volatility

Clicking on the UA symbol in OptionSlam’s historical stock screener

brings up the historical earnings statistics as of 2014-07-01. We discussed the
historical earnings statistics for UA earlier, when we were exploring how to
use the Basic spreadsheet in Chapter 3. Now that we are going to develop a
UA strategy, we will need to revisit the historical UA earnings statistics.
Rather than searching for the table in Chapter 3, the quarterly UA earnings
statistics are shown again in Figure 5.3. Except for the chart title, Figure 5.3
is identical to Figure 3.3.
 The 1-day closing percentage returns were calculated from the closing
price on the last trade day before earnings to the closing price on the next
trading day. You will recall that we entered the 1-day closing percentage
changes from Figure 3.3 (5.3) into the Basic spreadsheet, which calculated a
series of directional summary statistics and measures of historical earnings
volatility. The corresponding UA table from the Basic spreadsheet is
reproduced in Figure 5.4.
In Figure 5.4, there are three summary measures of the 1-day log returns.
The linear regression, median, and mean log-returns from the twelve previous
earnings events were +6.24% (C18), +3.11% (C19), and +2.41% (C20)
respectively. When combined with the Root Mean Squared Error (RMSE),
the resulting probabilities of a positive earnings return were 77% (E18), 64%
(E19), and 61% (E20). This indicates the presence of a positive historical
directional bias.
As shown in Figure 5.4, the annualized earnings volatility based on the
RMSE for the twelve previous UA earnings events was 136.76% (D22). We
will use this historical data for UA to create our own directional and volatility
forecasts, which will help us design an option earnings strategy for UA on
July 23, 2014.
 Reviewing the summary statistics for the past twelve earnings
announcements is a good start, but it is now time to revisit the issue that I
raised in Chapter 3. Based on a graph of the log-returns for UA, it was clear
that the variability of the earnings returns had increased recently. I recreated
this graph in Figure 5.5, but I added two rectangles that visually highlight the
recent increase in earnings volatility.
Note that the height of the rectangle surrounding the first six earnings
events is much shorter than the height of the rectangle encompassing the
second six earnings events. This is obviously a crude approach, but it does
make it easier to see the recent increase in earnings volatility.
 It is typically easier to identify patterns visually, but we also need to
quantify the recent increase in earnings volatility. Fortunately, if we clear the
contents of the six oldest earnings events, the Basic spreadsheet will
automatically recalculate all of the summary statistics for the most recent six
earnings events, including the historical earnings volatility. The UA earnings
table from the Basic spreadsheet for the most recent six earnings events is
provided in Figure 5.6.
The historical earnings volatility jumped from 136.76% (Figure 5.4 Cell
D22) to 168.18% (Figure 5.6 Cell D22), an annualized increase of over 30%.
I also used the Basic spreadsheet to recalculate the historical earnings
volatility for UA based on the most recent four earnings events. The
annualized implied volatility jumped again to 200.98%. I did not include the
Basic spreadsheet table based on the data for the four earnings events, but
you can easily replicate this calculation with your copy of the Basic
spreadsheet.
 The dramatic increase in historical earnings volatility from 136.76% over
twelve events, to 168.18% over six events, to 200.98% over four 4 events,
confirms our hypothesis that the UA earnings volatility had been increasing
prior to the July 2014 earnings announcement. The magnitude of the recent
increase in earnings volatility increased the likelihood of option pricing
anomalies prior to the July 2014 earnings announcement – anomalies that we
could exploit.
I will save you the suspense and reveal that the implied earnings
volatility (IEV) for UA on July 23, 2014 was only 137.2%. While the
market’s estimate of IEV was consistent with the historical earnings volatility
based on the twelve previous earnings announcements (136.76%), it failed to
recognize the large recent increase in earnings volatility. If the recent trend in
earnings volatility continued, then the market was grossly underestimating
UA’s earnings volatility prior to the July 2014 earnings announcement.

UA Directional Confirmation

We have established that UA’s IEV was understated on July 23, 2014 and
that there had been a persistent positive or bullish directional bias over the
previous three years of earnings events. Before we design an option earnings
strategy for UA, let’s attempt to independently confirm the bullish directional
forecast for the July 24, 2014 UA earnings announcement.
Analyzing the 1-day returns from past earnings announcements is the
most direct approach to quantifying directional bias, but we can also evaluate
earnings per share (EPS) and revenue forecasts. Stocks that beat their
consensus EPS and revenue forecasts are more likely to increase in price. If
we determine that UA was likely to beat its earnings and revenue
expectations on July 24, 2014, then that would support our case for a bullish
UA directional forecast.
Consensus revenue and EPS forecasts are widely available, but these
forecasts will not help us identify candidates that are likely to beat the
consensus. We need another data source that offers insights that go beyond
the consensus forecasts.
Estimize.com is a free (as I write this) online crowd-sourced platform that
aggregates EPS and revenue forecasts from investment professionals,
research analysts, statisticians, professors, individual traders, and even
finance students. If you are not familiar with the research on the “wisdom of
crowds,” then you would probably be surprised that aggregate estimates from
such a wide range of individuals could offer predictive value over Wall
Street’s consensus forecasts, but that is exactly the case.
Research studies have demonstrated that aggregate EPS and revenue
forecasts from the Estimize community are more accurate than those from
sell-side analysts and that it is possible to exploit these differences to generate
excess returns.
Estimize enhances their aggregate estimates by weighting each
community member’s forecast based on the analyst’s demonstrated accuracy
for that specific company. It is also possible to review the forecast for each
analyst and to rank the forecasts in order of the analyst’s historical accuracy.
Given this data, you could even develop your own EPS and revenue
aggregation model. The Estimize EPS and revenue forecasts are exactly what
we need to independently confirm our directional forecast.
The Estimize UA EPS and Revenue table from fiscal Q4 2012 through
fiscal Q2 2014 is provided in Figure 5.7 (with permission from Estimize). We
are interested in the fiscal Q2 2014 earnings data, which is located in the
column on the far right side of the table. Estimize does not offer
OptionSlam’s backtesting feature, so I have whited-out the actual EPS and
revenue data for the quarter.
As you can see from the table, the UA Estimize EPS estimate of $0.09 per
share exceeds the Wall Street consensus estimate of $0.08 per share. The
difference is not large, but the direction is consistent with the bullish
historical earnings return bias.
The difference between the UA Estimize revenue estimate and the Wall
Street revenue forecast is more significant. The Estimize aggregate estimate
of $595.40 million exceeds the Wall Street revenue forecast of $572.54
million by almost 4%. The Estimize EPS and revenue estimates both add
support for our bullish directional return forecast.
 The ability to evaluate the actual numerical values for EPS and revenue in
a table is critical, but it is often easier to identify patterns in data visually.
Figure 5.8 is a graph of the UA EPS data from fiscal Q4 2012 through fiscal
Q4 2014. I apologize if the chart is difficult to read, but it is a screen-capture
and I had limited control over the image. Fortunately, it is easy to read online,
but less so when reproduced. The solid green line represents the actual EPS
data, the light-blue dashed line illustrates the EPS estimate, and the dark-gray
dashed line depicts the Wall Street EPS estimate. The fiscal Q2 2014
observation is third from the right.
We are interested in evaluating fiscal Q2 2014 for UA, but in this case, I
did not white-out the most recent data. We already know from the table in
Figure 5.7 that the EPS and revenue estimates for Estimize supported our
bullish UA directional forecast for Q2 2014. This is more difficult to see in
the chart due to the small difference between the Estimize and Wall Street
EPS estimates.
The reason that I included the entire EPS history in the chart is because
there is a very prominent pattern in the data. Do you see it? Take a minute to
study the UA EPS chart in Figure 5.8.
Do you see the cyclical pattern in the data? UA is a retailer and retail is a
very seasonal business. The fiscal Q3 and Q4 EPS data is consistently higher
than the fiscal Q1 and Q2 data. This is not a random event. I emphasize this
point because it illustrates the potential for creating more sophisticated
directional earnings return forecasts that incorporate seasonal patterns.
Building a multi-variable directional earnings forecasting model is beyond
the scope of this book, but it is an exciting future research avenue that could
further expand your trading edge.
You can see from the EPS chart in Figure 5.8 that prior to fiscal Q2 2014,
UA had regularly beaten both Wall Street’s and Estimize’s EPS estimates. In
addition, Estimize’s estimates had been consistently more accurate than Wall
Street’s consensus forecasts.
 Figure 5.9 is a graph of the UA revenue data from fiscal Q4 2012 through
fiscal Q4 2014. The solid green line represents the actual EPS data, the light-
blue dashed line illustrates the EPS estimate, and the dark-gray dashed line
depicts the Wall Street EPS estimate. The fiscal Q2 2014 observation is third
from the right.
You will notice the same seasonal pattern in revenue that we saw in EPS.
In addition, note that in late 2013 and early 2014, UA’s actual revenues were
significantly higher than consensus revenue forecasts. Estimize’s revenue
forecast for fiscal Q2 2014 was significantly higher than the Wall Street
consensus and analysts had been consistently underestimating UA’s revenues
for several consecutive quarters. Both of these factors further strengthened
the bullish case for UA’s earnings announcement on July 24, 2014.
 Scenario Assumptions

We now have enough information to design an option earnings strategy

for UA on July 23, 2014. The next step is to enter this information into the
Integrated spreadsheet, which will use our forecasts in conjunction with the
comprehensive volatility model in an attempt to identify the UA option
strategy with the highest expected profit per unit of risk. For this example, I
specified a one-day holding period from July 23, 2014 to July 24, 2014 and
strategies requiring approximately $10,000 of capital (not shown). The
Integrated spreadsheet can accommodate a range of holding periods and
capital requirements. In this chapter, our discussion of the Integrated
spreadsheet will be limited to a few select screenshots.
Figure 5.10 is a screenshot from the “Specs” Tab of the Integrated
spreadsheet that shows the scenario assumptions used in all scenario and
optimization calculations. As was the case in the Basic spreadsheet, cells with
dark shaded backgrounds represent user inputs. Cells with white or light
colored backgrounds represent calculation cells and should not be modified.
Let’s review several of the most important user inputs.

Relative Value Correction

You will recall that the volatility model is used to identify the relative
overvaluation or undervaluation of every option in the matrix. The optimizer
will attempt to implement a strategy that exploits our volatility and
directional forecasts by buying undervalued options and selling overvalued
options. The “Percent UV/OV Correction” (C17) represents the amount of
undervaluation or overvaluation correction that occurs during the simulation
period.
For example, if an option is undervalued (cheap) by $0.20 and the
specified UV/OV correction was 100%, then the same option would be fairly
priced at the end of the simulation period. In other words, 100% of the initial
undervaluation would be realized. Similarly, if the specified UV/OV
correction was only 25%, then the same option would be undervalued by
$0.15 at the end of the simulation period; 25% of the undervaluation would
be realized ((0.20-0.15)/0.20).
This input variable gives the user a tremendous amount of control over
the simulation and optimization. When the holding period includes an
earnings event, a higher “Percent UV/OV Correction” would be justified. I
used 100% in the UA strategy example (C17).
Earnings volatility is arguably the greatest contributor to option pricing
anomalies. In addition, large price changes often occur following earnings
events, which forces market participants to take a fresh look at option values
across the entire matrix. As a result, assuming that 100% of the pre-earnings
option pricing anomalies are eliminated after earnings is not unreasonable. If
you prefer a more conservative assumption, I would suggest a value between
50% and 100%.
Conversely, if the holding period does not include an earnings event, then
the “Percent OV/UV Correction” should be much smaller, especially if the
holding period is short. When the holding period does not include an earnings
event, a value of 0% - 25% would be more appropriate.

Volatility Assumptions

Input cell (C16) represents the change in the 30-day at-the-money (ATM)
normal implied volatility (IV) of UA options in response to a 1% increase in
the price of the UA stock. The input value of negative 0.15 indicates that the
ATM IV will decrease by 0.15% per 1% increase in the price of the UA. For
example, if the UA ATM IV was 48% and the UA stock increased by 1%, the
UA ATM IV would decline to 47.85%. When scenario inputs are required,
the Integrated spreadsheet typically provides reference values to assist in the
development of the scenario forecast. In this case, the historical ATM
IV/price sensitivity was -0.26% (D16). This value is calculated elsewhere in
the spreadsheet. I reduced the magnitude of the forecast to -0.15% due to a
recent divergence from the historical value (not shown). We will explore this
concept more fully in the next few chapters.
The scenario assumptions also allow the user to specify a change in the
normal implied volatility for 30-day ATM options (F16). We do not have any
justification to forecast a change in the UA normal ATM IV for 30-day
options, but if we did, we would enter it here. The volatility model would use
the vertical and horizontal skew and the aggregate IV formula to calculate the
resulting IV changes for each option in the matrix.
Now it is time to enter our earnings volatility forecasts. The Integrated
spreadsheet calculated the historical earnings volatility of 136.76% (F19)
over the twelve previous earnings events, which was also calculated in the
Basic spreadsheet (Figure 5.4). Earlier in this chapter, we noted the dramatic
increase in historical earnings volatility from 136.76% over twelve events, to
168.18% over six events, to 200.98% over four events. Given the recent trend
in earnings volatility, I entered 152% (C19) as a conservative estimate of the
realized annualized earnings volatility, which was between the twelve quarter
earnings volatility (136.76%) and the six quarter earnings volatility
(168.18%).
The integrated spreadsheet also calculated an implied earnings volatility
of 137.2% (E19) using the comprehensive volatility model, which was
estimated using every option in the UA matrix. In addition to entering the
realized earnings volatility above, the Integrated spreadsheet also allows the
user to enter a new value for the implied earnings volatility (D19).
The realized earnings volatility affects the variability of the UA scenario
price changes over our one-day holding period, which will include the effects
of the earnings announcement. A user-specified change in implied earnings
volatility would affect the valuation of options that include multiple earnings
events (those with expiration dates beyond three months). The July 23, 2014
IEV of 137.2% was consistent with historical IEV values, so I used the value
of 137.2% in the scenario analysis.

Directional Assumptions

Earlier in this chapter, we discussed the directional bias of previous UA

earnings. The regression estimate (+6.237%), median (+3.109%), and mean
(+2.410%) log-returns for the twelve previous earnings events are all positive
(D21:F21), which indicated a bullish directional bias. We used the Estimize
estimates to confirm this directional bias was also present in UA’s EPS and
revenue forecasts for fiscal Q2 2014. Based on this information, I entered a
conservative directional forecast of +2.000% (C21).

Transaction Costs

Let’s ignore Row 21 of the Integrated spreadsheet for now and turn to
transaction costs, which have a major impact on the performance of all option
strategies. The Integrated spreadsheet allows the user to enter two variables
for transaction costs. The first value titled “Spread Slippage” represents the
difference between the purchase and sales prices relative to the mid-price of a
spread.
What is a spread? A spread is the simultaneous purchase and sale of two
options of the same type. In other words, a call spread involves buying one
call option and selling a second call option with a different expiration date,
strike price, or both. The optimization solutions will include some
combination of call spreads, put spreads, and straddles or strangles.
I have specified a Spread Slippage of negative $0.10 per spread, per
share, per contract (C25). If the mid-point of the spread was $1.25, a spread
slippage of negative $0.10 means that we would purchase the spread for
$1.35 and sell the spread for $1.15, which would result in a round-trip
transaction cost of negative $0.20 per spread. The resulting round trip
transaction cost of negative $0.20 would be multiplied by the number of
spread contracts and by the number of shares per contract.
For example, our optimal UA strategy will require 62 spread contracts. To
find the transaction costs resulting from slippage, we would multiply 62
spread contracts by 100 shares per contract, by negative $0.20. The total
round-trip transaction cost due to slippage would be negative $1,240.
Please note that the negative $0.10 spread slippage does not reflect the
quoted bid and ask prices for UA options, which would probably be much
wider than plus or minus $0.10 from the mid-price. As a rule, options should
never be sold at the quoted bid price or purchased at the quoted ask price.
Instead, limit orders should be used to execute option transactions near the
mid-point between the bid and ask prices.
In addition to spread slippage, the Integrated spreadsheet allows the user
to specify a second type of transaction cost: a per contract brokerage
commission. I have assumed a per contract commission of negative $0.75
(D25) per contract. The transaction cost calculation assumes that the negative
$0.75 per contract cost applies to both purchases and sales on every contract,
not on every spread.
The 62 spreads would include a total of 124 contracts. If we incurred a
transaction cost of negative $0.75 per contract purchased and negative $0.75
per contract sold, the total commission cost on 124 contracts (62 spreads)
would equal negative $186.00. When combined with the spread slippage
calculated above, the total round-trip transaction cost for the optimal UA
strategy would be negative $1,426.
For a strategy that only requires approximately $10,000 in total capital, a
round-trip transaction cost of negative $1,426 has a major impact on the
performance of the strategy. That is why it is expressly included in the
scenario analysis and optimization process.
 Optimal UA Solution

Figure 5.11 is a screenshot from the “Opt” tab of the Integrated

spreadsheet that shows the optimal UA solution found by the Solver
optimizer. There are eight columns from left to right: amount (number of
contracts), type (call or put), strike price, expiration date (YYMMDD), T-
Days (number of trading days until expiration), Mid Price, T-Delta (True
Delta), and Rich/Cheap (degree of over or undervaluation).
The call spread is shown in the first two rows, the put spread in the next
two rows, and the strangle or straddle in the last two rows. Let’s review the
call spread first. The optimal solution requires the sale of 11 call options with
a strike price of $67.50 and an expiration date of 10/18/2014 and the
simultaneous purchase of 11 call options with a strike price of $70.00 and an
expiration date of 8/16/2014. The mid-prices are shown for both call options
as are the number of trading days remaining until expiration and the True
Delta. The more interesting values are the rich/cheap values in the far right
column.
Negative values indicate options that were rich or overvalued. Positive
values indicate options that were cheap or undervalued. When relative values
correct or move toward fair value, rich options underperform and cheap
options outperform. The call option sold was rich or overvalued by $0.09 per
share, per contract. The call option purchased was cheap or undervalued by
$0.19 per share, per contract.
The optimal solution required the sale of 44 put options with a strike price
of $64.50 and an expiration date of 8/9/2014 and the simultaneous purchase
of 44 put options with a strike price of $62.50 and an expiration date of
8/30/2014. Let’s focus again on the rich/cheap values in the far right column.
The put option sold was rich or overvalued by $0.34 per share, per contract.
The put option purchased was cheap or undervalued by $0.23 per share, per
contract.
The final component of the optimal UA strategy was a diagonal strangle,
which required the purchase of 7 call options with a strike price of $65.00
and an expiration date of 7/26/2014 and the simultaneous purchase of 7 put
options with a strike price of $62.50 and an expiration date of 8/2/2014. In
this case, the call option purchased was cheap by $0.32 per share, per
contract and the put option purchased was also cheap by $0.09 per share, per
contract.
The optimizer will not always be able to purchase undervalued and sell
overvalued securities. In addition, the relative value correction is only one
component of value added. The optimizer attempts to maximize the
probability weighted expected profit of the strategy per unit of risk. The
expected profit is also heavily influenced by the volatility and directional
assumptions we specified earlier.
 The solution may look complicated, but only requires three spread
transactions, which makes it relatively easy to manage. However, it would be
virtually impossible to find this solution without the optimizer.

