1

Option Trading

Euan Sinclair
 2

Overview
Session One: Introduction to options, pricing and greeks.
Session Two: General trading principles. Volatility
measurement and forecasting.
Session Three: The variance premium. Hedging. Expiration
trading.
Session Four: Risk management. Some examples of trades.
Trade evaluation.

v10.1.1 2
 3

Option Trading
Session Three: Volatility Trading and
Hedging

v10.1.1
 4

Session Three Overview

• The variance premium.
– Indices
– Equities
– Bonds
– Volatility indices.
• Why does it exist?
• When can I expect it to exist?
• Strategy selection.
• Hedging.
• Trading expiration and very short-term options.

v10.1.1 4
 5

Volatility Trading
• Recall from Session One that when we derived BSM we
actually priced the replication value of the option.
• This depends on realized volatility.
• But option price in the market depends on implied volatility.
• So if realized volatility is not equal to implied volatility we
can trade the option, replicate it in the underlying and
profit.

v10.1.1 5
 6

Volatility Trading

• Theoretical profit:

𝑃𝐿 = 𝑉𝑒𝑔𝑎 𝜎𝐼 − 𝜎𝑅

• But only on average…

v10.1.1 6
 7

Volatility Premium
• Implied volatility > subsequent realized volatility.
• S&P shown below.
Volatility Premium
45.0000
40.0000
35.0000
30.0000
25.0000
20.0000
15.0000
10.0000
5.0000
0.0000

• Average premium=3.6 volatility points.

• Median premium=4.0 volatility points.
v10.1.1 7
 8

Volatility Premium
• Persistent even at very low volatility levels.
• 2017 shown below.
VP for 2017
18
16
14
12
10
8
6
4
2
0

• Average premium=4.4 volatility points.

• (about 40% of the implied volatility level).
v10.1.1 8
 9

Volatility Premium
• As a percentage, the premium is generally higher at low
volatility levels.
• Since 1990:
VIX Level Volatility Premium (as %)
<15 0.27
15<20 0.26
20<25 0.20
25<30 0.20
30<35 0.18
35<40 0.14
>45 0.10

v10.1.1 9
 10

Volatility Premium

• Selling 10 delta QQQ, 2nd month strangles

v10.1.1 10
 11

Volatility Premium

• Selling ATM QQQ, 2nd month straddles.

• (Same vega exposure as with strangles).
v10.1.1 11
 12

Why Does the Premium Exist? - Insurance Premium

• Both calls and puts provide insurance.

– Puts against a crash hurting an existing portfolio.
– Calls against FOMO (Fear Of Missing Out).
• All insurers charge a premium for their products so they
can make a profit.
• This is no different from any other shop.

v10.1.1 12
 13

Why Does the Premium Exist? - Fear of Atypical Events

• People vastly over-estimate the danger of extreme events.

• Terrorism kills 180 Americans a year (about 8 if we exclude
9/11).
• Heart disease kills 600,000 Americans each year.
• This mistake means options are overpriced.
• “Black Swans” don’t happen very often.

v10.1.1 13
 14

Why Does the Premium Exist? Correlation Risk

• When bad things happen, they tend to happen to an

investors entire portfolio.
• Diversification isn’t terribly effective in bad states.
• Options are the best performers and people pay a
premium to have a winner in bad times.

v10.1.1 14
 15

Why Does the Premium Exist? Unpleasant Risk Profile

• A short option portfolio has negative skewness, high

kurtosis, high downside deviation and large drawdowns.
• None of these are popular.
• A bad risk manager can easily blow up being short options.
• One short volatility fund lost 89% over two days in
February.

v10.1.1 15
 16

Why Does the Premium Exist? Skewness Premium

• Much of the short volatility premium comes from the

short puts.
• All of the reasons above are consistent with this
observation.
• Should also be obvious because the implied volatility of
puts is so much higher than calls.

v10.1.1 16
 17

Skewness Premium

• Selling second month 30 delta QQQ risk reversals

(hedged).
v10.1.1 17
 18

Volatility Premium in Equities

• On average, equities also display a volatility premium.
• About half that of indices.
• Some stocks have no premium or a negative premium.
• Highest premium in small cap stocks and value stocks.
• (Also these have the least liquid options).

v10.1.1 18
 19

Volatility Premium in Commodities

• Commodities with statistically significant volatility premia.
– Crude oil.
– Heating oil.
– Natural gas.
– Corn.
– Sugar.
– Copper.
– Cocoa.
– Oats.
v10.1.1 19
 20

Volatility Premium in Commodities

• Why copper and not gold?
• Why corn and not beans?
• Be very careful with commodities.
• “Equities trade on statistics. Commodities trade on
knowledge.”

v10.1.1 20
 21

Volatility Premium in Bonds

• About as persistent and as large (in percentage terms) as
in indices.
• But bond volatility is much lower (typically 3% to 6% as
opposed to 10% to 30%).
• So bond options have more gamma and positions can get
out of control quickly.

v10.1.1 21
 22

Volatility Premium in the VIX

• Probably exists.
• Evidence is weaker as we have a shorter time period to
examine and only one liquid product.

v10.1.1 22
 23

Volatility Premium in General

• Look for products with a significant and consistent implied
skew.
• Look for products with a significant and consistent term-
structure where front volatility is below back volatility.

v10.1.1 23
 24

Harvesting the Volatility Premium – Selecting an Expiration

• Trader’s heuristic: “Vega wounds. Gamma kills”.