P&L Graph

Even for an experienced option trader, the ability to fully appreciate the
nature of a strategy by mentally combining the three spreads would be very
difficult. That is why we use profit and loss diagrams, which are common to
all option analytical platforms. Figure 5.12 is a profit and loss (P&L) diagram
for the optimal strategy described in Figure 5.11. You are probably familiar
with P&L functions, but the P&L function in figure 5.12 is quite different.
The scenario values accurately incorporate the effects of earnings volatility
and normal volatility for every option in the matrix, as do the True Greeks.
The independent x-axis represents the UA stock price in dollars. The
dependent y-axis on the left side of the diagram denotes the profit and loss of
the optimal strategy for three different dates.
The solid green upper “P&L 0” represents the P&L at time zero, or the
instantaneous profit and loss of the strategy. The solid purple lower “P&L U”
line represents the P&L on the future date specified by the User, which is the
date used to optimize the strategy. In this example, the User date was one
trading day into the future, or July 24, 2014. The solid black lower “P&L X”
line represents the P&L on a second future date specified by the User, which
is intended to be the expiration date of the option with the shortest-term
option.
In this example, the “X” date was two trading days into the future, or July
25, 2014. Given that the “U” and “X” dates are only separated by one day,
the lower lines overlap and only appear as a single solid (lower) black line in
Figure 5.12.
The “U” and “X” probabilities are shown by the faded dashed lines and
correspond to the values on the dependent y-axis on the right side of the
diagram.
In addition to the P&L graph, I have included summary statistics for the
optimal strategy at the top of the diagram.
Let’s examine Figure 5.12 and see what we can learn about the strategy.
The optimizer always uses the User date, so let’s initially focus our attention
on the lower P&L line. My first observation was that virtually the entire P&L
line was above the break-even or zero profit line. The price of UA before
earnings on July 23, 2014 was $60.63.
The optimal strategy would have only underperformed over a very small
range of UA prices, centered slightly below the current price. Given that the
P&L line included round-trip transaction costs of negative $1,426, the
strategy performance was remarkable. This demonstrates the potential power
of selling overvalued options and buying undervalued options.
The shape of the P&L function was convex or curved upwards, indicating
the strategy would have performed well for large moves up or down in the
price of UA. You will recall that Gamma measures the curvature of the
option price function, so based on the shape of the diagram, we would expect
the optimal strategy to have positive Gamma and it does. The True Gamma
was +132.3 and the traditional Black-Scholes (BS) Gamma was +97.2.
The probability function was centered above the UA price of $60.63,
which reflects our bullish expected return of 2% (Figure 5.10, cell C21). The
width or dispersion of the probability curves is a function of the UA normal
and earnings volatility.
What explains the difference between the upper “P&L 0” line and the
lower “P&L U” and “P&L X” lines? The difference is quite large, especially
near the current UA price. The positive value of Gamma means that the
strategy must have had a negative Theta. The True Theta was negative
$1,415.5, which fully reflected the volatility crush due to passing the UA
earnings date. It is interesting to note that the Black-Scholes (BS) Theta was
only negative $74.70, which grossly understated the true cost of the strategy
due to the passage of time.
The True Delta was +$24.8 which reflected a small bullish bias, but the
BS Delta was actually negative (-$53.6), again misrepresenting the true risk
of the strategy. The strategy had a positive True Normal 30 Vega (+$ 19.10)
and a positive True Earnings Vega (+$19.3), which is consistent with our
forecast of a higher than expected level of earnings volatility.
As stated above, the total transaction costs were negative $1,426. The
expected profit (net of transaction costs) was $1,482 and the maximum loss
(net of transaction costs) was negative $137. The objective function, which is
what the optimizer was attempting to maximize, was the expected profit
($1,482) divided by the absolute value of the maximum loss ($137).
The resulting value of the objective function for the optimal strategy was
an astounding 10.78, which means the probability-weighted expected profit
from the strategy was 10.78 times the maximum loss. The objective function
value of 10.78 represents the expected profit per unit or dollar of risk. In this
case, risk was defined as the maximum loss (net of transaction costs) of the
strategy in any of the discrete scenarios ($137).
So does that mean that the optimal strategy could not have lost more than
$137 over the one-day holding period? No, unfortunately the strategy could
have lost more than $137. It is important to remember that the profits and
losses for each of the 100 probability-weighted scenarios were based on our
input assumptions.
For example, we assumed that 100% of the undervaluation or
overvaluation would correct or be eliminated during the one-day holding
period. We also assumed that normal implied volatility and implied earnings
volatility values would not change.
All of our input assumptions will affect the profits and losses of each of
the discrete scenarios. Fortunately, we have complete control over these input
values and we can use the Integrated spreadsheet interactively to examine the
sensitivity of the solution to each input assumption. The more you use the
spreadsheets, the more you will learn about your strategies.
 Confirmation

Just as we used supplemental evidence from Estimize.com to confirm our

directional earnings forecasts, it makes sense to corroborate our analytical
results as well. Unfortunately, the earnings volatility framework presented in
this book is new and is not widely used by the industry.
However, while writing this book I advised OptionVue on how to use and
apply the aggregate implied volatility formula to quantify the effects of
earnings volatility before and after earnings events. They subsequently
integrated my earnings volatility framework into their OptionVue software.
The spreadsheets included with this book and OptionVue’s software modules
are the only commercial tools that I am aware of that currently use my
earnings volatility framework.
As a result, I was able to enter the optimal UA solution from Figure 5.11
into the OptionVue software and performed a graphical analysis of the
strategy on July 23, 2014. OptionVue’s analytical results are depicted in
Figure 5.12B. OptionVue’s results do not include the directional or earnings
volatility forecasts that we developed throughout this chapter.
The UA stock price is shown on the independent x-axis and the strategy
profit or loss (in dollars) is shown on the dependent left-vertical axis. The
strategy return as a percentage of required capital is shown on the dependent
right-vertical axis. The dashed line represents the strategy results for the
“T+0” line, which stands for trade date plus zero days (instantaneous P&L).
The solid line depicts the strategy results for the “T+1” line, which represents
the P&L for a one-day holding period. These two lines are directly
comparable to the strategy results shown in Figure 5.12.
OptionVue also includes two horizontal bars at the bottom of the P&L
graph. The shorter top line represents a one-standard deviation price move in
the common stock of the underlying security for the time horizon selected, in
this case one trading day (T+1). The longer bottom line represents a two-
standard deviation move for the same holding period. The one and two-
standard deviation forecasts both use the aggregate implied volatility formula
to accurately incorporate the effects of earnings and non-earnings volatility
over the specified holding period.
While the Integrated spreadsheet model and OptionVue’s model both use
the aggregate implied volatility formula, the other components of the two
models are completely independent: estimates of earnings volatility,
directional bias, normal volatility, vertical skew, horizontal skew, and the
IV/price sensitivity. The Integrated spreadsheet and OptionVue results both
include estimated slippage and transaction costs. Despite the model
differences, the P&L forecasts are quite similar and both clearly support the
transaction.
 After publication, I plan to work with OptionVue further to help them
integrate any additional analytical concepts from this book that are applicable
to their software. Please see the resource section at the end of this book for
additional information on OptionVue’s products and services, including a
discounted subscription offer for Exploiting Earnings Volatility readers.
 Actual Strategy Results

The value of this analytical framework should not be evaluated based on

the results from a few strategy examples, but curiosity alone dictates that we
ask how the strategy actually performed and how accurately the framework
modeled the strategy results for the actual change in the price of UA on July
24, 2014.
Figure 5.13 is the same P&L graph from Figure 5.12 with one addition.
The Red diamond represents the actual profit of the optimal strategy (net of
transaction costs) from July 23, 2014 to July 24, 2014. The price of UA
increased from $60.63 to $69.53. Our bullish forecast was correct (but
understated) and the earnings volatility was also much greater than the
implied earnings volatility.
As you would expect, the optimal strategy performed well, earning a total
profit of $4,920 (net of transaction costs). The return on required capital of
$9,044 (not shown) was 54.40% ($4,920/$9,044), not bad for a one-day
holding period. At a price of $69.53, the model forecasted a strategy value of
$3,780, meaning that the actual strategy performance exceeded the expected
model performance. Given our assumption of a 100% UV/OV Correction,
this is a little unusual, but the large change in the price of UA (+14.7%)
probably increased the variability of the prices after earnings.
Conclusion
 The purpose of this chapter was to provide a practical example of how to
use the Exploiting Earnings Volatility analytical approach and toolset to
design a realistic earnings strategy, based on actual market prices and actual
earnings data. The emphasis in this chapter was on the investment process
and the resulting strategy, but we glossed over the mechanics of the
Integrated spreadsheet.
In the next three chapters, we will examine the steps required to use the
Integrated spreadsheet in practice. We completed a similar exercise in
Chapter 3 for the Basic spreadsheet. However, the Integrated spreadsheet is
far more complex than the Basic spreadsheet, which is why it will take a few
chapters to cover the material.
While the Integrated spreadsheet functions are complex, the functions are
automated and accessible via “push-button” macros, which should make the
process easier to understand and apply.
 6 – Integrated: Input & Import

This is the first of three chapters that will provide step-by-step instructions on
how to use the second of two spreadsheets that accompany this book: the
Integrated spreadsheet. Unlike the Basic Spreadsheet, the Integrated
spreadsheet uses macros and requires a more recent version of Microsoft
Excel to function properly. As was the case with the Basic spreadsheet, the
Integrated spreadsheet is populated with data for Under Armor (UA) as of
July 23, 2014.
The purpose of the next three chapters is to explain how to use the
Integrated spreadsheet, not how to construct the Integrated spreadsheet or
how to perform every intermediate calculation. If you have an interest in the
intermediate calculations, please review the formulas in the spreadsheet cells.
There are two tabs that are used in both the Integrated and Basic
spreadsheets that are very straightforward and will not be discussed in detail.
The first is the “Holidays” tab (not shown), where you will enter the dates of
exchange holidays (B3:B24) that will be used by the day-count algorithms to
calculate the number of trading days throughout the Integrated spreadsheet.
The other self-explanatory tab that will not be discussed is the “Saved”
tab (not shown). It is not used directly in any calculations; instead, it is
provided as a convenient location to copy and paste data from other tabs that
you might want to use again in the future. This will save you a great deal of
time if you continue to evaluate the same stocks quarter after quarter.
There is a third tab that is included in both the Basic and Integrated
spreadsheets (the “HistoricalEV” tab), but it was already discussed
extensively in Chapter 3. As a result, we will not repeat that explanation in
this chapter. However, please be aware that the row and column numbers in
the “HistoricalEV” tab are slightly different in the two spreadsheets, but the
functionality is the same.
As you read the next three chapters, keep in mind that the goal of the
Integrated spreadsheet is to provide a practical tool that will help you exploit
volatility, directional, and relative value anomalies to develop, optimize, and
evaluate option earnings strategies in real-world environments. Before we
delve into the minutia of the Integrated spreadsheet, I would like to provide
an overview that should help you put the individual steps in the proper
context.
 Overview

Before we can use the Integrated spreadsheet to design strategies, we

must provide information that will allow the Integrated spreadsheet to do all
of the required analytical calculations. This information includes user
specifications such as analysis dates and filter settings. Filter settings are used
to determine which options are used to estimate the volatility model and
which options are available as strategy candidates.
In addition, we will also have to provide information that is specific to the
underlying security, such as the next earnings and dividend dates, dividend
yield, and the multiplier (shares per contract). Finally, we will need to
provide historical price and implied volatility data to model the ATM
IV/price sensitivity.
After providing the information above, we will use the automated import
function (button macro) to pull in all of the required matrix data, apply filters
and error checking, and then sort and copy the resulting data to all of the
analytical tabs of the Integrated spreadsheet.
We will then use the automated Solver function (button macro) to solve
for the volatility parameters (including earnings volatility) that minimize the
sum of the squared pricing errors across all of the options in the matrix.
Automated simulation functions (button macros) will use the volatility
parameters and user-specific volatility and directional forecasts to calculate
scenario-specific values for every option in the matrix, which will be used for
optimization, risk management (True Greeks), and graphical analysis.
We will then enter several optimization parameters before using our
choice of automated Solver functions (button macros) to identify the option
earning strategy that delivers the highest expected return per unit of risk.
Finally, we will evaluate the profit and loss function and the True Greeks
of the resulting strategy, which will enhance our understanding of risk and
return.
The remainder of this chapter will focus on the steps required to enter
data and import the matrix. Chapter 7 will explain how to use the macros to
estimate the volatility model and generate the simulation data. Chapter 8 will
describe how to use the macros to optimize strategies and analyze the results.

ATM IV / Price Sensitivity

 As explained in Chapter 5, normal implied volatility responds to changes
in the price of the underlying security. Unfortunately, the magnitude and even
the direction of the changes in normal implied volatility vary widely among
securities and across time.
As a result, we need to supply a representative value for the ATM
IV/price sensitivity that can be used to calculate option values for each of the
discrete scenarios. The ATM IV/price sensitivity represents the change in the
30-day at-the-money (ATM) normal implied volatility (IV) of UA options in
response to a 1% increase in the price of the UA stock.
Before we can enter an assumed ATM IV/price sensitivity in the
Integrated spreadsheet, we need to examine historical ATM IV/price
sensitivities for UA, which requires historical data. We enter this data in the
“IVPrice” tab of the Integrated spreadsheet, which performs all of the
required sensitivity or slope calculations automatically.
Before we copy and paste the implied volatility and price data into the
“IVPrice” tab, we need to clear the existing values, which are probably from
an earlier analysis. To facilitate this process, I created a button with an
assigned macro titled “Clear IV & Price Data (~30 Sec.)” (Figure 6.1). The
“30 Sec” indicates that it takes roughly 30 seconds to run the macro, which
surprised me the first time it ran. Additional time is required due to the use of
Excel’s lookup function to align the price and volatility data.

After clearing the data, the next step is to copy and paste the daily
historical price and volatility data from your data source into the appropriate
cells of the “IVPrice” tab. Your broker may be able to supply this data and it
is available from a number of third-party data vendors, including OptionVue.
Figure 6.2 only shows the first few rows of price and IV data from the IV
price tab, but thousands of rows are available to store many years of
historical data. As is the case throughout the Basic and Integrated
spreadsheets, the blue shaded backgrounds with white text indicate input
cells. The input data begins in row three. The historical pricing dates are
entered in Column O and the adjusted closing prices for UA are entered in
Column P.
What are adjusted prices? Proportional adjustments should always be
made when analyzing historical prices, which will eliminate discrete price
changes due to dividends and splits. These large discrete price jumps would
create large undesirable and unrealistic changes in the ATM IV/price
sensitivities. In addition, please ensure that you correct any data errors and
missing data points. The spreadsheet will not do this for you.
The historical IV dates are entered in Column R and the 30-day ATM
normal IVs for UA are entered in Column S. If you do not have access to
normal IVs, then use implied volatilities. However, the IV values should
always be derived from one-month at-the-money (ATM) options. The
volatility model will automatically calculate the resulting normal IV changes
for all of the options in the matrix based on the estimated horizontal and
vertical skew parameters.
The price and IV dates do not need to line up in the spreadsheet. As
described above, Excel’s lookup function will align the data correctly.
 There are two additional user inputs required (Figure 6.3) to calculate the
ATM IV/price sensitivities. The first value (D1) represents the number of
trading days used to calculate the change in IV and the change in price. The
second input value (G1) represents the number of observations used to
calculate the ATM IV/price sensitivity or slope. I recommend using 22 for
both values, regardless of your holding period.
Calculating price and IV changes over 22 trading days (roughly one
calendar month) results in significant changes, which is useful when
estimating an ATM IV/price sensitivity that will be applied to pricing
scenarios that span a wide range of prices. Using 22 observations to calculate
the slope eliminates some of the day-to-day variability that would result from
using fewer observations.
Many of the cells in Figure 6.3 are intermediate calculations and therefore
do not require further explanation. However, there are a few values that are
relevant. The most important is the average slope of -0.263% (I1), which
represents the average slope calculation for UA over the entire price and IV
history. We saw this value earlier when we designed the option earnings
strategy for UA in Chapter 5. You will recall that this value was provided on
the “Specs” tab (Figure 5.10 cell D16) to give us a reference point when
entering our own ATM IV/price sensitivity assumption.
In addition to the average ATM IV/price sensitivity -0.263% (I1), the
ATM IV/price sensitivity is also calculated for every pricing date in the
spreadsheet and the daily values are provided in Column G. The ATM
IV/price sensitivity for UA on our analysis date of July 23, 2014 was
+0.214% (not shown), which was quite different from the historical average
of -0.263%. I entered an assumed ATM IV/price sensitivity of -0.15%
(Figure 5.1 cell C16) when we designed the UA strategy in Chapter 5. The
value of -0.15% was closer to the historical average of -0.263%, but shaded
slightly higher to acknowledge the ATM IV/price sensitivity of +0.214% on
July 23, 2014.
There are two more relevant values in Figure 6.3. The minimum (A1) and
maximum (B1) UA returns over any 22-day (D1) period are also provided on
the “IVPrice tab” and again on the “Specs” tab (Figure 5.10 Cells E23:F23).
These values will be useful when setting the range of return values where the
ATM IV/price sensitivity will apply (Figure 5.10 Cells C23:D23). To avoid
unusual implied volatility forecasts (or negative IV values), I always ensure
that my specified range is well inside the historical return values. For the UA
analysis, I entered a return range of plus and minus 30% (Figure 5.10 Cells
C23:D23). We will revisit these assumptions again in the next chapter.
 Dates

The next step in using the Integrated spreadsheet is to enter the required
dates. All of these dates are entered in the “Specs” tab (Figure 6.4). The first
is the “Analysis Date” of July 23, 2014 (H2), which is also the pricing date.
The spreadsheet assumes closing prices and does not calculate fractional
days, although this is something that could be implemented in the future and
would be particularly beneficial when evaluating short-term options.
The “1st User Date” is the date used for optimization purposes. For UA, I
specified a User date of July 24, 2014 (H4). As was the case above, closing
prices are assumed, which means the holding period for the UA optimization
was one trading day.
The “2nd User Date” was designed to represent the date of the earliest
option expiration. It is important to note that the optimizer will not buy or sell
any options that expire before the first or second “User Dates.” Since we do
not yet know which options will be included in the optimal strategy, I
recommend entering a “2nd User Date” equal to the “1st User Date” or equal
to the first option expiration after that date.
In the UA example, the “2nd User Date” was July 25, 2014 (H5), which
corresponds to the expiration date of the first weekly option series. After
solving for the optimal strategy, you could always return to the “Specs” tab
and enter a different “2nd User Date.” As long as the “2nd User Date” falls on
or after the option expiration date of the nearest-term option used in the
optimal strategy, changing the date will not affect the optimization result.
Both “User Dates” should obviously occur after the “Analysis Date.”

In addition to the “Analysis Date” and “User Dates” on the “Specs” tab,
we also need to enter the date of the first earnings announcement (July 23,
2014 in Cell H7) and the first dividend date (August 7, 2014 – H16) that fall
after the “Analysis Date.” It is extremely important to note that the earnings
announcement date must correspond to the date of the close immediately
preceding the next earnings announcement, not the date of the actual
earnings announcement. In this case, the UA earnings were scheduled to be
released before the open on July 24, 2014. However, if you entered an
earnings date of July 24, 2014, the Integrated spreadsheet would assume the
earnings announcement occurs after the close on July 24, 2014. As a result,
you must always enter the date of the close immediately preceding the next
earnings announcement.
As of July 23, 2014, UA did not pay dividends, so the first ex-dividend
date of August 7, 2014 (H16) was arbitrary and has no effect on the analysis.
The next earnings date (H7) and next dividend date (H16) are entered on the
“Specs” tab.
The Integrated spreadsheet assumes that earnings will be announced
quarterly and will automatically populate the subsequent seven quarterly
earnings announcement dates. The Integrated spreadsheet will use these dates
to determine the number of earnings trading days for each option in the
matrix. Unlike the Basic spreadsheet, the Integrated spreadsheet calculates
the actual number of earnings trading days for every option.
 The Integrated spreadsheet also populates the dates for the subsequent
seven ex-dividend dates. These dates are calculated based on the number of
dividend payments per year (4), as specified in Cell D2 of the “Specs” tab,
which is the first of the General Specs we will discuss (Figure 6.6).