• Volatility reverts, so it is reasonable to expect that an
adverse move that causes Vega losses will reverse.
• So Vega losses can be ridden out (sometimes this might
not be recommended).

v10.1.1 24
 25

Aside: Stops
• Standard advice: Get out of a trade when it loses a
certain amount.
• “Losers add to losers”.
• BAD ADVICE!
• If you exit based only on losses, you traded too big.
• Exit when you are wrong.
• Sometimes this correlates with losses, sometimes not.

v10.1.1 25
 26

Effect of Stops
• Naive guess: stops truncate the distribution of results.
• Trade with an average return of 10%.

v10.1.1 26
 27

Effect of Stops
• But 11% of trades will lose more than 15%.
• So we put in a stop at a loss of 15%.

v10.1.1 27
 28

Effect of Stops
• Not even close.
• Instead, many trades that would have been small losers or
winners get stopped out (the probabilities still need to add
to one).

v10.1.1 28
 29

Effect of Stops
• Stops don’t just stop losers.
• They also stop some winners.
• The net result is that they reduce returns.
• No free lunch.
• “Dumb traders think smart traders always use stops”.

v10.1.1 29
 30

Harvesting the Volatility Premium – Selecting an Expiration

• But gamma losses will be locked in by the process of delta

hedging.
• Short option positions will require selling low and buying
high to hedge.
• So it seems plausible that much of the premium is
concentrated in the short term options that have the most
gamma.

v10.1.1 30
 31

Harvesting the Volatility Premium – Selecting Strikes

• I’m not aware of any rigorous theoretical work on this.

• My thinking:
– The ATM volatility is closest to fair (but still overpriced).
– Teeny puts have most overpriced volatility, but in dollar terms the
very low strikes have little premium (or vega). So you need to sell
too many.
– Sell the strike with the highest excess dollar premium.
– Simulations tend to support this.

v10.1.1 31
 32

Harvesting the Volatility Premium – Selecting Strikes: Example

• One month options: Stock=$100, ATM volatility=30%.

Strike Implied Price Price at 30% Premium

Volatility
80 45% 0.2 0.01 0.19

85 41% 0.41 0.09 0.32

90 37% 0.84 0.44 0.40

95 33% 1.70 1.42 0.28

v10.1.1 32
 33

Hedging

• In order to convert an option into a volatility trade, we

need to maintain delta neutrality.
• In theory, we hedge continuously.
• In practice that leads to infinite transaction costs.

v10.1.1 33
 34

Hedging: First Idea

• Whenever the option is in the money, hedge it as a 100

delta option.
• Example: Stock is $100 and we sell a 110 call.
• Don’t hedge.
• When stock goes above $110, buy a share.
• S<X: keep entire premium, C.
• S>X: Option P/L=C-(S-X)
: Hedge P/L = S-X
: Total P/L = C

v10.1.1 34
 35

Hedging: First Idea

• Sadly, this also leads to infinite transaction costs.

• And a very volatile P/L.
• Simulation: 30 day call, implied and realized volatility both
0.2, rates zero.
• Call value is 2.75.
• Theoretical PL=0 - costs

v10.1.1 35
 36

Hedging: First Idea

• Assume zero transaction costs

• Average PL=0
• Median PL= -0.8
PL
0.16
0.14
0.12
0.1
Frequency

0.08
0.06
0.04
0.02
0
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10
PL

v10.1.1 36
 37

Hedging: First Idea

• Assume 1 as a cost whenever we hedge.

• Average PL=-3.5
• Median PL= -4.0
• Obviously not infinite as the option has a finite life, but if
we observe each hour the average loss is 14.
• In the limit it is infinite.

v10.1.1 37
 38

Hedging: Heuristics

• Need something more subtle and closer to continuous.

v10.1.1 38
 39

When to Hedge?

• Continuous hedging leads to infinite transaction costs.

• Hedge at a given time.
• Hedge at a given price move, either in terms of $ or
standard deviations.
• Hedge to a set delta band.
• All are sub-optimal in terms of cost relative to risk
reduction.

v10.1.1 39
 40

Utility Based Methods

• Utility is the concept of balancing risk and reward.