General Specs

Another General Spec is the annual dividend yield (0%), which is entered
in Cell F2. As mentioned above, UA did not pay dividends, so the annual
dividend yield was 0%. If a company does pay dividends, those future
dividends will affect the value of the company’s options. When a stock
passes its ex-dividend date, the price of the stock declines and this effect
must be included when modeling option prices and implied volatilities.
Three additional General Specs are included in Figure 6.6: the annual
risk-free interest rate, the option multiplier, and the American option flag.
The risk free interest rate (0.25%) is entered in Cell C3 and is used in option
valuation and risk analysis. The option multiplier (100) is entered in Cell C5
and is important in the calculation of the True Greeks and scenario values. In
the case of UA, the American flag (Y) is entered in Cell E5, which signifies
that (Yes) UA options may be exercised prior to expiration. A discrete
binomial lattice that models optimal exercise at each node is required to
accurately value American options.
Due to the use of Normal and Earnings volatility, which are not constant,
the resulting binomial lattice will not recombine, which makes this solution
method impractical. The number of nodes in a non-recombining binomial
lattice increases exponentially. However, the American flag does impose
some crude boundary conditions during the option valuation process.
Please note that many of the user input values in Figure 6.6 will not
change when you evaluate different stocks. Most companies pay dividends
quarterly, the risk free interest rate is not company specific, the standard
option multiplier is 100, and most stock options are American. All of the
remaining values in the table will be calculated automatically.
 Estimation Filter Specs

We are almost ready to import the file, but we still need to specify two
sets of filters (Figure 6.7). The minimum and maximum values for the first
set of filters (C7:D14) will determine which options will be included when
solving for the volatility parameters.
The maximum number of trading days until expiration is the only user
input required in the second set of filters (F9). The minimum number of
trading days remaining until expiration will also be used as a candidate filter,
but this value will be determined by the second “User Date” we entered
earlier (Figure 6.4 Cell H5). As I mentioned at that time, the optimizer will
only consider options that expire on or after the “User Dates.”
The two sets of filters serve two very different purposes. The objective in
specifying the filters that will determine the options used in volatility
modeling is to be as inclusive as possible, while preemptively avoiding
potential pricing errors that could compromise the integrity of the volatility
model. By definition, any options excluded from the volatility estimation
would not be available as strategy candidates either.
Let’s review the filters used in the UA analysis on July 23, 2014 (Figure
6.7). Columns C and D represent the minimum and maximum acceptable
filter values for the volatility estimation. I have specified a minimum option
price of $0.20 (C7). Low priced options are typically either very short-term,
out of the money, or both. In such cases, large changes in implied volatility
are required to affect the option values. If these options were mispriced, even
by seemingly insignificant amounts, the resulting implied volatility errors
would be very large, which would compromise the accuracy of our volatility
model parameters.
Notice that the maximum filter for the option price is grayed out (D7),
which means that it is not applicable and should not be used during the
filtering process. This will be true for several of the filter rules. Remember
that user input cells are always shown with a blue shaded background with
white text.
I have entered two days as the minimum number of trading days (C8) and
504 as the maximum number of trading days (D9) remaining until expiration
for the estimation filters. To ensure an accurate volatility model, leaps should
be included. I probably could have excluded the maximum number of trading
days filter entirely. Unless long-term options are systematically and
dramatically mispriced, they should be included in the volatility estimation.
I included liquidity filters for minimum volume (C10) and minimum open
interest (C11), but both have zero values and I typically do not recommend
filtering on volume or open interest. Even options with zero open interest or
volume can be quite liquid. Liquidity is more a function of the total option
volume and total open interest than a function of the volume or open interest
of an individual option in the matrix. Obviously it would not make sense to
apply maximum volume and open interest filters and these values are grayed
out.
I entered the minimum and maximum Call Delta (per contract) filters of
+4.0 (C12) and +96.0 (D12) respectively. The rationale for these filters is
similar to the argument used for the minimum option price. Given an option
multiplier of 100, the minimum and maximum possible Call Deltas (per
contract) are zero and 100.
By excluding call options with very high or very low Deltas, we eliminate
options that are extremely in or out of the money, both of which are impacted
little by changes in implied volatility. As a result, if these options were
mispriced, the resulting implied volatility errors would be very large, which
would compromise the accuracy of our volatility model parameters.
The same argument applies to extreme Put Deltas. Put Deltas are
obviously negative and I have filtered out put options with (per contract)
Deltas lower than -96.0 (C13) or higher than -4.0 (D13). Finally, I have
limited the maximum estimation skew (the vertical skew coefficient) to a
value of 1.0 (D14). I will not explore this value here, other than to state that
capping the maximum vertical skew coefficient avoids extreme and unusual
volatility skew solutions and this value should not be modified. The vertical
skew coefficient will be explained further in Chapter 7.
 Scenario Filter Specs

I specified a maximum of 75 trading days remaining until expiration for

the scenario filter (Figure 6.7 Cell F9). Why would I exclude all of the option
candidates with greater than 75 trading days remaining until expiration?
There are three reasons: liquidity, the relative impact of earnings volatility,
and practical optimizer considerations.
By limiting strategy candidates to options with fewer than 75 trading days
remaining until expiration, the entire weekly series and the first three monthly
options should all be included. Near-term options tend to be more liquid and
are also disproportionally affected by earnings volatility. Since we are
attempting to design option earnings strategies, it makes sense to ensure we
use the most liquid options that are most likely to exploit earnings-related
mispricing anomalies.
The other argument for eliminating option candidates with greater than 75
trading days remaining until expiration is to make it easier for the optimizer
to find a solution. We will discuss strategy optimization extensively in
Chapter 8, but I designed the Integrated spreadsheet to work with the free
version of Solver, which has significant limitations regarding problem size,
speed, and solution algorithms. Limiting the number of candidates makes it
easier for the free version of Solver to find good solutions in a reasonable
period of time.
The Integrated spreadsheet is complex and has many input fields.
However, most if not all of the input fields in Figure 6.7 are not company-
specific. In other words, once you enter your desired filter specifications, you
could use the same specs for every stock and every option matrix. As you
gain more experience with the spreadsheet, you will probably tweak the filter
specs from time to time, but you will not need to change them every time you
use the spreadsheet.

Import File Settings

I created a convenient import function (button macro) that will pull in all
of the required matrix data, apply filters and error checking, and then sort and
copy the resulting data to all of the analytical tabs of the Integrated
spreadsheet. Before we can use that function, we need to specify the location
of the import file and define the file specifications.
It should be no surprise that we will enter the file specifications on the
“Specs” tab (Figure 6.8). To open and import the file, the import macro
requires a file path and name, which is defined by the following user fields:
symbol (UA), file path (C:\Users\Brian2012\Desktop\OVFiles\), pricing date
(20140723), pricing time (1600), file type (.CSV). These user inputs are
entered in Cells B29:B33. The Integrated spreadsheet then aggregates the
input fields to create a filename (B28) and a complete file path & file name
(B27), which is used by the macro.
Your file path will obviously be different, but I suggest placing all of your
import files in the same directory to make it easier to use the import function.
Since each file is time and date specific, you can even have multiple files for
the same symbol in the same directory. The import function reads comma-
separated-value or CSV files.
 Import File Format

Before we can use the import macro, we must ensure the file is in the
proper format and the data fields are in the correct order. I am a paying
customer of OptionVue and I use their option platform to export CSV files,
which I import directly into the Integrated spreadsheet using the import
function (button macro). Trading Insights also has an affiliate relationship
with OptionVue; please refer to the Resources section at the end of this book
for additional information on OptionVue’s discount offer.
Figure 6.9 is a screenshot of OptionVue’s ASCII Export dialog box (with
permission from OptionVue). Note the selected asset is UA and the
OptionVue export file path and name exactly matches the import file path
and name in Figure 6.8 of the Integrated spreadsheet. If you decide to use
OptionVue to generate your import files, please use the settings from Figure
6.8.
 The good news is that it is quite easy to use the OptionVue file export
function in conjunction with the import function in the Integrated
spreadsheet. The bad news is that we also need to define the precise format of
the OptionVue export file, to ensure that it will work correctly with the
import function in the Integrated spreadsheet. Fortunately, this only needs to
be done once.
To define the OptionVue export file format, click on the “Line Format”
button in Figure 6.9. That will bring up OptionVue’s Line Format screen
(Figure 6.10). The right hand side (RHS) of the screen provides a list of all
available parameters. The left had side (LHS) of the screen includes the list of
parameters included in the export file, in the order listed.
 Simply use the buttons to move the desired fields from the RHS to the
LHS and move the parameters up and down until they match the order
required by the import function. The actual OptionVue parameters (in the
correct order) required by the Integrated import function are provided in
Figure 6.11.
Once you create your desired export file format in OptionVue, all of your
export files will automatically be created in this format. You will not need to
use the “Line Format” button every time you export a file.
 Data Field Formats

If you are using OptionVue to generate these files, you can proceed to the
import functions. However, if you are using another vendor or your broker to
create these files, then we need to discuss the format further.
Figure 6.12 is a screenshot of the left-hand-side (LHS) of the first 26 lines
of the “IMPORT” tab of the Integrated spreadsheet (after importing the UA
file). Figure 6.13 is a screenshot of the right-hand-side (RHS) of the first 26
lines of the “IMPORT” tab of the Integrated spreadsheet. The number of
fields (columns) made it impossible to include all of the “IMPORT” data
fields in a single screenshot. Please refer to the screenshots of the “IMPORT”
tab as I review the format requirements for each import field in the CSV file.
The CSV file begins with a header row, with one title for each of the
fields or columns. The first row of actual data must always be for the
underlying stock. If not, then the import function will not work properly.
The first column (existing position) is not imported into the spreadsheet,
but must be present in the import file. The next two columns are symbol
(Column B) and description (Column C), which should not require a specific
format. The analysis date field (Column D) is very important and it must use
the YYMMDD format. The accompanying time field (Column E) does not
require a specific format.
The strike price field (Column F) is next and should include a decimal
place when needed. The next field (Column G) is used to designate a Call (C)
or Put (P) option; note the use of capital letters. The expiration date field
(Column H) is also very important and must also use the YYMMDD format.
The MID IV field (Column I) represents the implied volatility at the mid-
price between the bid and ask prices. Decimals must be included and a value
of 0.895 would represent an annualized implied volatility of 89.5%.
The next three fields are Theoretical Price (Column J), Bid Price (Column
K), and Ask Price (Column L), all of which should include decimals when
necessary. The values in these columns represent option prices per share, not
per contract. All of the above fields are shown in Figure 6.12.
 The remaining fields are shown in Figure 6.13. The Percent O/U field
(Column M) represents the percent overvalued or undervalued. The next two
fields are Volume (Column N) and Open Interest (Column O). The next five
columns represent the Greeks: Delta (Column P), Gamma (Column Q), Vega
(Column R), Theta (Column S), and Rho (Column T). All of the Greek
values are per contract, not per share. In other words, the per contract Greeks
equal the per share Greeks multiplied by the option multiplier. The next field
(Column U) is the probability of finishing in the money (ITM) and the last
field (Column V) is the last transaction price.
Given that the Integrated spreadsheet calculates the True Greeks, it might
seem strange to import Greek values from an outside source, but we have a
chicken and an egg problem. It is beneficial to use Delta to filter the data
before estimating the volatility model, but we can’t calculate the value of
True Delta before estimating the volatility model. That is why we need to
import this data. The same reasoning applies to the Mid IV data.
In addition, some of the import data is not used by the Integrated
spreadsheet. I use the same OptionVue export file format for different option
research projects and all of the data in the Export file is not required by the
Integrated spreadsheet. However, the column format and order is important,
because the Integrated spreadsheet retrieves the required data from specific
columns.
The most important data fields are all date fields, all price fields, strike
price, call/put, MID IV, Delta, Vega, and volume and open interest (if you
plan to filter on those fields). The description and symbol fields are important
for identification purposes, but not for analytical purposes.
 Import Functions

Now that we have an import file in the proper format, we can run the
import function. Actually, there are several Import functions (button macros),
all of which are located on the LHS of the “IMPORT” tab (Figure 6.12). The
good news is that you should only need to use the last import function titled
“IMPORT Combo,” which performs all of the import functions in the proper
sequence.
When I created the spreadsheet, I wrote all of the import-related macros
individually. The top five buttons on the LHS of the “IMPORT” tab
correspond to the five individual macro functions performed by the
“IMPORT Combo” macro discussed above. All of the functions must be
performed in the proper sequence before beginning the analytical process.
Even though you should not need to use the individual macros, I left the
individual macro buttons in the spreadsheet to allow me to identify the source
of any errors that might occur during the import process.
Here is a brief explanation of the individual macro functions. The
“IMPORT CSV” macro does exactly what its name implies; it imports all of
the matrix data from a correctly formatted CSV file into the appropriate
columns of the “IMPORT” tab.
The “User Filter” macro function applies the filter settings from Figure
6.7 to the option matrix, eliminating any options that do not meet the user-
specified filter criteria.
The “Error Filter” macro applies some basic error-checking procedures to
the remaining options and removes any options that are flagged during the
process. For example, in order to apply the volatility model, an implied
volatility must be calculated for each option. If a large pricing error prevented
the option model for solving for implied volatility (due to an option price
below the intrinsic value for example), then that option would be excluded.
Keep in mind that these error filters can only identify glaring pricing errors
and that the user is responsible for ensuring the integrity and accuracy of the
data. Inaccurate or incorrectly formatted data will result in errors, invalid
results, or both.
The “OPT Sort” function performs a multi-layered sort on all of the
options in the matrix that satisfy all of the filters and pass the error-checking
process. The options are sorted first by type with calls at the top and puts at
the bottom. Within the calls and puts, the options are sorted in ascending
order of trading days remaining until expiration (shortest to longest). The
final layer sorts the options in descending order of strike price.
All of the analytical tabs in the Integrated spreadsheet require a common
dataset. “Copy IMPORT” is the final import-related function, which copies
the filtered, error-checked, sorted option matrix from the “IMPORT” tab to
each of the analytical tabs of the Integrated spreadsheet.
All of the import-related macro functions (including the sort) must be
performed before beginning the analytical processes. To reduce potential
problems with the analytical functions or the optimizer, I suggest that you
always run the “IMPORT Combo” macro function and ignore the individual
import macro functions. The “IMPORT Combo” macro function runs the
individual import-macro functions in the proper order and prepares the
Integrated spreadsheet to perform all of the analytical calculations.

Summary

At first glance, the amount of input data required by the Integrated

spreadsheet might seem daunting; so let’s review the day-to-day input
process, assuming that we have already developed procedures to supply the
necessary data in the proper format - which is a one-time project.
The first step is to copy and paste the price and IV history. We then need
to enter the three analysis dates, the next earnings date, the next dividend
date, and the dividend yield. The remaining general specs and filter
specifications should be consistent for the vast majority of stock candidates
and should require few changes.
To describe the import file name, we need to enter the symbol, the
analysis date, and the time. One final step: push the “IMPORT Combo”
button. That’s it. Once you have developed a procedure to generate the input
data, the input process for the Integrated spreadsheet should be fast and
efficient.
 7 – Integrated: Volatility Model & Simulation

This is the second of three chapters that provide step-by-step instructions on

how to use the Integrated spreadsheet. As a reminder, the Integrated
spreadsheet is populated with data for Under Armor (UA) as of July 23,
2014, which is the same stock for which we designed an option earnings
strategy in Chapter 5.
In Chapter 6, we entered all of the required input data and imported the
UA option matrix. In this chapter, we will review the steps required to use
that data to simultaneously solve for all of the parameters of the volatility
model, including the implied earnings volatility (IEV). For those of you who
would like to learn more about the structure of the volatility model, I will also
explain how the volatility model parameters are used to describe the vertical
and horizontal skews.
After solving for and explaining the volatility model parameters, we will
examine the steps required to generate the simulation results, which will be
used to calculate the True Greeks and to optimize earnings strategies.

Volatility Model Estimation

Estimating the parameters of the volatility model is an optimization

problem. As was the case for the Basic spreadsheet in Chapter 3, the
Integrated spreadsheet uses Solver, which is included with Microsoft Excel as
a free Add-in. As I explained in Chapter 3, if you have never used Solver
before, then you will first need to enable the Solver Add-in in your version of
Excel. This exact procedure will depend on your version of Microsoft Excel.
In MS Office 2010, you would use the following button sequence:

File => Options => Add-Ins => Manage Excel Add-ins => Go

This will bring up the Add-in screen we reviewed in Chapter 3 (Figure

3.5), which you can use to enable the Solver Add-in. You may need to
browse to find the Solver Add-in.
The optimization problems in the Integrated spreadsheet are more
complex than the problems we solved using the Basic spreadsheet. As a
result, I created macros that run Solver automatically in the Integrated
spreadsheet. However, you will still need to enable the Solver Add-in and
may also need to add the Solver VBA reference to allow Solver to be run
from Visual Basic (VBA). The macro uses VBA Solver function calls, so
these macros may not be compatible with non-Microsoft, out-of-date, or non-
English language versions of Excel. Please refer to your Excel documentation
for further instructions on managing and enabling Add-ins.
Figure 7.1 is a screenshot of the Solver problem used to estimate the
volatility model parameters. We will use the Solver macro to solve this
problem, so you will not see this screen in practice, but it is still a convenient
way to describe the problem. In this section, I will also refer to Figure 7.2,
which is a partial screen capture from the “SolveIV2” tab. Figure 7.2 includes
a table with the results of the UA volatility model estimation on July 23,
2014. It also shows the macro buttons that are used to estimate the volatility
model. The Solver problem displayed in Figure 7.1 references cells that are
depicted in Figure 7.2.

Solver Volatility Model

Every optimization problem requires an objective or goal, typically

something that we are attempting to minimize or maximize. In the case of the
volatility model (Figure 7.1), we are attempting to minimize the value in Cell
S12 (Figure 7.2), which represents the root mean squared error (RMSE) of
the prices of every option in the matrix that met our filter criteria. The pricing
errors are approximated using the errors in implied volatility and the input
Vega, adjusted for the percentage of normal trading days.
To meet the objective, an optimizer must systematically change the
adjustable or decision variables, which in this case represent the volatility
model parameters that quantify the vertical skew, the horizontal skew, and
the implied earnings volatility (IEV). The volatility model parameter values
are contained in the following Cells: S4, S6:T8, U6:V6, U8:V8. The list of
decision variables is shown in Figure 7.1 and the specific adjustable cells can
be seen in the table in Figure 7.2.
Notice that the “Make Unconstrained Variables Non-Negative” check-
box in Figure 7.1 is unchecked, indicating that the unconstrained volatility
parameters may be negative. Finally, the solving method shown in Figure 7.1
is “GRG Nonlinear” because the volatility functions are non-linear and
Solver uses the Generalized Reduced Gradient (GRG) Algorithm to process
these nonlinearities.
 I will explain the volatility parameters and the associated constraints in
detail later in this chapter. For now, let’s continue to focus on running Solver
and on the UA volatility model solution’s summary statistics.
There are two Solver macro buttons shown in Figure 7.2. The
“Randomize Volatility Parms” macro should be run first, followed by the
“Estimate Volatility Model” macro. As described earlier, the volatility model
contains non-linear functions, which makes it more difficult for an optimizer
to solve. Optimizers often get stuck on a local optimal solution of a non-
linear function, which may prevent the software from finding the global
optimal solution. As a result, the success of a non-linear optimization is
dependent on the initial conditions. That is why I included the “Randomize
Volatility Parms” macro: to provide different starting points for the
optimization procedure.
Unlike a simple linear programming optimization problem, there is no
guarantee that an optimizer will be able to identify the best possible solution
for non-linear problems. Fortunately, the Solver volatility model estimation
runs relatively quickly and you can rerun the optimization with a new set of
initial parameter values if you are not satisfied with your results.
So how can we evaluate the estimation results? The first step is to check
the Solver return value, which will allow us to verify that Solver ran correctly
and did not experience any errors. In the UA estimation, the Solver return
value was zero (X4). A list of Solver return values and their descriptions
from the Frontline (Solver Vendor) website is shown in Figure 7.3. A return
value of zero indicates that “Solver found a solution. All constraints are
satisfied.”
Now we are interested in evaluating how well the volatility model
solution explains the volatility structure of the UA matrix. The Solver
solution resulted in an objective function value of $0.175 (S12), which
represents the root mean squared pricing error for the 348 (R3) options used
in the estimation process. The root mean squared normal volatility error was
12.23% (S10).
You may be familiar with the term “R-Squared” which is often used in
linear regression to explain how well a statistical model fit the underlying
data. R-Squared represents the percentage of variance in the data explained
by the model. As a result, the value of R-Squared must be between zero and
one (100%). A value of one or 100% would indicate that 100% of the
variance in the data is explained by the model. The R-Squared of the UA
model was 0.869 or 86.9% (S3).
 When I initially ran the UA volatility model estimation, I was
disappointed by the magnitude of the $0.175 (Figure 7.2 Cell S12) pricing
error and by the 12.23% (Figure 7.2 Cell S10) normal volatility error. I had
used the estimation model many times before and the magnitude of the UA
errors was higher than usual. I reviewed the individual pricing errors in
Column Q of the “SolveIV” tab (subset shown in Figure 7.5) and determined
that there were systematic pricing anomalies in the UA matrix. Note, the
anomalies were not pricing errors; they represented undervalued and
overvalued options. We exploited these pricing anomalies when we designed
the optimal option earnings strategy for UA in Chapter 5.
If you are not satisfied with the results of the Solver volatility model
estimation and rerunning Solver with different initial parameter values does
not resolve the problem, there are two other options that may improve the
model results. The first is to modify the parameters Solver uses to run the
optimization algorithm (Figure 7.4).
The Solver macro reads the parameter values from the blue-shaded cells
in Figure 7.4. If you modify these values, the Solver macro will use your
parameter values during the volatility model optimization process. A
description of Solver’s solution parameters is beyond the scope of this book.
Please refer to the Solver documentation for more information on the solution
parameters. If you are not a proficient and experienced Solver user, I
recommend using the initial Solver solution parameters provided in the
Integrated spreadsheet.
 If you discover one or two extreme pricing errors that are adversely
affecting the quality of the volatility model estimation, you could exclude
them from the estimation without modifying the filters or reimporting the
matrix.
Column D on the “SolveIV2” tab represents an Exclude flag (see Figure
7.5). Normally the Exclude flag is set to zero for every option, which means
that it is included in the estimation. However, if you would like to exclude
one or more options due to extreme pricing errors, enter the value of one (1)
in the appropriate row of Column D. Before closing and saving the Integrated
spreadsheet, remember to reset the Exclude values to zero. Otherwise, you
would exclude the option in that row in all future optimizations.
 After editing the Solver solution parameters (Figure 7.4) and/or the
Exclude values in Column D (Figure 7.5), rerun the Solver macros to re-
estimate the volatility model.

Volatility Model Parameters

We initially used Figure 7.2 to evaluate the Solver UA volatility model

solution, but we deferred a discussion of the volatility model parameters until
now. To eliminate the need to refer back to Figure 7.2, I created a second
copy of the volatility model solution and parameters in Figure 7.6.
The volatility model simultaneously estimates eleven different volatility
model parameters: the annualized implied earnings volatility (S4), three
horizontal skew coefficients for call options (S6:S8), three horizontal skew
coefficients for put options (T6:T8), two vertical skew slope coefficients (U6,
U8) and two vertical skew curve coefficients (V6, V8).
We have discussed the implied earnings volatility (IEV) extensively
throughout the book, so I will not go into more detail here, except to note that
the IEV is estimated simultaneously with all of the other volatility model
coefficients. By definition, IEV is an estimate of implied earnings volatility,
which is the same for all of the options in the matrix. All of the other
volatility model parameters are used to explain normal volatility, which is
different for every option in the matrix. The aggregate implied volatility
formula is used to combine earnings volatility and normal volatility.
 Before we can estimate the normal volatility model parameters, we must
first define the structure of the volatility model. In other words, we need a set
of formulas that can be used to calculate the normal volatility for every
option in the matrix. Those formulas must rely exclusively on objective
variables that can be calculated or are observable for every option. The
resulting formulas will describe two option pricing phenomenon: the
horizontal skew and the vertical skew.
The normal horizontal and vertical skew sections that follow are optional
material. In other words, you do not need to know the structure of the
horizontal and vertical skew calculations to use the Integrated spreadsheet or
to create earnings strategies.
However, understanding the horizontal and vertical skew is valuable; the
volatility structure has a profound effect on option risk and returns. Gaining
insight into the volatility structure will help you design and evaluate all types
of option strategies, which will make you a better option trader. The formulas
are not overly complicated, but are not required reading.
Unlike the aggregate implied volatility formula presented earlier, which
correctly aggregates independent components of volatility and is consistent
with option pricing theory, the volatility model presented in the following
sections is one of many possible structures that could be used to model option
volatility. There is no one definitive volatility structure or model. I designed
the model with sufficient power to explain the distribution of normal implied
volatilities across time and among strike prices. The model is continuous,
which means that it can be used to solve for the normal implied volatility for
any strike price and time to expiration, now or in the future.

Normal Horizontal Skew

Normal implied volatility varies as a function of time to expiration,

reflecting the market’s implied forecast of future volatility. If the current
level of normal volatility is perceived as too low, then implied volatilities will
increase as a function of time to expiration. If the current level of normal
volatility is viewed as too high, then implied volatilities will decrease as a
function of time to expiration. The normal horizontal skew function will
calculate the at-the-money normal implied volatility (ATM NIV) as a
function of the time remaining until option expiration (in years). The number
of remaining trading days is used to calculate the time remaining until
expiration.
As explained earlier, we will estimate two horizontal skew functions, one
for calls and one for puts. Both functions will use the same formula or
structure, but will have different coefficients. The normal horizontal skew
function has three terms: an intercept, a slope, and a measure of curvature.
Let’s begin with the normal horizontal skew function for at-the-money
UA call options on July 23, 2014. An actual UA call option with a time to
expiration of 0.08730 years (~1-month) will be used in the calculation
example below.

ATM NIVC = At-the-money call option normal implied volatility

EXP represents the Excel function e raised to a power, e =
2.718281828
aC = call option intercept = 26.52% (S8)
bC = call option slope = 299.78% (S6)
cC = call option curvature = 0.67% (S7)
 T = Time to expiration in years = 0.08730 (example value)

ATM NIVC =aC+bC*(1-EXP(-cC*T))

ATM NIVC =26.52%+299.78%*(1-EXP(-0.67%*0.08730))

ATM NIVC =26.52%+299.78%*(1-EXP(-0.000585))
ATM NIVC =26.52%+299.78%*(0.000585)
ATM NIVC =26.52%+0.17% = 26.69%

Let’s continue with the normal horizontal skew function for at-the-money
UA put options on July 23, 2014. An actual UA put option with time to
expiration of 0.08730 years (~1-month) will be used in the calculation
example below.

ATM NIVP = At-the-money put option normal implied volatility

EXP represents the Excel function e raised to a power, e =
2.718281828
aP = put option intercept = 32.34% (T8)
bP = put option slope = -3.81% (T6)
cP = put option curvature = 600.00% (T7)
T = Time to expiration in years = 0.08730 (example value)

ATM NIVP =aP+bP*(1-EXP(-cP*T))

ATM NIVP =32.34%+ -3.814%*(1-EXP(-600.00%*0.08730))

ATM NIVP =32.34%+ -3.814%*(1-EXP(-0.5238))
ATM NIVP =32.34%+ -3.814%*(0.407734)
ATM NIVP =32.34%+ -1.56% = 30.78%

The above volatility model formulas and UA coefficients can be used to

calculate the at-the-money normal implied volatility for any UA call option
or put option, regardless of the amount of time remaining until expiration.
The solid red line in Figure 7.7 depicts the modeled ATM NIV for UA put
options on July 23, 2014. The solid green line shows the estimated ATM NIV
for UA call options on the same date.
 We also need to define formulas and estimate the resulting coefficients of
the vertical skew model to find the normal implied volatility for every option
in the UA matrix. The dotted and dashed lines in Figure 7.7 illustrate the
curve and slope coefficients for the vertical skew model as a function of time
to expiration. We will discuss the vertical skew model in detail in the next
section.
 Normal Vertical Skew

Normal implied volatility varies as a function of the strike price of an

option in relation to the current underlying stock price. This relationship is
called the vertical skew, which is driven by the market’s expected magnitude
of directional price movement. In other words, if the market expects the
prices of the underlying security to fall much faster than they rise, options
with relatively low strike prices will have higher implied volatilities than
options with relatively high strike prices. The implied volatility relationship
will be reversed when the market expects prices to rise much faster than they
fall.
To model the vertical skew, we must first come up with a metric that
quantifies the relative strike price, which we will call the vertical skew
coefficient (VSC). We will then create a formula that expresses the normal
implied volatility of any option in the matrix as a function of its VSC.
I considered several possible VSC functions, but ultimately decided to use
the same function used by OptionVue in their vertical skew modeling. Their
metric fit the normal implied volatility data well, and it is consistent with the
Black-Scholes Option Pricing Model (BSOPM). In fact, elements of the VSC
formula are taken directly from the BSOPM.
The following VSC formula expresses the difference between the strike
price and the underlying stock price, adjusted for the time remaining until
expiration. The following example uses specific data values for an actual UA
put option on July 23, 2014.

VSC = Vertical Skew Coefficient

LN represents the Excel natural log function
K = Strike price ($56.00)
S = Current Underlying Stock Price ($60.63)
T = Time to expiration in years = 0.08730

VSC = LN(K/S)/(T^0.5)

VSC = LN(56.00/60.63)/(0.08730^0.5)
VSC = LN(0.923635)/(0.295466)
VSC = -0.079438/(0.295466)
 VSC = -0.269

Now that we have the calculated the value of the vertical slope coefficient
(VSC), we can use the formula below to calculate the normal implied
volatility for the one-month (0.08730 years) UA put option with the strike
price of $56. In fact, after we calculate the VSC, we have all of the
information necessary to calculate the normal implied volatility for any
option in the UA matrix. The following example is a continuation of the one-
month (0.08730 years) UA put option for which we calculated the ATM NIV
earlier in the horizontal skew section.

NIV = Normal Implied Volatility

VSC = Vertical Skew Coefficient
ATM NIV = At-the-money normal implied volatility (calculated
earlier: 30.78%)
T = Time to expiration in years = 0.08730
bVS = Vertical Skew Slope
cVS = Vertical Skew Curvature

NIV = ATM NIV + (bVS*(VSC)+cVS*(VSC^2))

Unfortunately, we cannot solve for the NIV until we calculate the vertical
skew slope and vertical skew curvature values. Just as the ATM NIV varies
as a function of the amount of time remaining until expiration (horizontal
skew), the vertical skew is not constant across the term structure of
volatilities either. In other words, the shape of the vertical skew curve is
different for options with different expiration dates. As a result, the values of
the vertical skew slope (bVS) and the vertical skew curvature (cVS) are both
unique to options with 0.08730 years remaining until expiration.
However, it would not be practical or desirable to estimate independent
vertical skew slope and curvature values independently for every option
expiration date. We are ultimately interested in calculating the normal
implied volatility for any option in the matrix on the initial pricing date and
at our choice of several projected dates in the future. As a result, we need to
model the vertical skew slope and curvature as a continuous function of the
time remaining until expiration.
 The vertical skew slope and curvature functions were depicted earlier in
Figure 7.7 estimation results. The dotted and dashed lines in Figure 7.7
illustrate the curve and slope coefficients for the UA vertical skew model
respectively. Both the vertical skew slope and vertical skew curve functions
are linear, which means that both functions can be explained by slope and
intercept coefficients, which are estimated simultaneously with all of the
other volatility model parameters.
Below are the formulas used to calculate the vertical skew slope and
vertical skew curve coefficients, both of which are expressed as a function of
time remaining until expiration. The formulas below are a continuation of our
earlier example, which used the one-month UA put option with a $56 strike
price.

bVS = Vertical Skew Slope

A = Intercept (Figure 7.6 Cell U8) = 2.556%
B = Slope (U6) = -5.742%
T = Time to expiration in years = 0.08730

bVS = A + B*(T)
bVS = 2.556% + -5.742%*(0.08730)
bVS = 2.556% + -0.501%
bVS = 2.055%

cVS = Vertical Skew Curvature

A = Intercept (V8) = 21.192%
B = Slope (V6) = -10.136%
T = Time to expiration in years = 0.08730

bVS = A + B*(T)
bVS = 21.192% + -10.136%*(0.08730)
bVS = 21.192% + -0.501%
bVS = 20.307%

To refresh your memory, here is the normal implied volatility calculation

that we began earlier for the one-month UA put option with the $56 strike
price. The values of the vertical skew slope (2.055%) and curvature
(20.307%) for the one-month UA put option were derived above.

NIV = Normal Implied Volatility

VSC = Vertical Skew Coefficient
ATM NIV = At-the-money normal implied volatility (30.78%)
bVS = Vertical Skew Slope (2.055%)
cVS = Vertical Skew Curvature (20.307%)

NIV = ATM NIV + (bVS*(VSC)+cVS*(VSC^2))

NIV = 30.78% + (2.055%*(-0.269)+20.307%*(-0.269^2))
NIV = 30.78% + (-0.55%+20.307%*(0.07236))
NIV = 30.78% + (-0.55%+1.7%)
NIV = 30.78% + (0.92%)
NIV = 31.70%

Volatility Model Estimation Summary

Before we continue, let’s take a step back and review the preceding
volatility model calculations intuitively. The volatility model simultaneously
estimates eleven different volatility model parameters: the annualized implied
earnings volatility (S4), three horizontal skew coefficients for call options
(S6:S8), three horizontal skew coefficients for put options (T6:T8), two
vertical skew slope coefficients (U6, U8) and two vertical skew curve
coefficients (V6, V8).
After estimating the volatility parameters, the next step is to use the three
horizontal skew coefficients to solve for the at-the-money normal implied
volatility (IV). Unfortunately, normal implied volatility also varies as a
function of the strike price in relation to the current price of the underlying
security. As a result, we must calculate the vertical skew coefficient (VSC) to
quantify the strike price relationship.
The vertical skew also varies as a function of the time remaining until
expiration, so we also need to calculate the vertical skew slope and curvature
coefficients for each option. Finally, we can use all of these values to
calculate the vertical skew, which can be combined with the ATM NIV to
determine the normal implied volatility for each option.
Figure 7.8 shows the normal implied volatility (dependent y-axis) as a
function of the vertical skew coefficient (independent x-axis) for expiration
dates ranging from one week to two years (RHS legend).
 Aggregate Implied Volatility Revisited

Now that we have estimated the normal implied volatility, how do we

find the aggregate implied volatility? We use the aggregate implied volatility
formula from Chapter 2 to combine the normal implied volatility and the
implied earnings volatility into a single implied volatility value. Here is the
aggregate implied volatility calculation for the one-month UA put option that
we have used throughout this chapter.

NTDE = Number of Earnings Trading Days

NTDN = Number of Normal Trading Days
WE = Earnings Volatility Weight
WN = Normal Volatility Weight

WE = NTDE/( NTDE + NTDN)

WE = 1/( 1 + 21) = 4.5455%

WN = NTDN/( NTDE + NTDN)

WN = 21/( 1 + 21) = 95.4545%

IV = Aggregate Implied Volatility (Annualized)

VE = Earnings Volatility (Annualized) = 137.20%
VN = Normal Volatility (Annualized) = 31.70%

IV = ((WE*(VE^2))+(WN)*(VN^2))^0.5
IV = ((4.5455%*(137.20%^2))+(95.4545%)*(31.70%^2))^0.5
IV = ((8.5564%)+(9.5921%))^0.5
IV = (18.1485%)^0.5
IV = 42.60%

The volatility model estimated that the annualized implied volatility of the
one-month (8/23/2014) UA $56 put option was 42.60% at the close on July
24, 2014. The actual annualized implied volatility of the 8/23/2014 UA $56
put option was 42.23%.
 Solver Volatility Model Constraints

The volatility model Solver solution was shown in Figure 7.2 and again in
Figure 7.6. We initially focused on the Solver solution statistics and then on
the volatility model parameters. In both cases, we ignored the input Cells
with the blue shaded background (U4:W4) which we use to constrain the
Solver solution. Now that we have reviewed all of the volatility model
parameters, we can revisit the Solver constraint cells.
I specified a minimum implied earnings volatility (IEV) of 1.10% (U4),
which is only applied to decision variable cell S4. I also set a minimum
intercept value of 1.10% (V4), which applies to all four intercept decision
variables (S8:V8). Finally, I set a minimum and maximum range of negative
600% to positive 600% (W4) for all of the decision variable cells (S6:V6),
which excludes the implied earnings volatility (S4). These values should
work well for many Solver problems, but you may modify these values if you
would like more control over the Solver solution. If you find the shape or
curvature of the functions too extreme, I suggest reducing the maximum
parameter range from plus or minus 600% to plus or minus 300%, or even
lower.

True Greeks & Simulation

Now that we have estimated the volatility model parameters for UA on

July 24, 2014, we can use these values to calculate the True Greeks. In
Chapter 5, we specified a number of assumptions that were used to simulate
price changes for all of the options in the UA matrix. These assumptions
were initially shown in Figure 5.10 and are depicted again in Figure 7.9 for
easier reference. We will not review the user assumptions again, but these
assumptions are used to calculate all of the scenario assumptions, including
the discrete option price values used in the True Greek sensitivity
calculations.
 You will remember from Chapter 4 that the True Greeks represent risk
measurements or sensitivities, which are derived from discrete changes in
price, time, normal volatility, earnings volatility, etc. As a result, to calculate
the True Greeks, we need to re-calculate how the value of each option would
change in response to specific changes in each of the independent variables,
including any resulting effects on the other dependent variables.
This is nothing more than a succession of price simulations, but one that
requires the parameters of the comprehensive volatility model described in
this chapter to accurately model normal volatility, earnings volatility, and
aggregate volatility. The ability to calculate scenario-specific volatility values
for each option allows us to calculate the True Greeks. All of the True Greek
calculations were described in detail in Chapter 4.
The True Greek intermediate and final calculations are all stored on the
“Greeks” tab of the Integrated spreadsheet, but you may run the True Greek
macro from any of the following tabs: “Greeks,” “TZero,” “CHGTU,” and
“CHGTX.” To run the True Greeks macro from any of these tabs, click on
the button titled “T-Greeks.”
Figure 7.10 is a screen shot from the “Greeks” tab that shows the location
of the “T-Greeks” macro button and the “SIM ALL” macro button. The “SIM
ALL” button runs the price simulations for the current pricing date (“SIM
0”), the first user specified pricing date used in the optimization (“SIM U”)
and the second user pricing date which typically corresponds to the expiration
date of the shortest dated option (“SIM X”).

The “TZero” tab represents the current pricing date (zero time elapsed).
On the “TZero” tab, you can use the following button macros: “T-Greeks,”
“SIM ALL,” and “SIM 0” (Figure 7.11). The values for the instantaneous
price simulation (July 23, 2014) are stored on the “TZero” tab.

The “CHGTU” tab represents the pricing simulation results for the first
user simulation date (July 24, 2014) which are used for optimization
purposes. On the “CHGTU” tab, you can use the following button macros:
“T-Greeks,” “SIM ALL,” and “SIM U” (Figure 7.12). The values for the
price simulation on the first user date are stored on the “CHGTU” tab.
 The “CHGTX” tab represents the pricing simulation results for the second
user simulation date (July 25, 2014), which was intended to represent the
expiration date of the nearest-term option considered in the optimization
process. On the “CHGTX” tab, you can use the following button macros: “T-
Greeks,” “SIM ALL,” and “SIM X” (Figure 7.12). The values for the price
simulation on the second user date are stored on the “CHGTX” tab.

The simulation macros apply the user assumptions (Figure 7.9) to

generate a series of discrete terminal prices (and the corresponding scenario
probabilities) for the underlying stock. For each terminal price, the volatility
model parameters are used to calculate the resulting normal and aggregate
implied volatilities for every option in the matrix. All of these calculations
include the effects of the vertical and horizontal skews, which are different
for every option, but are all derived from the same volatility model
framework and volatility model parameters. The aggregate implied volatility
values are then used to calculate the terminal values of each option on the
desired date.
The simulated option values and corresponding probabilities provide the
necessary data to help us identify the option earnings strategy with the
highest expected profit per unit of risk. The simulated option values are also
used to evaluate the most promising strategies graphically. We will explore
optimization and graphical analysis in Chapter 8.

Summary
 There is a lot going on behind the scenes in the Ingegrated spreadsheet,
especially with respect to the volatility model and simulation calculations.
Fortunately, all of these functions are automated and are very easy to use.
First push the “Randomize Volatility Parms” and “Estimate Volatility
Model” macro buttons on the “SolveIV2” tab to estimate the volatility model.
If your data is accurate and you use the appropriate filter settings, it is
unlikely that you will need to customize the Solver solution parameters or
constraints on a regular basis.
After reviewing the volatility model results, push the “T-Greeks” and
“SIM ALL” buttons on any of the following tabs: “Greeks,” “TZero,”
“CHGTU,” or “CHGTX.” This will generate and save all of the Greek and
simulation results for every option in the matrix.
Except in special cases, that’s all you will need to do: push four buttons in
the proper sequence. That will prepare the spreadsheet for the final steps:
optimization and graphical analysis.
 8 – Integrated: Optimization & Analysis

This is the third of three chapters that provide step-by-step instructions on

how to use the Integrated spreadsheet. The Integrated spreadsheet is
populated with data for Under Armor (UA) as of July 23, 2014.
In Chapter 6, we entered all of the required input data and imported the
UA option matrix. In Chapter 7, we solved for the optimal parameters of the
volatility model, including the implied earnings volatility (IEV). In this
chapter, we will review the steps required to solve for the optimal option
earnings strategy and analyze strategy candidates graphically. We will use the
“Opt” tab for strategy optimization and the “Analysis” tab for graphical
analysis.
Identifying the option strategy with the highest expected profit per unit of
risk is an optimization problem. The optimization problem in this chapter
uses Solver. As I explained in previous chapters, if you have never used
Solver before, then you will first need to enable the Solver Add-in in your
version of Excel and you may also need to add the Solver VBA reference.