• How much are we prepared to pay to reduce risk?
• Since BSM really prices replication, we can do same thing
with costs.
• At some point, you will be indifferent to risk and costs of
removing it.
• But now personal preferences matter.

v10.1.1 40
 41

Utility Based Methods

• Redo the BSM model but allow for a spread in the

underlying.
• Hedge to the edge of a band that is gamma dependent.
• Hedge short gamma more aggressively.
• (play aggressive defense and let profits from long gamma
run).
• High costs=> wider band.
• High risk aversion=> narrow band.

v10.1.1 41
 42

Hedging a Long Option

v10.1.1 42
 43

Hedging a Short Option

v10.1.1 43
 44

Utility Based Methods

• The modified delta bands don’t necessarily include the

BSM delta, particularly for far OTM options.
• This is because a long option position means we buy at
the offer and sell at the bid, so it “sees” a slightly lower
volatility.
• This means a long option has a slightly lower delta (due to
delta dependence on volatility).

v10.1.1 44
 45

Wilmott-Whalley Approximation

• This theory results in having to solve a nasty PDE.

• Can’t be done in real time (or even close).
• But, using an asymptotic expansion, we can come up with
an easily used approximation.

v10.1.1 45
 46

Wilmott-Whalley Approximation
1
𝜕𝑉 3 𝑒𝑥𝑝 −𝑟 𝑇 − 𝑡 𝜆𝑆Γ 2 3
∆= ±
𝜕𝑆 2 𝜅

• Lambda is proportional transaction cost.

• Kappa is risk aversion.
• So choose a risk aversion parameter for a position you
understand, then use this for all positions.
• This can save about 10% of costs over ad hoc methods.

v10.1.1 46
 47

Wilmott-Whalley Approximation

v10.1.1 47
 48

Hedged Position Results

• Volatility edge means you win on average.
• Any particular trade is hugely variable, even if your vol.
forecast is right.
• I.E. implied is 20%, your forecast is 15% so you sell
options and can (easily) still lose.

v10.1.1 48
 49

Path Dependency
• Discrete hedging creates path dependence.
• Simple example: a big move the day of expiration (when an
option has a lot of gamma) will give a different PL than the
same move a year out.
• In each case the volatility is the same but the result isn’t.
• Dispersion is roughly inversely proportional to square root
of number of hedges.

v10.1.1 49
 50

Hedging Frequency
• Dispersion is roughly inversely proportional to square root
of number of hedges.
• EG $1000 vega of one year options traded with no
volatility edge.

v10.1.1 50
 51

Hedging Frequency
• So why not hedge more often?
• Again, costs.

v10.1.1 51
 52

Drift VS No Drift
• When the underlying is drifting and we are long gamma,
our hedges are going to be losers. We will be selling into
a rising market. So we want to hedge less often.
• If we use a higher volatility, we see a lower gamma so we
hedge less.
• Traders refer to this hedging trick as letting their deltas
run if they are long in a trending market,
• or hedging defensively if they are short gamma in a
trending market.

v10.1.1 52
 53

Drift VS No Drift

Position Market Volatility Bias for

Hedging
Short Gamma Trending Low

Short Gamma Range Bound High

Long Gamma Trending High

Long Gamma Range Bound Low

v10.1.1 53
 54

Choosing a Hedging Volatility

• The volatility you use will determine delta and gamma and
hence when and how you hedge.
• You are free to choose whatever you want…
• Implied vol. => low MTM variance but an uncertain final
number.
• (True) Realized vol. => higher MTM variance but a
guaranteed profit (if your forecast is correct).

v10.1.1 54
 55

Hedging at Implied

v10.1.1 55
 56

Hedging at Realized

v10.1.1 56
 57

Expiration Trading
• Very close (hours) to expiration, options lose most
optionality.
• If ITM behave as underlying.
• If OTM behave as worthless.
• The tricky part is what happens at the strike.
• Note that pricing models don’t “break” at expiration. They
are still a helpful guide.

v10.1.1 57
 58

Expiration Trading: Pinning

• The underlying will settle near a strike much more than
randomness would imply.
• This is due to market makers hedging long gamma, as a result
shorts don’t need to hedge.
• Sell ATM options close to expiration, ideally those with high
open interest.

v10.1.1 58
 59

Expiration Trading: Pin Risk

• If the market settles at a strike, it can be very difficult to
predict which options will be exercised.
• Be aware that slightly OTM options can be exercised.
• Sometimes done to squeeze the underlying or to avoid
slippage in liquidating a large underlying position.

v10.1.1 59
 60

Expiration Trading: Cash Settlement

• Cash settled options will be hedged with futures.
• At expiration, the futures position will still be there.
• Need to allow for unwinding.

v10.1.1 60
 61

Expiration Trading: Greeks

• Greeks are still correct, but may be misleading.
• For example, a one-hour straddle will often have theta 20
times higher than its value.
• Also, gamma is actually infinite right at the strike.
• Sometimes useful to set implied volatility to zero to avoid
possible confusion.

v10.1.1 61
 62

Session Four Preview

• A real trade.
– Selection.
– Implementation.
• Trade evaluation: in general and in this specific case.

v10.1.1 62