Introduction to Optimization

A comprehensive explanation of optimization would require one or more

college or graduate level courses, which is well beyond the scope of this
book. However, a brief introduction to optimization will help you understand
how and why the Integrated spreadsheet was constructed; it will also explain
how the strategy optimization problem was designed to operate within the
strict size constraints imposed by the free version of Solver. This introduction
will help you get the most out of the strategy optimizer in the “Opt” tab of the
Integrated spreadsheet.
We will start with basic concepts and then examine how additional real-
world complexity affects optimization problems. Optimizers attempt to either
maximize or minimize an objective function. Finding the optimal option
earnings strategy means maximizing the expected profit per unit of risk.
Optimizers attempt to identify the best possible combination of values for a
group of decision variables, subject to a set of constraints.
The simplest type of optimization problem is called a linear program or
LP. In a linear program, the objective is a linear function of the decision
variables, which results in a very simple (flat) geometric landscape. Given
this simplicity, if any linear program with continuous decision variables has
an optimal solution, the Simplex algorithm will find that solution – and will
do so relatively quickly. That’s right; a single algorithm published in 1947
called the Simplex algorithm can find the optimal solution to any linear
program, provided that the decision variables are continuous.
What if one or more of the decision variables are not continuous and can
only take on integer values? Then all bets are off. This seemingly benign
integer constraint makes otherwise simple linear problems exceedingly
onerous to solve. This type of optimization problem is called an integer linear
program. A large linear program might take only a few seconds to solve on a
desktop computer. The identical linear program with integer decision
variables (integer linear program) might take a few days to solve!
Unfortunately, the decision variables for many real-world problems can only
accept integer values, and the strategy optimization problem in the Integrated
spreadsheet is no exception. Since we cannot buy or sell a fractional number
of option contracts, our decision variables must be integers.
We now need to add one final layer of complexity. What if the objective
function is not linear? In other words, what if the objective that we are trying
to maximize or minimize is a non-linear function of the decision variables?
As you might expect, this is not going to make our life any easier. Non-linear
objective functions are analogous to an n-dimensional geographic landscape,
with rolling hills, peaks, mountains, cliffs, and valleys.
If the geometric landscape is smooth and convex over the entire range of
solutions, then gradient-based techniques will be able to find the global
maximum or minimum, which is what we want. How does a gradient-based
technique work? At any given point, it evaluates the slope, derivative, or
gradient in each direction. To find the global maximum, it takes several small
steps in the direction of the steepest gradient, then stops and recalculates a
new set of gradients. It repeats the process until it reaches the global
maximum. This algorithm is only guaranteed to find the optimal global
solution if the entire solution region is smooth and convex. If not, then
gradient-based solution algorithms may get stuck at a local optimum, when
our goal is to find the best possible solution over the entire geometric
landscape, which is called the global optimum.
Unfortunately, our geometric solution landscape is not uniformly smooth
and convex, so we need a different approach, one that has a better chance of
exploring the entire geometric solution landscape and finding the global
optimum. That solution technique is called a genetic or evolutionary
algorithm.
Genetic algorithms begin with a large population of solutions and
repeatedly “breed” parent solutions, taking gene sequences (decision variable
values) from both parents, to create child solutions. This is called crossover.
Evolution is cruel but effective, only allowing the “fittest” child solutions to
survive and populate subsequent generations. Evolution is powerful, but it
does take time, and so do genetic algorithms. Fortunately, Moore’s law and
the consistent increase in processing speeds have made genetic algorithms an
invaluable tool in non-linear optimization.
However, there is an obvious problem with genetic algorithms. What if
the characteristics of the initial population are inferior and subsequent
generations can only inherit the mediocre genes of the initial parent
population? Sure, the solution will improve somewhat over time, but it will
always be limited by the substandard genes of the initial population. In
optimization terms, this means that the global optimal solution will not be
found in the geometric landscape represented by the initial parent population.
For maximum effectiveness, geometric algorithms must search the entire
solution space. How do geometric algorithms solve this inbreeding problem?
Mutation. A user-specified percentage of the parent genes spontaneously and
randomly mutate, bringing new genes to the child generations that were not
present in either parent’s genes. Many gene mutations generate poor solutions
and simply die out in the first generation.
However, others open up promising new regions of the geometric
landscape that lead to superior objective function values. Once the promising
new mutated gene improves the quality of the offspring, it will survive and
flourish and the new region of the solution space will be explored in
subsequent generations.
Increasing the percentage of mutations helps the geometric algorithm
explore the entire geometric landscape, but also makes it more difficult for
the algorithm to pass attractive genes on to future generations. Changing the
mutation percentage involves tradeoffs between landscape coverage and
efficiency of convergence.
Another problem with geometric algorithms is that they abhor constraints
and perform much better when the only constraints are simple boundary
conditions. When feasible parent solutions repeatedly produce infeasible
child solutions due to multiple constraints, it would be very challenging for
geometric algorithms to progress toward the optimal solution.
There is one final problem with genetic algorithms: there is no way to
know when they have found the optimal solution. Genetic algorithms are
really nothing more than a sophisticated version of trial and error and they
never give up. They will continue to breed and evaluate new offspring until
the power goes out. We literally need to tell them when to stop.
As you probably already guessed, strategy optimization is a non-linear,
discontinuous, integer programming problem with hundreds or even
thousands of prospective option candidates. As a result, it is the most difficult
type of optimization problem to solve. Using the free version of Solver will
make it even more challenging.

Free Solver’s Limitations

When I wrote this book, I wanted to make the material and spreadsheet
tools as accessible and practicable as possible, so forcing the reader to
purchase expensive optimization software just to experiment with strategy
optimization seemed unreasonable. As a result, I designed the optimization
problem to use the free version of Solver that is included with recent versions
of Microsoft Excel.
Unfortunately, the free version of Solver is artificially restricted to work
with a limited number of decision variables and constraints. In addition, the
solution algorithms in the free version of Solver are purposely inefficient and
relatively slow. These deliberate deficiencies are intended to induce Solver
users to upgrade to one of the premium versions of Solver. While I am an
ardent proponent of capitalism and this is an effective marketing technique, it
necessitated some compromises and creativity when designing the strategy
optimization framework.
Normally, when designing a strategy optimization, I would assign one (or
more) decision variables to each investment candidate. In the simplest case,
each candidate would have its own decision variable, which would represent
the number of contracts purchased or sold for that option. This is a very
efficient and intuitive approach, but it requires a large number of decision
variables.
Unfortunately, even after using filters to reduce the number of option
candidates from the matrix, our UA example had almost 350 candidates,
which would have far exceeded free Solver’s 200 decision variable limit. As
a result, my usual approach would not work with the free version of Solver.

Optimization Design Considerations

Instead, I had to make some compromises. First, I decided to limit

solutions to combinations of three spreads: a call spread, a put spread, and a
straddle or strangle. This may seem like a significant concession, but
combinations of these three spreads can be used to create almost every
conventional option strategy: butterflies, broken wing butterflies, condors,
calendars, credit spreads, debit spreads, diagonals, straddles, strangles, plus a
surprising variety of unusual hybrid combinations.
The call and put spreads both require an offsetting purchase and sale,
which is obviously the definition of a spread. This structure helps limit
potential losses, which is especially important when trading options over
earnings events.
The strategy optimization only permits straddle/strangle purchases.
Holding short positions in straddles or strangles when earnings are
announced could lead to unlimited losses. Given that earnings are typically
announced when the market is closed, there is no way to hedge or manage
these losses. As a result, the strategy optimizer only permits long straddles or
strangles. However, you may still short volatility by using the put and call
spreads to construct butterflies, condors, or similar covered hybrid
derivatives.
Given the above structure, there are only six potential option positions:
two for the call spread, two for the put spread, and two for the straddle or
strangle. The row number in the “Opt” tab of the Integrated spreadsheet will
be used to describe the specific options purchased or sold for each spread.
This will only require a total of six integer decision variables, each of which
will contain the row number of the option purchased or sold to construct the
spread.
The values of these six decision variables will describe the positions used
to construct each spread, but we still need to know how many contracts to
buy or sell. For that we will need three additional integer decision variables,
one for each spread. We only need three additional decision variables (not
six), because the spreads all require the same number of contracts on both
sides of the trade. The resulting integer problem solution only requires nine
decision variables, instead of potentially hundreds or even thousands. As a
result, this problem structure does not violate free Solver’s decision variable
limits.
That is the good news. The bad news is that this approach is not nearly as
efficient as assigning one or more decision variables to each option
candidate. As a result, the Integrated spreadsheet employs a specific sorting
algorithm in the “IMPORT Combo” macro to help mitigate some of the
inefficiencies resulting from this approach.

Several Solution Methodologies

The success of a genetic algorithm depends heavily on searching the

optimal region of the solution landscape, which in turn is governed by the
initial starting point/solution. As a result, the initial values of the decision
variables used to seed the genetic algorithm are critical to the success of the
algorithm.
Many different traders will use the Integrated spreadsheet, all with unique
backgrounds, goals, and time constraints. As a result, the spreadsheet was
designed to accommodate several different methodologies for searching the
solution landscape for the optimal solution. For traders who have a specific
strategy solution in mind, the Integrated spreadsheet includes macros that
allow them to use their solution as the starting point in a single Solver
optimization run.
Keep in mind that starting-point solutions must be expressed as a
combination of the spreads described earlier: call spread, put spread, and/or
straddle/strangle. One, two, or all three spreads can be used and the user will
be able to specify the time to expiration and Deltas for each option in the
initial solution.
Unfortunately, strategy optimization is complex and most traders will not
have a specific strategy in mind. Given the importance of searching the entire
solution space, I created ten pre-set strategies for those traders to use as
starting points in the optimization: bull call spread, bear put spread, neutral
condor, neutral butterfly, bullish butterfly, bearish butterfly, neutral strangle,
neutral straddle, bullish strangle, and bearish strangle. The pre-set strategy
names and descriptions are not important. They simply represent ten
solutions centered in different regions of the solution landscape with a range
of risk characteristics or Greeks.
 For users who would like to use the pre-set strategies, but would prefer to
identify an attractive strategy as quickly as possible, the Integrated
spreadsheet includes a macro that will calculate the values of the objective
function for each of the pre-set strategies and a second macro that will use the
most promising macro as the starting point for a single Solver optimization
run.
If you have more time and would like to maximize exploration of the
solution set, the Integrated spreadsheet includes a macro that will use the pre-
set solutions as starting points for ten different Solver optimization runs. In an
attempt to improve on the findings of the genetic algorithm, the same macro
will then run a second series of Solver optimization runs using the
Generalized Reduced Gradient (GRG) nonlinear algorithm, with the initial
ten optimal Solver solutions as starting points. The result is twenty Solver
runs, from twenty different starting points, using two different solution
algorithms. This sounds complicated (and it is), but it can be accomplished
by pushing a single macro button and taking a break for a cup of coffee (or
two).
It is interesting that the final solutions will often deviate significantly
from the starting solutions. In fact, the genetic algorithm is likely to produce
different solutions every time you run the optimizer. That probably seems
strange if you are used to solving linear programs with the Simplex
algorithm, which will always find the single best solution, if one exists.
The genetic and GRG nonlinear algorithms are not guaranteed to locate
the single global optimum of a non-linear, discontinuous, integer
optimization problem. Fortunately, we do not need the “best” solution to
generate a trading edge; we only need to identify solutions with an attractive
expected profit per unit of risk.

Objective Function

So what do I mean by expected profit per unit of risk? The expected profit
is easy to understand. Expected profit equals the probability-weighted profit
across 100 discrete pricing scenarios. The prices and probabilities were
derived from a binomial lattice that was generated from the expected normal
volatility, the expected earnings volatility, and the holding period used in the
optimization.
To calculate the expected profit per unit of risk, we also need to specify a
risk measure. I included three different risk measures, one of which must be
selected from a drop-down list in Cell A1 of the “Opt” tab of the Integrated
spreadsheet (not shown). The first risk measure is sigma, which represents
the standard deviation of the expected profit. This measure is used by many
traders, which is why I included the risk measure in the spreadsheet.
However, I did not use this risk measure in the examples from this book. The
reason is that standard deviation or variation in expected profits is only a
problem on the loss side. Scenarios with large gains also increase sigma,
which would lower the objective function, making those solutions look
worse.
As a result, I used the second risk measure (maximum loss) to calculate
the objective function for all of the examples in the book. In other words, the
resulting objective function would equal the expected or probability-weighted
profit divided by the absolute value of the maximum loss. We use the
absolute value because the maximum loss is negative and we want to
maximize the objective function.
Finally, I included a third risk measure with a value of one (1.0). Using a
constant value of 1.0 probably seems strange as well, but if we divide the
expected profit by 1.0, the result equals the expected profit. In other words,
by including a risk measure of 1.0, this allows users to maximize the
expected profit if they desire, with no consideration of risk.

User Input Settings

Before running Solver, the Integrated spreadsheet must translate the user
or pre-set strategy descriptions into row numbers, which are used to describe
the three spreads. The row numbers are also used to calculate the objective
function and various strategy summary statistics. The user and pre-set
strategy descriptions both use the section of the “Opt” tab illustrated in
Figure 8.1 to translate trading days until expiration and Delta into the row
numbers for the corresponding options from the matrix that most closely
approximate the desired values.
For now, let’s focus on how we use this section of the spreadsheet to
enter the user-defined strategy that will serve as the starting point for Solver.
You will recall that cells with a blue-shaded background and white text
represent input cells. The most important of the user input cells indicate
whether the three spreads (Call, Put, Straddle/Strangle) will be included in
the solution (Q18:S18). A value of one indicates the spread will be included
and a value of zero directs the spreadsheet to exclude the spread. The only
acceptable values for these three cells are one and zero.
Next, we will need to enter the desired number of trading days until
expiration for each of the six options used in the three spreads (K16:P16).
The desired number of trading days for the call sale and purchase are entered
in Cells K16 and L16. The desired number of trading days for the put sale
and purchase are entered in Cells M16 and N16. Finally, we enter the trading
days for the call and put components of the straddle/strangle in Cells O16 and
P16. In this example, 22 days is approximately equivalent to one month.
Even if a spread is excluded from the initial strategy, as is the case for
both the call and put spread in Figure 8.1 (Q18:R18), we still need to enter a
plausible number of trading days that is consistent with the range of option
candidates used in the optimization.
It is also important to note that the number of trading days entered in
Cells K16:P16 will be used for all strategies. This includes user-defined
strategies as well as the ten pre-set strategies. It would have been too
cumbersome to specify a different set of values for every strategy.
In addition to trading days until expiration, we also need to enter the
desired Delta for each of the six options used to construct the three spreads
(K18:P18). Obviously the calls will have positive Deltas and the Puts will
have negative Deltas. The Deltas are expressed per contract, not per share. In
the case of UA, the call Deltas will range from zero to 100 and the put Deltas
will range from zero to negative 100.
 The spreadsheet uses Excel’s Match function to find option candidates
that closely match the desired number of trading days until expiration and the
desired Delta. The row numbers for the resulting options are stored
temporarily in the solution Cells (K20:P20).
In addition, the Integrated spreadsheet also calculates the number of
contracts for each spread (Q20:S20) to be consistent with the user-specified
amount of required capital. In all of the strategy examples in the book, I
specified a required capital range of $9,000 to $11,000 (not shown), to give
the optimizer some flexibility in finding an integer solution. This flexibility is
very important.
The spreadsheet calculations of required capital look at each spread
individually and are relatively simplistic. In reality, every broker uses their
own margin calculation algorithms, which are based on FINRA margin
requirements and on their proprietary portfolio margin risk models.
Let’s look at the options selected by the spreadsheet for the initial
solution. The call and put spreads are excluded, so we will focus on the
straddle. The nearest call option candidate was found in row 106 (O20) and
the closest put option was located in row 281 (P20). For both the call and put
purchase, we specified 22 trading days until expiration (O16:P16). The call
and put options found by the spreadsheet for the initial strategy exactly
matched our desired number of trading days (O17:P17).
We specified Deltas of 60 (O18) for the call option and negative 40 (P18)
for the put option. The call and put options found by the spreadsheet had
Deltas of 60.7 (O19) and negative 39.4 (P19) respectively. In addition, the
spreadsheet reports the option type (K15:P15), option price (K21:P21), and
strike price (K22:P22) for each of the options identified by the spreadsheet.
Once we are satisfied with the initial solution, we can use the user macros
to evaluate or optimize the initial strategy. To calculate the objective function
value (and other summary statistics) for the initial strategy, use the “Copy
User” button macro shown at the right-side of Figure 8.1. The Copy User
macro will copy the initial solution from Cells K20:S20 to the calculation
region of the spreadsheet shown in Figure 8.2 (C3:K3).
Figure 8.2 shows the initial user-strategy before optimization. The value
of the objective function (before the strategy is optimized) equals +0.25 (L3),
which is not bad, especially for a (non-optimized) starting point. The
objective function value of +0.25 represents the expected profit after
transaction costs, divided by the absolute value of the maximum loss.
If you are curious, the $1,212 in Cell L2 represents the expected profit
before transaction costs and the value of 0.17 in Cell L4 represents an
alternative objective function value, which in this case equals the expected
profit divided by sigma. The optimizer will always attempt to maximize the
value shown in Cell L3. The other values are shown for informational
purposes only.
The “MaxLoss” shown in Cell A1 indicates that the objective function
(L3) equals the expected profit divided by the absolute value of the maximum
loss. The solution (C3:K3) is also used to calculate a number of other
summary statistics, which we will review later in this chapter. For now, we
will continue to focus on optimization procedures.
 If you are satisfied with the initial objective function value (L3), use the
“Opt User” push-button macro (Figure 8.1) to optimize the user-defined
strategy. The optimized strategy results are shown in Figure 8.3. The
objective function value increased from 0.25 (Figure 8.2 cell L3) to 0.63
(Figure 8.3 cell L3), which is a very good solution. You will also notice that
the expected profit excluding transaction costs (L2) and the expected profit
divided by sigma also increased.
If we look a little closer, there is another very interesting observation: the
optimized solution includes a 14-contract put spread (J3), which was not even
part of the initial solution. The initial solution only used a straddle.

This was the best solution found by Solver based on the initial user
solution. If we ran Solver again, it might even improve upon this value.
When we use the ten pre-set solutions later in this chapter, Solver will be able
to find a strategy solution with a much larger expected profit per unit of risk.
As a result, we will delay our discussion of the optimal solution until then.
Before we explore the pre-set optimization macros, let’s take another look
at the Solver settings (Figure 8.4). The Solver settings used by the strategy
optimizer are located in the Cells V10:V20 on the “Opt” tab. As was the case
when using Solver to identify the optimal volatility model parameters, you
should review the Solver documentation before modifying these values.
However, modifying these settings might improve Solver’s performance.
Before continuing, I want to explain a few of the Solver settings in Figure
8.4. The “MaxTime” value of 300 (V10) indicates that Solver will run for a
maximum of 300 seconds (five minutes) per optimization run. Obviously
increasing this value will give Solver more time to explore the solution space,
but keep in mind that the Solver macro uses both the genetic algorithm and
the GRG algorithm.
The “MaxTimeNoImp” value of 60 (V20) instructs Solver to stop an
optimization if the objective function value has not improved in the past 60
seconds. This setting will prevent Solver from wasting additional time when
its search algorithm is ineffective.
The “PopulationSize” of 1000 (V16) is a relatively large initial
population, which will diversify the “gene pool” and help Solver explore the
entire solution space. However, increasing the population will also increase
the run-time.
I mentioned the mutation rate before. The initial “MutationRate” of 0.25
(V19) will allow Solver to consider solutions that were not represented in the
initial population, but should not compromise Solver’s ability to converge on
a solution. Varying this setting could improve or degrade Solver’s
performance.
Finally, the “RandomSeed” of zero (V17) instructs Solver to generate a
different population every time it runs the analysis. This allows you to re-run
Solver in an attempt to improve upon your initial solutions. Solver only
retains improved solutions, so re-running Solver cannot reduce your objective
function.
 Pre-Set Strategy Optimization

As explained earlier, the “Opt” tab includes ten pre-set strategies to use as
starting points in the optimization: bull call spread, bear put spread, neutral
condor, neutral butterfly, bullish butterfly, bearish butterfly, neutral strangle,
neutral straddle, bullish strangle, and bearish strangle. The strategies shown
in Figure 8.5 represent the ten pre-set solutions, which are centered in
different regions of the solution landscape.
The strategy descriptions use the same framework we used earlier to enter
the user strategy. The number of trading days entered in Cells K16:P16
(Figure 8.1) are used for all strategies, so we do not need to enter these
values again. However, the pre-set strategy descriptions do include the
desired Delta values (Figure 8.5 Cells AA25:AF34) as well as the
include/exclude values (Figure 8.5 Cells AG25:AI34) for each spread. The
values in the last row of Figure 8.5 (AA35:AI35) are automatically populated
with the user settings from Figure 8.1 (K18:S18). The user strategy is not
used by the pre-set strategy optimization macros.
 The easiest, most exhaustive (and most time consuming) way to search
for the optimal strategy is to use the “Opt All” push-button macro (Figure
8.6).

The “Opt All” macro copies the Delta and include/exclude spread values
for the first pre-set strategy (AA25:AI25) to the Delta row of the user matrix
described earlier (Figure 8.1 K18:S18). The user matrix finds the appropriate
row numbers for each option (Figure 8.1 K20:S20) and the macro then copies
these values to the calculation region of the “Opt” tab (C3:K3). The macro
then uses these values as the starting point to run the Solver optimization.
The results for the first pre-set strategy are then copied to the first row of the
strategy results section of the “Opt” tab (Figures 8.7 & 8.8, Cells K25:Y25).
The “Opt All” macro then repeats the same procedure for the remaining
nine pre-set strategies. The optimized results for all of the pre-set UA
strategies are shown in Cells K25:Y34 of Figures 8.7 & 8.8. The results were
split into Figures 8.7 & 8.8 due to the width of the region. This macro
requires twenty different optimization runs, which takes a lot of time.
If you are not satisfied with the results, you could use the “Re-Opt All”
push button macro (Figure 8.6) to re-optimize all of the strategies. However,
instead of starting from the pre-set strategy definitions, Solver would start
with the results from the previous optimization runs (K25:S25). The “Re-Opt
All” macro would then copy the results from the re-optimization to Cells
K25:Y34, replacing the original optimization results. This macro also
requires twenty different optimization runs and will probably take nearly as
much time as the “Opt All” macro.
The “Calc All” push-button macro in Figure 8.6 follows the exact same
procedure as the “Opt All” macro, except it does not optimize the initial
strategies. Instead, it calculates the objective function and summary statistics
for each pre-set strategy (before optimization) and copies the non-optimized
results to Cells K25:Y34. Why would we want to do this? Because the “Calc
All” macro is very fast and we might not have enough time to run 20
different optimization runs.
Instead, we could run the “Calc All” macro and then use the “Opt (Nth)”
push-button macro in Figure 8.7 to optimize the single pre-set strategy with
the highest (pre-optimization) objective function value. Simply enter the
desired solution number (1-11) in input Cell J24 and push the “Opt (Nth)”
button. The optimized results will replace the initial solution values in the
appropriate row in Figure 8.7 & 8.8.
In addition to the “Opt (Nth)” macro, there are three additional macros
that only use one user-specified pre-set solution at a time. The “Re-Opt
(Nth)” push-button macro uses the original optimization results for the pre-
set strategy of our choice to re-optimize the strategy. It then copies the results
to the corresponding row of Figures 8.7 & 8.8.
The “Calc (Nth)” macro calculates the (pre-optimization) objective
function and summary statistics for the pre-set strategy of our choice and
copies the results to the appropriate row of the results section of the “Opt”
tab.
Finally, the “Copy (Nth)” push-button macro copies the solution of our
choice back to the calculation region of the “Opt” tab (C3:K3), which
recalculates the objective function and all of the summary statistics. The
solution in the calculation region of the “Opt” tab is also used to generate the
values for the graphical analysis, which will be discussed later in this chapter.
The row numbers for each option and the number of contracts for each
spread are shown in Figure 8.7, as are the four macros that run one of the pre-
set strategies at a time.
 Figure 8.8 is a continuation of Figure 8.7. Each row in Figure 8.8
corresponds to the same strategy from the same row in Figure 8.7. Figure 8.8
shows the objective function (T), average profit after transaction costs (U),
standard deviation of the profit (V), maximum loss (W), the required capital
(X), and the Solver result (Y).
As you can see in Figure 8.8, the ten pre-set strategies produced a wide
range of optimal solutions, which illustrates the advantage of using the “Opt
All” macro whenever possible. The first pre-set strategy generated the
optimal solution with the highest objective function value, a remarkable
10.78 (T25). The expected profit for the optimal solution derived from the
first pre-set strategy was 10.78 times the absolute value of the maximum loss.
Both the expected profit and maximum loss were calculated net of assumed
transaction costs. This was an unusually large objective function value and
was only possible due to the optimizer’s ability to exploit large systematic
pricing anomalies in the UA matrix.
 Solver Strategy Optimization Problem

All of the optimizations generated return values of either zero or fourteen.

To refresh your memory, the Frontline Solver return values are shown again
in Figure 8.9.
A return value of zero indicates that “Solver found a solution. All
constraints and optimality conditions are satisfied.” A return value of
fourteen indicates that “Solver found an integer solution within tolerance. All
constraints are satisfied.” In this case, all of the optimizations returned
feasible solutions, but the first solution was clearly the best. Always
remember to check the Solver return values for each solution before
proceeding.
 If you have a recent version of Microsoft Excel, you should be able to use
the push-button macros to run Solver automatically, provided you have
enabled the Solver Add-in and the visual basic component. In the event that
you are unable to run the Solver macros or you would like to modify the
Solver problem, you would need to run Solver manually. To help you get
started, I have included Figure 8.10, which is a screen-shot of the Solver
problem used during the strategy optimization process.
Solver attempts to maximize the expected profit per unit of risk, which is
located in Cell L3. It does this by systematically modifying the decision
variables, which are located in Cells C3:K3. As you will recall, these are the
same cells the macros populated with the initial starting point solutions.
As explained earlier, genetic algorithms do not function as efficiently
with constraints, so they were kept to a minimum. However, some constraints
were necessary. Here is a brief description of the most important constraints.
There are three constraints related to the decision variables, which restrict the
decision variables (C3:K3) to integer values with row numbers that
correspond to the appropriate option type (call or put) for each spread
component.
The required capital (P3) is bounded by a user-specified minimum (P4)
and a user-specified maximum (P2). For all of the strategy examples in the
book, I specified a range of $9,000 (P4) to $11,000 (P2).
I also included constraints that allow the user to force the time to
expiration for the call spread and put spread purchases to exceed the time to
expiration for the call spread and put spread sales. If you want to enforce this
limit, enter the value of zero in Cell AG1.
Cell AG1 represents the maximum difference between the trading days to
expiration for spread purchases minus sales. In other words, the trading days
to expiration for spread purchases, minus the trading days to expiration for
spread sales, must be greater than or equal to the value of Cell AG1. If the
time to expiration for spread purchases is greater than or equal to the time to
expiration for spread sales, the Financial Industry Regulatory Agency
(FINRA) considers the spread “covered,” which results in reduced margin
requirements.
I use portfolio margin in my trading account, which allows much greater
flexibility in option trading. As a result, the value of Cell AG1 in the UA
example was negative 75, which permitted spreads that did not meet the
FINRA requirements. The optimal UA solution used this flexibility and
identified a solution that would not have been possible within the FINRA
covered spread requirements.
 Be careful not to equate this flexibility with excessive risk. For the call
and put spreads, the number of options purchased will always equal the
number of options sold. The time to expiration of the option purchased might
be shorter than the time to expiration of the option sold, but the strategy
would only be held until the expiration date of the shorter-term option. It
would not be appropriate to hold the naked short position.
 After running the “Opt All” push-button macro, it was obvious that the
first pre-set strategy produced the optimization solution with the highest
objective function. To reevaluate this solution, we enter the value of one into
Cell J24 (indicating the first strategy solution) and execute the “Copy (Nth)”
push-button macro to copy the first solution into the calculation region of the
“Opt” tab. Columns A through L of the UA solution are shown in Figure
8.11. This section shows the row numbers for each option, the number of
spread contracts, and the objective function value.
 Figure 8.12 is a continuation of the UA solution shown in Figure 8.11.
Each row of Figure 8.12 is a continuation of the rows in Figure 8.11. Row
three represents the UA solution. Figure 8.12 shows the average profit after
transaction costs (M3), the standard deviation of the average profit (N3), the
maximum loss (O3), the required capital (P3), and all of the True
Greeks(Q3:W3).
As discussed earlier, input Cells P2 and P4 represent the user-specified
maximum and minimum required capital constraints entered by the user and
imposed by Solver. The other cells in row two and row four with the blue
shaded background are also user inputs, but are not used in the optimization.
Instead, they provide visual warnings (red shaded background) if the
maximum (Row 2) or minimum (Row 4) values are violated. We already
examined the profitability statistics and True Greeks for the optimal UA
strategy in Chapter 5, so we will not revisit the results again here.
 Figure 8.13 is a continuation of the UA solution shown in Figures 8.11
and 8.12. Each row of Figure 8.13 is a continuation of the rows in Figures
8.11 and 8.12. Row three represents the UA solution. The Deltas of the
options in the optimal UA solution are shown on the left-side of Figure 8.13
in row three (X3:AC3). The cells with the blue shaded background in rows
two and four are warning cells only and are not used to constrain the optimal
Solver solution. In this case, they are designed to warn the user about
solutions that contain deep in-the-money options.
The remaining blue shaded input cell (AG1) was discussed earlier. Cell
AG1 represents the maximum difference between the trading days to
expiration for spread purchases minus sales. If you want to enforce this limit,
enter the value of zero in Cell AG1. If you want to relax this constraint, enter
a large negative number in Cell AG1. The remaining cells all contain
intermediate calculations and are not directly applicable to our discussion of
the optimization process.
 Looking at row numbers is not the easiest way to visualize a complex
strategy. Figure 8.14 reproduces the table that we used in Chapter 5 (Figure
5.11) to summarize the options used to construct the optimal UA strategy.
For each component of the three spreads, the number of contracts, option
type, strike price, expiration date, trading days until expiration, mid-price,
True Delta, and the degree of undervaluation (Rich/Cheap) are provided in
columns J through Q respectively.
 I personally prefer to examine the P&L graph before implementing any
option strategy. Figure 8.15 reproduces the P&L graph that we used in
Chapter 5 to examine the optimal UA strategy. All of the macro buttons,
tables, and data discussed in this chapter are located on the “Opt” tab of the
Integrated spreadsheet. The P&L graph in Figure 8.15 is located on the
“Analysis” tab.
The Integrated spreadsheet will automatically generate the P&L graph for
the solution in the calculation region of the “Opt” tab (C3:K3). The only
user requirement is to execute the “Scale Axes” push-button macro located
near the top-left of the P&L graph (not shown). The “Scale Axes” macro
automatically scales the vertical and horizontal axes based on the range of the
scenario data. If you forget to use the “Scale Axes” macro, the P&K graph
may not be readable.
Summary
 The strategy optimization process is complex and I have included a lot of
detail to help you understand the process. However, the optimization
procedure is almost entirely automated. The easiest way to use the strategy
optimizer is to enter a constant number of trading days (that exceeds your
desired holding period) in Cells K18:P18. Then push the “Opt-All” button
and let Solver search for the strategy with the highest expected profit per unit
of risk.
If you would like to use a more interactive (and faster) approach, feel free
to use your own custom solution or pick the most promising of the pre-set
solutions as the starting point for the optimization.
Regardless of the optimization process you choose, once you find an
attractive solution, examine the P&L graph on the “Analysis” tab to ensure
that you understand every aspect of the profit distribution before
implementing any option earnings strategy.
 9 – CREE Earnings Strategy

In Chapter 5, I demonstrated how the tools in this book could be used to

create a trading edge by exploiting earnings-related pricing anomalies in
Under Armor (UA) options on July 23, 2014. In that chapter, we first used
OptionSlam.com to screen the entire database of stocks for candidates with
high levels of historical earnings volatility and a demonstrated directional
earnings return bias.
We used OptionSlam.com’s earnings return data to evaluate UA’s
historical earnings volatility and we also calculated past and current levels of
implied earnings volatility. This analysis allowed us to forecast the expected
level of realized earnings volatility for UA’s upcoming earnings
announcement (before the open on July 24, 2014).
Similarly, we studied UA’s historical one-day returns immediately
following the previous twelve quarterly earnings announcements and
identified a systematic bullish earnings directional bias. We confirmed that
bullish directional bias by reviewing the Estimize.com consensus revenue
and EPS forecasts relative to Wall Street’s revenue and EPS estimates.
The Integrated spreadsheet employed Solver in an attempt to identify the
UA option earnings strategy with the highest expected profit per unit of risk.
The expected risk and return calculations for every option in the UA matrix
were derived from the comprehensive volatility model parameters, our
specific volatility and directional forecasts, and from the expected corrections
in each option’s relative value during the holding period. Finally, we
examined the profit and loss diagram and the True Greeks for the optimal
strategy.
In Chapters 6, 7, and 8, we examined how to use the Integrated
spreadsheet to perform all of the above functions. We have now covered all
of the concepts, formulas, definitions, and procedural steps required to
evaluate, optimize, and trade option earnings strategies. In this chapter, we
will apply the same investment process that we used in Chapter 5 (described
above) to design one more real-world option earnings strategy. Since we have
covered all of this material in previous chapters, this chapter will focus more
on the investment process and trade strategy and less on definitions, data
sources, and spreadsheet mechanics.
The objective of this exercise is to reinforce your understanding of the
concepts that we have used throughout this book and to gain some additional
practical insights into trading option earnings strategies. As we proceed, keep
in mind that our goal is to use our tools and resources to identify and exploit
a trading edge.

Screening Earnings Candidates

Figure 9.1 is a screenshot of OptionSlam.com’s historical stock screener

with all of the filter settings that I used on September 25, 2014 to find
prospective candidates with earning announcements scheduled during the
subsequent three months. I limited my selection to stocks with weekly
options and an average daily volume of over one million shares. I specified a
minimum EVR of 2.0 and a minimum stock price of $20 per share.
EVR is OptionSlam’s proprietary earnings volatility measure, which is
based on the most recent three years of quarterly earnings announcements,
with the most recent data weighted more heavily. EVR ranges between zero
and ten, with ten being the most volatile.
I further restricted prospective candidates to stocks that experienced a
minimum mean and median absolute price change of 5% or more and a
median raw price change of negative 4% or less over the past twelve earnings
announcements. The resulting candidates experienced excess earnings
volatility and consistently surprised analysts and traders to the downside,
generating median returns of less than negative 4% per quarter.
 Figure 9.2 is a screenshot of OptionSlam.com’s historical screener’s
results on September 25, 2014. The filters I used returned only nine stocks.
For each of the candidates in Figure 9.2, the symbol, next earnings date,
market, sector and EVR are provided on the left side of the table.
The right side of the table includes the mean and median absolute 1-day
percentage price changes over the twelve previous earnings events, followed
by the mean and median raw 1-day percentage price changes over the twelve
previous earnings events. The absolute percentages are a measure of volatility
and the raw percentages indicate directional bias.

After examining the nine candidates, I selected Cree, Inc. (CREE) and
chose to implement the option earnings strategy on September 26, 2014,
several weeks before the next scheduled CREE earnings announcement on
October 21, 2014.
 Why attempt to initiate an option earnings strategy several weeks before
an earnings announcement? Option traders know that short-term options
decay faster than long-term options. They also know from experience that the
implied volatility of options expiring after earnings announcements will
increase as time passes and the earnings date nears.
Many traders presume that we could sell short-term options that expire
before the earnings announcement and buy options that expire after the
earnings announcement and profit from the rapid decay of the short-term
options and from the increasing implied volatility of the long-term options.
This strategy is called an earnings calendar spread.
If this hypothesis is true, then we should be able to use our tools to
identify a candidate with an unusually low level of implied earnings volatility
(IEV), which would further enhance the returns of this strategy. The ideal
candidate’s current IEV would be lower than both previous levels of implied
earnings volatility and historical earnings volatility. This would increase the
probability that the IEV would rise approaching the date of the earnings
announcement, which would enhance the value of our long option position.
Let’s test this hypothesis. To do so, we will need to evaluate CREE’s
implied earnings volatility and historical earnings volatility. The short option
positions would expire before the earnings announcement, so we could not
own the calendar spread strategy past the earnings date, but we might want to
modify the earnings calendar strategy before the earnings announcement or
even employ a different strategy. As a result, to ensure that we are fully
prepared to design strategies for any holding period, we will research both
CREE’s historical earnings volatility and directional bias.

CREE Historical Earnings Volatility

Figure 9.3 is OptionSlam.com’s historical earnings volatility table for

CREE as of September 26, 2014. The 1-day closing percentage returns were
calculated from the closing price on the last trade day before earnings to the
closing price on the next trading day. You will recall that we need to enter the
1-day closing percentage changes from Figure 9.3 into the Basic or Integrated
spreadsheet, which will calculate a series of directional summary statistics
and measures of historical earnings volatility.
 In Figure 9.4, there are three summary measures of the 1-day log returns.
The linear regression, median, and mean log-returns from the twelve previous
earnings events were -8.28% (C18), -4.20% (C19), and -3.61% (C20)
respectively. When combined with the Root Mean Squared Error (RMSE),
the resulting probability of a positive earnings return were 26% E(18), 37%
(E19), and 39% (E20). This indicates the presence of a negative or bearish
directional bias. As shown in Figure 9.4, the annualized earnings volatility
based on the RMSE for the twelve previous CREE earnings events was
206.37% (D22).
 Figure 9.5 is a chart of the log-returns for CREE’s previous twelve
earnings events. As you can see from the graph, only one of the last six
earnings returns was positive, indicating that the directional bias (descending
red trend line) might have worsened recently.
 As we did when we evaluated UA in Chapter 5, it would be useful to re-
calculate the earnings summary return statistics using only the past six
observations, which will allow us to gain further insight into recent changes
in directional bias and earnings volatility.
Figure 9.6 is the Historical Earnings Volatility Table using only the
previous six CREE earnings events. The linear regression, median, and mean
log-returns from the six previous earnings events were -7.09% (C18),
-11.09% (C19), and -10.20% (C20) respectively. The resulting probabilities
of positive earnings returns were 31% (E18), 22% (E19), and 24% (E20).
The linear regression returns for the two sets of earnings events were
similar, but the mean and median returns were almost 7% lower for the past
six earnings events, suggesting an increasing bearish trend. The probabilities
of a positive return derived from the mean and median of the past six
earnings events both declined by 15%. The market had been consistently
disappointed by CREE’s recent earnings announcements. If we ultimately
decide to create an option earnings strategy that we will hold over the
earnings date, we will need to incorporate our directional earnings forecast
into the optimization.
The annualized earnings volatility based on the RMSE for the six
previous CREE earnings events was 232.39% (D22), which was
approximately 26% higher than the annualized earnings volatility for the
twelve previous earnings events (206.37% in Cell D22 of Figure 9.4). This
suggests that not only were the results increasingly bearish, the magnitude of
the market’s reaction to CREE’s earnings announcements had increased as
well. We will factor the recent increase in CREE’s earnings volatility into our
volatility forecasts during the strategy optimization process.
 CREE Historical Implied Earnings Volatility

While it is imperative to know CREE’s historical earnings volatility, we

also need to review how the market has priced past levels of CREE’s
earnings volatility. If we do not hold our earnings strategy past the earnings
announcement, the realized earnings volatility will not affect our strategy.
The change in CREE’s implied earnings volatility over the holding period
will affect the strategy through the True Earnings Vega.
In Chapter 3, we examined how the “ImpliedEV” tab of the Basic
spreadsheet uses the aggregate IV formula and implied volatility data
immediately before and after earnings to solve for the implied earnings
volatility (IEV). Figure 9.7 shows the Historical Market Implied Earnings
Volatility analysis for CREE’s 8/12/2014 earnings announcement. Solver
found that CREE’s IEV immediately before the 8/12/2014 earnings
announcement was 170.5% (C3).
I repeated this analysis for the preceding two earnings announcements as
well. CREE’s IEV immediately preceding the 4/22/2014 and 1/21/2014
earnings announcements were 179.6% and 227.6% respectively (not shown).
The average CREE IEV for the three earnings announcements prior to our
current analysis date was 192.5% (C1). All three of the IEV values exceeded
170% and the average of 192.5% was slightly less than the actual historical
earnings volatility experienced over the previous twelve (206.37%) and six
(232.39%) earnings events.
CREE Directional Confirmation
 Before we proceed to strategy optimization, we will use one additional
source in an attempt to confirm the bearish directional bias that we
discovered in CREE’s historical earnings return data. As we did in Chapter 5
for UA, we will review the Estimize.com consensus EPS and revenue
forecasts for CREE and compare those forecasts to the Wall Street estimates.
We are interested in the 10/21/2014 CREE earnings announcement, which
corresponds to the first fiscal quarter of 2015 (FQ1’15) in the Estimize.com
EPS table in Figure 9.8 and the Revenue table in Figure 9.9 (the far right-
hand-side columns).
The FQ1’15 Estimize.com EPS consensus of $0.33 per share was below
the Wall Street estimate of $0.34 per share. Similarly, the Estimize.com
Revenue consensus of $431.71 million was also below the Wall Street
estimate of $434.55 million. This confirms the bearish directional forecast.
However, if we investigate the data further, the bearish implications are even
more significant. FQ1’15 was the first time in the last six quarters that the
Estimize.com EPS or Revenue consensus was below the Wall Street estimate.
This represents a significant change in the direction of CREE’s EPS and
Revenue consensus estimates for the Estimize.com community, further
supporting the likelihood of a bearish directional surprise from CREE’s
10/21/2014 earnings announcement.
 Scenario Assumptions

We now have enough information to design an option earnings strategy

for CREE on September 26, 2014. For this strategy example, I specified a
holding period from September 26, 2014 to October 17, 2014. The ending
date was selected to coincide with the last option expiration date before the
CREE earnings announcement on October 21, 2014. I also specified that the
resulting strategies should require between $9,000 and $11,000 of capital. We
reviewed the input procedures and Integrated spreadsheet images for these
specifications in the last few chapters and they are not shown here. In this
chapter, our discussion of the Integrated spreadsheet will be limited to a few
select screenshots.
Figure 9.10 is a screenshot from the “Specs” Tab of the Integrated
spreadsheet that shows the assumptions used in all scenario and optimization
calculations. Cells with dark shaded backgrounds and white text represent
user inputs. Cells with white or light colored backgrounds represent
calculation cells and should not be modified. Let’s review several of the most
relevant user inputs.
I entered a value of 10.00% (C17) for the CREE “Percent UV/OV
Correction,” which represents the amount of undervaluation or overvaluation
correction that occurs during the simulation period. You may recall that I
used 100% in the UA strategy example, which was justified because the UA
holding period included the earnings event, which is the principal source of
pricing anomalies in individual stocks. The CREE holding period did not
include an earnings event, which warrants the much lower “Percent OV/UV
Correction” of 10%.
At the close on September 26, 2014, the comprehensive volatility model
used all of the options in the CREE matrix (that passed the input filters) to
estimate an implied earnings volatility of 168.27% (E19). As you will recall,
CREE’s historical earnings volatility over the past twelve earnings events
was 206.37% (F19) and was even higher for the past six earnings events
(232.39% in Cell D22 of Figure 9.6). Earlier in this chapter, we also
calculated the average historical implied earnings volatility for the three
preceding earnings events: 192.5% (Figure 9.7 Cell C1). Based on the
historical values, I assumed (conservatively) that the implied earnings
volatility would increase from 168.27% (Figure 9.10 Cell E19) to 180%
(D19) during the holding period. In other words, the implied earnings
volatility would increase toward CREE’s historical actual and implied
earnings volatility.
The 190% (C19) realized earnings volatility will not affect the simulation
because the holding period does not include an earnings event. Similarly, the
assumed 0% (C21) directional earnings price return will not affect the
simulation either, because the return bias is only realized after an earnings
announcement.
Finally, I assumed slippage of negative $0.10 (C25) per spread trade and
negative 0.75 per contract (D25). This slippage estimate is conservative for
CREE, but slightly overestimating slippage increases the probability that
optimal solutions can be implemented in practice. It should be possible to
execute a very high percentage of CREE spreads within $0.05 of the spread
mid-point.
 Optimal CREE Solutions

I executed the “Opt All” macro, which ran twenty different optimization
runs, from twenty different starting points, and used two different solution
algorithms. The results for each of the pre-set starting points are shown in
Figure 9.11. Figure 9.11 shows the objective function (T), average profit after
transaction costs (U), standard deviation of the profit (V), maximum loss
(W), the required capital (X), and the Solver result (Y).
The objective function represents the expected return divided by the
absolute value of the largest loss incurred in any of the discrete holding
period scenarios. The largest objective function value was 0.01 (T26 & T29)
and the largest expected profit (net of transaction costs) was $125 (U29). The
standard deviation of profits (Column V) and the maximum losses (Column
W) incurred were all significant. The required capital amounts for all
solutions (Column X) were within the specified range of $9,000 to $11,000
and the Solver solutions were all feasible (Column Y).
The Solver solutions were clearly not as impressive as the UA optimal
solution. So what do we do now? Nothing!
This is probably the single most important observation in the book and it
is not limited to option strategies: when we do not have a trading edge, we
should not trade. There are no exceptions to this rule. If we trade without a
demonstrated probabilistic edge, it would be gambling and we would be
destined to lose money as the odds played out against us over time.
 What have we learned from this exercise? This was only one example, but
we assumed that implied earnings volatility would increase as the earnings
date approached and we also had the supposed advantages of accelerating
time decay on short option positions expiring before earnings and increasing
implied volatility of long option positions expiring after earnings. Even with
those supposed advantages, Solver could not construct a single option
earnings strategy that was able to cover transaction costs, even by buying
overvalued options and selling undervalued options.
This should cause us to seriously question the hypothesis that earnings
calendar spreads inherently offer positive expected returns. Our example does
not imply that earnings calendar spreads will never offer attractive risk-
adjusted returns, but the above analysis does suggest that those opportunities
will probably be case-specific, not systematic.
So what do we do now? We invested all of our time and effort into
researching CREE and we have nothing to show for it. Well, that is not
exactly true. We analyzed CREE on September 26, 2014, but we still have
several weeks before the CREE earnings announcement on October 21, 2014.
The historical actual and implied earnings volatility data for CREE will
not change, so neither should our directional and earnings volatility forecasts.
In fact, very few of the user specifications in the spreadsheet will change.
Simply save a copy of the spreadsheet in a CREE directory and re-run the
analysis as often as time permits.
The Integrated spreadsheet provides an automated macro that makes it
quick and easy to import CSV files from the CREE matrix at any time before
the next earnings event. The “Estimate Volatility Model” macro is also
automated, as are the strategy optimization macros. Option prices are not
constant and neither are option pricing anomalies, especially approaching
earnings.
As our OptionSlam.com historical scan indicated, CREE has experienced
a high degree of earnings volatility and a bearish directional bias in the past
and our ability to precisely calculate CREE’s implied earnings volatility gives
us an advantage when evaluating and constructing option earnings strategies,
especially when holding a strategy over an earnings event. We only have to
wait for an opportunity that we can exploit.

CREE Revisited

For the second CREE strategy optimization example, I specified a one-

day holding period from October 21, 2014 to October 22, 2014, which was
selected specifically to include the earnings event. I again specified that the
resulting strategies should require between $9,000 and $11,000 of capital.
The Integrated spreadsheet images for these specifications are not shown
here.
Figure 9.12 is a screenshot from the “Specs” Tab of the Integrated
spreadsheet that shows the scenario assumptions used in all scenario and
optimization calculations on October 21, 2014. Let’s review several of the
most significant user inputs.
I entered a value of 100% (C17) for the CREE “Percent UV/OV
Correction,” which represents the amount of undervaluation or overvaluation
correction that occurs during the simulation period. This assumption is the
same as the 100% “Percent UV/OV Correction” used in the UA strategy
example. The 100% correction is plausible because the holding periods
include the earnings event.
At the close on October 21, 2014, the comprehensive volatility model
used all of the options in the CREE matrix (that passed the input filters) to
estimate an implied earnings volatility of only 135.46% (E19). This is much
lower than CREE’s IEV that we estimated on September 26, 2014 (168.27%
in Figure 9.10, Cell E19) and is significantly below CREE’s historical actual
earnings and implied earnings volatility, which should offer attractive
strategy solutions.
CREE’s historical earnings volatility over the past twelve earnings events
was 206.37% (F19) and was even higher for the past six earnings events
(232.39% in Figure 9.6, Cell D22). In addition, the average historical implied
earnings volatility (IEV) for the three preceding earnings events was 192.5%
(Figure 9.7 Cell C1).
Based on the historical values, I assumed that the implied earnings
volatility would increase from 135.46% (Figure 9.12 Cell E19) to 160%
(D19) during the holding period. The value of 160% was still well below
previous IEV levels for CREE, but the length of the holding period was only
one day. However, earnings price shocks can result in material changes in
implied earnings volatility.
This time, the 200% (C19) realized earnings volatility will affect the
simulation because the holding period now includes an earnings event.
Likewise, the assumed negative 4% (C21) directional earnings price return
will also affect the simulation results, because the return bias is realized after
an earnings announcement.
The 200% (C19) realized earnings volatility assumption is consistent with
the realized earnings volatility over the previous twelve earnings events
(206.37% in Cell F19), but is still lower than the historical earnings volatility
over the past six events (232.39%). The negative 4% (C21) directional
earnings price return assumption is conservative, especially relative to the
mean and median earnings returns over the past six earnings events. The
actual earnings return could be substantially worse than negative 4%, but it is
wise not to use extreme assumptions when generating optimization scenarios.
As was the case in the first CREE example, I assumed slippage of
negative $0.10 (C25) per spread trade and negative 0.75 per contract (D25).
Again, this slippage estimate is conservative for CREE, but slightly
overestimating slippage increases the probability that optimal solutions can
be implemented in practice.
 Optimal CREE Solution

Figure 9.13 is a screenshot from the “Opt” tab of the Integrated

spreadsheet that shows the optimal CREE solution found by the Solver
optimizer. There are eight columns from left to right: amount (number of
contracts), type (call or put), strike price, expiration date (YYMMDD), T-
Days (number of trading days until expiration), Mid Price, T-Delta (True
Delta), and Rich/Cheap (degree of over or undervaluation).
The call spread is shown in the first two rows, the put spread in the next
two rows, and the strangle or straddle in the last two rows. The number of
contracts in the call spread was zero (J7:J8), so let’s review the put spread
first.
The optimal solution required the sale of 2 (J9) put options with a strike
price of $28.00 (L9) and an expiration date of 11/01/2014 (M9) and the
simultaneous purchase of 2 (J10) put options with a strike price of $37.50
(L10) and an expiration date of 12/20/2014 (M10). The put option sold was
rich or overvalued by $0.05 (Q9) per share, per contract. The put option
purchased was cheap or undervalued by $0.09 (Q10) per share, per contract.
The final component of the optimal UA strategy was a diagonal strangle,
which required the purchase of 18 (J11) call options with a strike price of
$33.50 (L11) and an expiration date of 11/01/2014 (M11) and the
simultaneous purchase of 18 (J12) put options with a strike price of $35.00
(L12) and an expiration date of 12/20/2014 (M12). In this case, the call
option purchased was cheap by $0.01 (Q11) per share, per contract and the
put option purchased was also cheap by $0.11 (Q12) per share, per contract.
The solution only requires two spread transactions, which makes it relatively
easy to manage and to execute.

P&L Graph

The ability to fully appreciate the nature of a strategy by mentally

combining multiple spreads would be very difficult. That is why we use
profit and loss diagrams, which are common to all option analytical
platforms. Figure 9.14 is a profit and loss (P&L) diagram for the optimal
strategy described in Figure 9.13. The scenario values accurately incorporate
the effects of earnings volatility and normal volatility for every option in the
matrix, as do the True Greeks.
 The independent x-axis represents the CREE stock price in dollars. The
dependent y-axis on the left side of the diagram denotes the profit and loss of
the optimal strategy for three different dates.
The solid green upper “P&L 0” represents the P&L at time zero, or the
instantaneous profit and loss of the strategy. The solid purple lower “P&L U”
line represents the P&L on the future date specified by the User, which is the
date used to optimize the strategy. In this example, the User date was one
trading day into the future, or October 22, 2014. The solid black lower “P&L
X” line represents the P&L on a second future date specified by the User,
which is intended to be the expiration date of the option with the shortest-
term option.
In this example, the “X” date was also one trading day into the future, or
October 22, 2014. Given that the “U” and “X” dates are the same, the lower
lines overlap and only appear as a single solid (lower) black line in Figure
9.14.
The “U” and “X” probabilities are shown by the faded dashed lines and
correspond to the values on the dependent y-axis on the right side of the
diagram. In addition to the P&L graph, I have included summary statistics for
the optimal CREE strategy at the top of the diagram.
The optimizer always uses the User date, so let’s focus our attention on
the lower P&L line. The majority of the P&L line was above the break-even
or zero profit line except for a narrow region of the diagram centered between
the prices of $31 and $37. The maximum simulated strategy loss of negative
$1,855 (including transaction costs) occurs at the lowest point in the diagram,
which corresponds to a post-earnings price of $33.49. The price of CREE
before earnings on October 21, 2014 was $33.15.
The optimal hybrid strategy would have performed very well for price
declines in CREE, which should be no revelation. Our research indicated that
the price of CREE was likely to decline due to a bearish earnings surprise and
we used a negative 4% price shock to create our simulated price distribution.
This explains why the probability distributions (dashed lines) are centered
below the pre-earnings price of $33.15. The optimizer recognized our bearish
forecast and designed a strategy to profit from the expected price decline.
Delta measures the directional exposure to price and the True Delta of the
strategy was -305.80, reflecting a strategy that would perform well when the
price of CREE declines.
The shape of the P&L function was convex or curved upwards, indicating
the strategy would have performed well for large moves up or down in the
price of CREE. You will recall that Gamma measures the curvature of the
option price function, so based on the shape of the diagram, we would expect
the optimal strategy to have positive Gamma. The True Gamma was +355.3.
The difference between the upper “P&L 0” line and the lower “P&L U”
and “P&L X” is due to the effects of the passage of time, which is measured
by Theta. The positive value of Gamma means that the strategy must have
had a negative Theta. The True Theta was -$1,579.3, which fully reflected the
volatility crush due to passing the CREE earnings date. The Black-Scholes
(BS) Theta was only -$136.6, which grossly understated the true cost of the
strategy due to the passage of time.
The strategy had a positive True Normal 30 Vega (+$83.4) and a positive
True Earnings Vega (+$19.9), which is consistent with our forecast of a
higher than expected level of earnings volatility.
The total transaction costs were -$460. The expected profit (net of
transaction costs) was $1,803 and the maximum loss (net of transaction costs)
was -$1,855. The objective function, which is what the optimizer was
attempting to maximize, was the expected profit ($1,803) divided by the
absolute value of the maximum loss ($1,855). The expected profit can be
found on the “Opt” and “Analysis” tabs, but was not shown in Figure 9.14.
The resulting value of the objective function for the optimal strategy was
an attractive 0.97, which means the probability-weighted expected profit
from the strategy was 0.97 times the maximum loss. The objective function
value of 0.97 represents the expected profit per unit or dollar of risk. In this
case, risk was defined as the maximum loss (net of transaction costs) of the
strategy in any of the discrete scenarios ($1,855).
The optimal objective function value of 0.97 for CREE was appealing,
but much lower than the remarkable 10.78 for UA. The CREE option matrix
was much more efficiently priced than the UA matrix. As a result, we could
not exploit undervalued and overvalued CREE options as effectively as we
could when we designed the UA strategy. The average pricing error for the
CREE option matrix was $0.099 (not shown), which was much less than the
$0.175 (Figure 7.2 Cell S12) average pricing error for the UA option matrix.
Confirmation
 As we did earlier when we created the UA option earnings strategy, we
will use OptionVue’s graphical analysis to look for any major disparities in
the strategy results that would warrant further research. I entered the optimal
CREE solution from Figure 9.13 into the OptionVue software and performed
a graphical analysis of the strategy on October 21, 2014. OptionVue’s
analytical results are depicted in Figure 9.14B. OptionVue’s results do not
include the directional or earnings volatility forecasts that we estimated
earlier in this chapter.
The CREE stock price is shown on the independent x-axis and the
strategy profit or loss (in dollars) is shown on the dependent left-vertical axis.
The strategy return as a percentage of required capital is shown on the
dependent right-vertical axis. The dashed line represents the strategy results
for the “T+0” line, which stands for trade date plus zero days (instantaneous
P&L). The solid line depicts the strategy results for the “T+1” line, which
represents the P&L for a one-day holding period. These two lines are directly
comparable to the strategy results shown in Figure 9.14.
The two horizontal bars at the bottom of OptionVue’s graphical analysis
show the expected one and two-standard deviation price moves over the one-
day holding period. The one and two-standard deviation forecasts both use
the aggregate implied volatility formula to accurately incorporate the effects
of earnings and non-earnings volatility over the specified holding period.
Despite the many differences between the two models, the P&L forecasts
were generally consistent.
 Actual Strategy Results

So how did the optimal hybrid CREE strategy actually perform? Figure
9.15 is the same P&L graph from Figure 9.14 with one addition. The Red
diamond represents the actual profit of the optimal strategy (net of transaction
costs) from October 21, 2014 to October 22, 2014. The actual price of CREE
decreased from $33.15 to $27.28. Our bearish forecast was correct and the
realized earnings volatility was also greater than the implied earnings
volatility.
As you would expect, the optimal strategy performed well, earning a total
profit of $6,482 (net of transaction costs). The return on required capital of
$9,488 (not shown) was 68.32% ($6,482/$9,488) over the one-day holding
period. At a price of $27.28, the model forecasted a strategy value of $6,344,
meaning that the actual strategy performance was almost identical to the
modeled strategy performance.
Conclusion
 The purpose of this chapter was to provide another practical example of
how to use the Exploiting Earnings Volatility analytical approach and toolset
to design a realistic option earnings strategy, based on actual market prices
and actual earnings data.
The first pre-earnings CREE example taught us a very important lesson:
we do not always have a trading edge. When we do not have a trading edge,
we should not trade. When we do have an edge, we should exploit our
advantage, but must always manage our risk.
I normally advocate using stop orders to exit positions at predetermined
levels and position sizing to ensure our total losses are small relative to our
trading capital. Unfortunately, stop loss orders are of limited use when
trading options, primarily due to the importance of managing execution levels
and minimizing slippage. Even if we could place stop loss orders for our
hybrid option spreads, they would be completely ineffective in limiting risk
during earnings announcements.
First, earnings are typically announced when the option market is closed.
Second, prices make discrete jumps when earnings are announced. Even if
the option market was open, the stop orders would not work.
As a result, it is critical to use positing sizing formulas to limit risk when
trading option earnings strategies. The maximum loss that could be incurred
(which will exceed the maximum loss depicted in the P&L diagram) should
always be considered when calculating the position size for your option
earnings strategies. Your position size should be reduced as the maximum
potential loss of the strategy increases. This ensures that your total loss will
remain below the desired threshold, which is expressed as a percentage of
your total capital.
 10 – Practical Considerations

The purpose of this final chapter is to provide some practical suggestions on

how to get the most out of the tools and concepts introduced in this book. We
will examine how to use these tools efficiently to exploit earnings volatility,
and we will also explore how this framework could be applied to improve
volatility research and to generate new profit opportunities.

Efficient Use of Tools

One important benefit of option earnings plays is that we have hundreds

of new trade opportunities every quarter. We know that earnings will be
announced every quarter and we even know the exact timing of the
announcement in advance. Given that we now have the ability to calculate
actual historical earnings volatility, historical implied earnings volatility, and
real-time implied earnings volatility, we have a material advantage over the
typical retail option trader. To fully exploit our trading edge on the full
spectrum of earnings candidates, we have to use our new tools efficiently.
The first step is to automate the data collection process. To calculate
historical earnings volatility, we need historical earnings return data. I use
OptionSlam.com, which also provides extensive screening tools to help
identify prospective trading candidates. Analyzing earnings volatility requires
a lot of data and it would be impractical to gather this data manually,
especially for a number of different candidates. It would also increase the
likelihood of data errors, which would compromise the integrity of our
analytical results and could lead to erroneous trades.
To make full use of the input macros in the Integrated spreadsheet, we
also need a source of current and historical option data. I use OptionVue,
which can export the option matrix data in a format that is compatible with
the import macros in the Integrated spreadsheet. The import macros read
CSV files in a specific format, which could potentially be generated from
other sources. Regardless, we will want to use the tools in the Integrated
spreadsheet often, so it is critical to find a reliable source of option matrix
data in the designated CSV format.
The next step is to generate and monitor earnings candidates with a
history of volatile or directionally biased earnings returns. I use
OptionSlam.com’s custom screening tools to identify earnings candidates and
regularly calculate each candidate’s implied earnings volatility and normal
volatility. OptionVue also has an “Earnings Plays” subscription service (see
Resources section) that could be used to identify earnings strategy candidates.
The best way to develop your earnings volatility intuition is to frequently
use the Basic and Integrated spreadsheets to calculate implied earnings
volatility and normal volatility. If possible, calculate these values daily for all
of your earnings candidates. At a minimum, begin one month before the
earnings announcement and end shortly after the announcement.
Save this data in a database or spreadsheet for further research. Once you
have saved the historical actual and implied earnings volatility values in your
database, you will only need to add new data to the database every quarter;
the historical values never change. This data will be invaluable as you create
option earnings strategies in the future.
I also recommend saving the data you use in the Integrated and Basic
spreadsheets in a separate spreadsheet. Simply copy and paste the values into
the separate spreadsheet or tab for future use. When you want to re-run the
analysis on a previous candidate, copy and paste the saved data back into the
Basic or Integrated spreadsheet and add the most recent data before re-
running the analysis.
During earnings season, regularly calculate the implied earnings volatility
for each candidate and look for large anomalies that you can exploit. If the
anomalies are small, there is no need to waste your time by running an
optimization. However, if the anomalies are large, run the strategy optimizer
in the Integrated spreadsheet and evaluate the expected profit per unit of risk
for the resulting strategies.
If the risk-adjusted expected profits are attractive, examine the actual
market prices and confirm that execution at the desired prices is likely. If so,
enter limit orders for the option spreads. If executed, manage your strategy
using the P&L graph, expected profit, maximum loss, and other strategy
summary statistics from the Integrated spreadsheet to make informed exit and
risk management decisions.
That may sound like a lot of work -- which is why we get paid for it.
Exploiting market inefficiencies takes time and effort, which is why the
inefficiencies exist. If you only have time to track a few candidates, then
become an expert in those candidates and only execute trades when your
strategies’ expected risk-adjusted profits are attractive.
 Volatility Research

I primarily trade algorithmic strategies and many of my strategies use

volatility-based indicators for decision rules and filters. The volatility-based
indicators use historical volatility, implied volatility, volatility skews, and
option volume in a variety of interesting ways, which include the use of
trends, oscillators, and divergences. These volatility-based tools work
extremely well for strategies that are based on equity indices and futures, but
on not individual stocks. Why? Because the discrete effects of earnings
events on the underlying prices, implied volatility, and realized volatility of
individual stocks compromise the effectiveness of actual and implied
volatility indicators.
As we have discussed throughout this book, implied volatility equals the
weighted-average of earnings and non-earnings volatility, and the weighting
is a function of the relative number of earnings and non-earnings trading days
remaining until option expiration. As a result, implied volatilities of options
on individual stocks are not comparable across time or among different
candidates. Implied volatility represents the aggregation of two independent
sources of volatility, both of which should be analyzed independently.
Until now, it had not been possible to separate earnings and non-earnings
volatility. The aggregate implied volatility formula and the spreadsheet tools
introduced in this book allow us to isolate and quantify implied earnings
volatility, normal implied volatility, and all of the components of normal
volatility.
If we are interested in building a forecasting model for realized or implied
earnings volatility, we have to evaluate historical earnings volatility and
historical levels of implied earnings volatility directly. It will not help us to
study implied volatility, which includes the effects of both earnings and non-
earnings volatility.
Similarly, if we want to build volatility-based indicators to help us make
timing decisions on individual stocks, we need to use historical and current
levels of normal or non-earnings volatility. We could go even further and
study the historical volatility model parameters derived by the comprehensive
volatility model solutions in the Integrated spreadsheet.
The key is to isolate and quantify implied earnings volatility, normal
implied volatility, and all of the components of normal volatility before
attempting to develop volatility-based indicators or strategies. The
components of volatility are comparable across time and among individual
stocks, while implied volatility is not.

Trading Edge

Separating and quantifying volatility components and saving the resulting

data is a great beginning, but our ultimate goal is to use our volatility insights
to improve our trading edge. How could we go further?
We have already examined the basic technique of evaluating historical
earnings volatility statistically to identify directional biases. We could also
use more sophisticated tools to forecast directional price moves due to
earnings events: linear regression, non-linear regression, multiple regression,
time series analysis, or even neural network models. In addition to evaluating
past earnings events for each company, we could also evaluate the earnings
volatility for each sector.
This would allow us to answer some interesting questions. How does the
earnings volatility for each sector change over time? Is it a function of the
economic environment? Is the relative earnings volatility of each company in
a sector stable over time? Could the relative ranking of each company within
the sector be used to forecast the earnings volatility of each company more
accurately?
How do the directional earnings returns for each sector compare over
time? Are there observable patterns in the sector directional earnings returns?
Could the directional earnings returns for other companies in the sector (that
have already reported) help us forecast directional returns for companies in
the same sector that have yet to report?
Are the earnings volatilities of some sectors chronically mispriced?
Which sectors have offered the best directional and non-directional earnings
opportunities historically? Why? Will these trends continue? The spreadsheet
tools provided with this book allow us to generate the required data to answer
all of these questions, which should greatly enhance our trading edge.
We used the Estimize.com consensus EPS and Revenue estimates to
confirm our directional earnings return forecasts in Chapter 5 and Chapter 9.
Could we use this data directly to identify trading opportunities? In
discussing this book with Euan Sinclair, a proprietary option trader and
author of Volatility Trading, Sinclair suggested that when the dispersion of
earnings estimates is low, earnings surprises are larger. In other words,
realized earnings volatility is inversely correlated with the dispersion of
earnings estimates. This is an interesting hypothesis and one that we could
confirm and exploit by downloading and analyzing the Estimize.com data.
Before we explore other potential applications of this framework, I
wanted to pass along a practical insight from a professional option trader with
decades of experience. Frank Fahey is a former floor trader and still trades
options professionally. In addition, he is a mentor for Discover Options, the
educational arm of OptionVue. Fahey recommended contacting the investor
relations department of the company and confirming the time and date of the
earnings announcement before implementing any option earnings strategy.
Most earnings databases are reasonably accurate, but there will be errors.
Think about the cost of using the wrong earnings date in our analysis. If our
earnings date was off by a single day, it could change the implied volatility of
short-term options by 25% or more. It would compromise our entire analysis.
All of our volatility model, valuation, risk, and optimization results would be
wrong. Learning from experienced and successful traders is one of the best
ways to ensure our investment process will work in practice.

New Applications

We have applied the aggregate implied volatility formula specifically to

earnings events, but the concept and formula are not limited to earnings
announcements. We could use the same conceptual framework and aggregate
implied volatility formula to quantify the implied volatility of any discrete
event that provides material new information to market participants. The
only requirement is that we know the expected time and date of the release of
that information in advance.
One example would be FDA drug approvals, which obviously have the
potential to create extreme price moves in the underlying stock. If we know
the approximate FDA announcement date, we could use that date and the
aggregate implied volatility formula to solve for the IEV.
Throughout this book the abbreviation “IEV” signified implied earnings
volatility. In this case, IEV would represent the implied event volatility. The
framework, formulas, and spreadsheet tools that we used to evaluate earnings
volatility could be used to analyze the volatility associated with any discrete
event. If the exact announcement date was not known in advance, we could
study the term structure of volatility to determine the market’s expected
distribution of possible announcement dates.
There is only a single FDA announcement for any drug, so it would be
impossible to compare past actual event volatilities or historical IEV levels to
current IEV levels. However, it would be possible to analyze actual event
returns and historical IEV levels for similar classes of drugs or drugs that had
similar potential impact on a company’s future revenue. We could compare
these values to the current IEV for the new prospective drug to identify
potential event volatility anomalies.
We could even perform a similar set of IEV calculations for corporate
legal judgments and Supreme Court rulings. Obviously the scope of these
legal rulings would be limited to specific companies in certain industries, but
for these companies, the impact could be dramatic. Our ability to calculate
the IEV and evaluate it in the context of a wide range of other market moving
events gives us an advantage. The framework and aggregate implied
volatility formula are applicable to any discrete event that provides material
new information to market participants.
What about releases of economic data? These releases move the market
and the release dates are known in advance. Non-Farm Payroll (NFP) is one
of the major monthly economic releases. We could use index options to
calculate the IEV associated with the monthly NFP report. We could analyze
the historical index returns on past NFP dates and we could even calculate
historical IEV values. This would give us the appropriate context to evaluate
the IEV associated with the current NFP report.
We could solve for the NFP IEV for several equity indices. What if they
were materially different? Could that create a trading opportunity? What if
the NFP IEV was the same as the normal volatility, but the historical NFP
volatility was greater than the norm. That would suggest that the NVP IEV
was too low and index options did not reflect the additional volatility
associated with the NFP report.
We could calculate similar IEVs for other major economic releases, Fed
meetings, etc. All of the resulting IEVs would be analogous. In other words,
they would all be expressed as annualized standard deviations, which are
directly comparable. This analysis might give us additional insight into
optimal timing of initiating our market-neutral option income strategies. The
number and type of potential applications are almost limitless.
In our analysis of earnings volatility, we implicitly assumed that the
implied earnings volatilities for all future company earnings announcements
were equal. This is probably a reasonable assumption. However, if we use the
aggregate implied volatility formula to calculate the IEV for multiple
independent events, then we would need separate variables for each IEV and
each event would have its own weight. The logical extension of the aggregate
implied volatility formula to multiple independent events is straightforward.

NTDE1 = Number of Event-1 Trading Days

NTDE2 = Number of Event-2 Trading Days
NTDEN = Number of Event-N Trading Days
NTDN = Number of Normal Trading Days
NTDT = Total Number of Trading Days

WE1 = Event-1 Volatility Weight

WE2 = Event-2 Volatility Weight
WEN = Event-N Volatility Weight
WN = Normal Volatility Weight

WEN = NTDEN/( NTDT)

WN = NTDN/( NTDT)

IV = Aggregate Implied Volatility (Annualized)

VE1 = Event-1 Volatility (Annualized)

VE2 = Event-2 Volatility (Annualized)
VEN = Event-N Volatility (Annualized)
VN = Normal Volatility (Annualized)

IV = ((WE1*(VE1^2))+(WE2*(VE2^2)) +…(WEN*(VEN^2))+(WN)*
(VN^2))^0.5

As you can see from the general aggregate implied volatility formula
above, we can use this version of the formula to solve for aggregate implied
volatilities that are influenced by any number of independent discrete events.
We could even use this formula to estimate the forward volatilities implicit in
the term structure of volatilities, otherwise known as the horizontal skew. As
was the case with the aggregate implied volatility formula presented earlier,
all of the weights in the general aggregate implied volatility formula must
sum to one (100%).
A thorough exploration of the new topics introduced in this chapter is
beyond the scope of this book. Nevertheless, I wanted to share with you some
ideas to get you started on your own research projects and to help convey the
potential for this innovative new analytical framework.

Conclusion

Exploiting Earnings Volatility has introduced a wide range of concepts

and practical tools:
- Two versions of the aggregate implied volatility formula, one for
earnings announcements and a second for multiple independent discrete
events.
- An analytical framework with practical tools that simultaneously
estimated IEV, normal implied volatility, and all components of normal
implied volatility including vertical and horizontal skews.
- A calculation methodology with practical tools that calculate True
Greeks for every option in the matrix and for option earnings strategies.
- A non-linear integer strategy optimization tool that attempts to identify
the option earnings strategy with the highest expected profit per unit of
risk.
- A graphical framework to evaluate the resulting strategies.
- Real-world examples with actual market data that demonstrate how to
use this analytical framework and tool set to evaluate, optimize, and trade
option earnings strategies.- A graphical framework to evaluate the
resulting strategies.
- Suggestions for applying this approach in new ways to further expand
your trading edge.

The primary objective of this book was to introduce a theoretically sound

framework for quantifying the effects of earnings volatility on all options and
to provide tools that will help traders apply that framework in practice.
I also strived to present the material in a way that would be accessible and
educational for new traders, while still offering some interesting challenges
and insights for experienced option professionals. The diversity in readers’
backgrounds was the principal reason that I provided two separate
spreadsheets.
I recognize that very few readers will immediately apply all of the
concepts and tools included in the book and spreadsheets, but trading is a
journey, not a destination. Our goals should be to continuously expand our
knowledge and investigate new opportunities to achieve and maintain our
trading edge. The concepts and tools presented here were designed to help
you pursue those goals – now and in the future.
Thank you for investing your time and effort to review the many new and
challenging concepts presented in this book. I hope you found the earnings
volatility framework and the insights in this book to be helpful and I hope
they greatly enhance your ability to evaluate, construct, and manage your
option earnings strategies. Good luck in your trading.
 About the Author

Brian Johnson designed, programmed, and implemented the first return

sensitivity based parametric framework actively used to control risk in fixed
income portfolios. He further extended the capabilities of this approach by
designing and programming an integrated series of option valuation,
prepayment, and optimization models.
Based on this technology, Mr. Johnson founded Lincoln Capital
Management’s fixed income index business, where he ultimately managed
over $13 billion in assets for some of the largest and most sophisticated
institutional clients in the U.S. and around the globe.
He later served as the President of a financial consulting and software
development firm, designing artificial intelligence-based forecasting and risk
management systems for institutional investment managers.
Mr. Johnson is now a full-time proprietary trader in options, futures,
stocks, and ETFs primarily using algorithmic trading strategies. In addition to
his professional investment experience, he also designed and taught courses
in financial derivatives for both MBA and undergraduate business programs.
His first book, Option Strategy Risk / Return Ratios: A Revolutionary New
Approach to Optimizing, Adjusting, and Trading Any Option Income
Strategy, was published in 2014 and has received outstanding reviews. He
has also written articles for the Financial Analysts Journal, Active Trader,
and Seeking Alpha and he regularly shares his trading insights and research
ideas as the editor of www.TraderEdge.Net.
Mr. Johnson holds a B.S. degree in finance with high honors from the
University of Illinois at Urbana-Champaign and an MBA degree with a
specialization in Finance from the University of Chicago Booth School of
Business.

Email: [email protected]
 Resources

I write a wide range of free, informative articles on www.TraderEdge.Net.

The goal of Trader Edge is to provide information and ideas that will help
you enhance your investment process and improve your trading results. The
articles cover many different topics: economic indicators, technical analysis,
market commentary, options, futures, stocks, exchange traded funds (ETFs),
strategy development, trade analysis, and risk management. You will find
educational articles that appeal to the beginner, as well as advanced tools and
strategies to support more experienced traders.
Trader Edge also offers a subscription to one of the proprietary strategies
that I developed and trade in my own account. The Trader Edge Asset
Allocation Rotational (AAR) Strategy is a conservative, long-only, asset
allocation strategy that rotates monthly among five large asset classes: large-
cap U.S. stocks, developed country stocks in Europe and Asia, emerging
market stocks, U.S. Treasury Notes, and commodities. The strategy was
inspired by the “Ivy League portfolio” and uses trend and technical filters to
reduce downside risk. The AAR strategy has generated approximately 20%
annual returns over the 20+ year combined back and forward test period.
Please use the following link to learn more about the AAR strategy:
http://traderedge.net/order/aar-strategy/.
During the course of my research for this book, I collaborated with the
owners of OptionSlam.com on several enhancements for their site that will
help all traders who use option strategies to trade earnings announcements.
Examples from the OptionSlam.com site were used throughout this book;
many thanks to OptionSlam.com for their permission to use these
screenshots.
Given the strong synergies between OptionSlam.com and the tools in this
book, OptionSlam.com has agreed to offer an exclusive 15% discount on
annual INSIDER Memberships to readers who purchased this book.

OptionSlam.com INSIDER Membership

The following benefits are provided to all INSIDER OptionSlam.com

Members:
- View Earnings History of Individual Stocks
 - View Volatility History of Individual Stocks
- View Straddle Tracking History of Individual Stocks
- View and Customize the Upcoming Earnings Filter
- View and Customize the Earnings Calendar
- View Weekly Implied Volatility Report
- View and Customize the Best Trending Stocks Report
- View and Customize the Current Monthly and Weekly Straddles Report
- View and Customize the Historical Straddles Report
- View Trades from All Members
- Customize and Schedule Email Alerts of Personalized Reports
- Export Earnings Statistics to Excel
OptionSlam.com’s historical earnings data provides all of the return and
volatility data necessary to evaluate past earnings performance. The
“Upcoming Earnings Filter” is a powerful and flexible tool that will help you
efficiently identify both directional and non-directional trading candidates.
I encourage you to visit
https://www.optionslam.com/partner_info/traderedge and take advantage of
the exclusive 15% Trader Edge referral discount. Note the underscore (“_”)
between “partner” and “info” in the above link.

OptionVue

Several of the graphs in this book represent screenshot images captured

from OptionVue’s software; many thanks to Len Yates and OptionVue for
their permission to include these images. Trading options without a
comprehensive option analytical platform is not advisable and the OptionVue
software is one of the most powerful tools available. Unlike most broker
platforms, OptionVue evaluates both the horizontal and vertical volatility
skews, resulting in much more realistic calculations and more accurate risk
and valuation metrics. In addition, I worked with OptionVue to help them
apply the aggregate implied volatility formula to quantify the effects of
earnings volatility before and after earnings events in the OptionVue
software.
The OptionVue software also includes a very powerful “Trade Finder”
module, which is similar to the strategy optimization tool in the Integrated
spreadsheet. Trade Finder allows the user to specify an objective, strategy
candidates, filters, and forecast adjustments and uses those inputs to search
for the best possible strategy. Most important, Trade Finder uses the
aggregate implied volatility formula to accurately incorporate the effects of
earnings volatility in its analysis.
OptionVue also recently released a new subscription service specifically
designed for “Earnings Plays.” OptionVue’s description of the five Earnings
Play’s strategies follows:
- Prime Movers: Stocks that make big moves - options tend to be
undervalued.
- Prime Non-Movers: Stocks that make smaller-than-expected moves,
options tend to be overvalued.
- Earnings Pairs: Two stocks in the same industry, only one of which is
announcing earnings.
- Echoes - Two stocks in the same industry, with one announcing 1-18
days after the other.
- Runners - Stocks that tend to "run" in price after the earnings
announcement.
This system is based on the hypothetical results actual trades would have
experienced in the past and shows you a quality ranking for each trade along
with its past success rate.
OptionVue also offers real-time and historical option prices, which can be
used to backtest option strategies, even with adjustments. Finally,
DiscoverOptions, the educational arm of OptionVue, offers one-on-one
personal option mentoring from professional option traders with decades of
experience.
Through our referral agreement, OptionVue is offering an exclusive 15%
discount on the initial purchase of any annual subscription of any OptionVue
product and on all DiscoverOptions educational products. However, the
discount is not available to current OptionVue clients with an active
OptionVue subscription. Please use the coupon code “traderedge” (lower
case with no spaces or quotation marks) to receive your 15% discount when
ordering applicable products from OptionVue online or over the phone.
I encourage you to visit http://www.optionvue.com/traderedge.html and
take advantage of the exclusive 15% Trader Edge referral discount. If you
would prefer to evaluate the OptionVue software before placing an order, the
above link will also allow you to enroll in a free 14-day trial of OptionVue’s
option analytical platform.
 CSI

Reliable prices are essential for developing and implementing systematic

trading strategies. Commodity Systems Inc. (CSI) is one of the leading
providers of market data and trading software for institutional and retail
customers. Please use the following link to learn more about CSI’s pricing
subscriptions: https://csicheckout.com/cgi-bin/ua_order_form_nw.pl?
referrer=TE. Note the underscore (“_”) between “ua” and “order” and
between “order” and “form” and between “form” and “nw” in the above link.
I am a paying customer of OptionSlam.com, OptionVue, and CSI. My
company, Trading Insights, LLC, has an affiliate referral relationship with
OptionSlam.com, OptionVue, and CSI.

Spreadsheet Tools

Purchasing this book entitles you to an individual user license to

download and use the associated risk/return Excel spreadsheets for your own
research. However, you may not transfer or share the copyrighted
spreadsheet, passwords, or download links with others.
There are two Excel spreadsheets that accompany this book. The name of
the first spreadsheet is EEVBasic.xls. The Basic spreadsheet uses some
simplifying assumptions to reduce the amount of input data required from the
user, but still estimates current and historical earnings volatility and uses
those estimates to forecast future levels of implied volatility around earnings
announcements. The Basic spreadsheet does not use macros and should be
compatible with many different versions of Excel.
The name of the second spreadsheet is EEVIntegrated2.xlsm. The
Integrated spreadsheet requires more data, but is far more sophisticated and
powerful than the Basic spreadsheet. It does not use simplifying assumptions.
Instead, it simultaneously estimates all of the volatility parameters required to
model every option in the matrix. These parameters completely describe the
vertical and horizontal skews as well as level of earnings volatility implied by
option prices.
The Integrated spreadsheet uses this comprehensive volatility model to
calculate “True Greeks” and accurate simulated option values. Excel’s
integrated “Solver” optimization tool is then used by the spreadsheet to
identify optimal strategies to maximize risk-adjusted returns. The Integrated
spreadsheet uses push-button macros to perform all functions. Due to the
extensive use of macros and other functions, the Integrated spreadsheet may
only be fully compatible with the latest versions of Microsoft Excel.
The spreadsheets are included in an encrypted, self-extracting, zip file.
Many cells in the spreadsheet are protected or validated to ensure the
formulas function correctly. However, you may still use the worksheets
interactively.
To download the zip file, go to the following page on Trader Edge:
http://traderedge.net/eev-spreadsheets-2/ and follow the download
instructions.
You will need the following case-specific password to unencrypt the zip
file: UnlockEEVZip891254
You will also need a separate case-specific password to open the Excel
files. The password is the same for both Excel files:
OpenEEV245887
If the Trader Edge website is not accessible, please send me an email with
an explanation of the error received to ([email protected]). Include
your copy of the electronic receipt for the purchase of this book and I will
send you a copy of the zip file as an email attachment.
Given the complexity of the spreadsheets and macros, it is possible that a
few coding or formula errors survived the debugging process. If so, it is
likely that these will be discovered after publication. Please send me an email
with a detailed description of any coding or formula errors that meet ALL of
the following criteria:
1. IS reproducible in the latest version of Microsoft Excel.
2. IS NOT a function of a specific set of user data or input values.
3. IS NOT due to Solver’s inability to find an optimal or feasible
solution.
I do not offer user spreadsheet support, but if I can replicate and correct
the error, I will upload a corrected copy of the spreadsheet to the Trader Edge
download page and I will update the file origination date on the same page.
Please check the download page periodically for the latest versions of the
spreadsheets.
I hope you enjoy these tools and find them useful in your option trading
and research.