Maximum Adverse Excursion

IMI AXIMUM
IAI DVERSE
B-l XCURSION
ANALYZING PRICE FWCTUATIONS
FOR TRADING MANAGEMENT
JOHN SWEENEY
JOHN WILEY & SONS, INC.

New York
Chichester
Weinheim
Toronto
Singapore
Brisbane
This text is printed an acid-free paper.

Copyright 0 1997 by John Sweeney.
Published by John Wiley 8z Sons, Inc.
All rights reserved. Published simultaneously
in Canada.
Reproduction or translation of any part of this work beyond

that permitted hy Section 101 ur 108 of the 1976 United
States Copyright Act without the permission of the copyright
owner is unlawful. Requests for permission or further
information should be addressed to the Permissions Department,
John Wiley & Sons, Inc.
This publication is designed to provide accurate and authoritative
information in regard to the wbject matter covered. It is sold
with the understanding that the publisher is not engaged in
rendering legal, accounting, or other professional services. If
legal advice or other expert assistance is required, the services
of a competent professional person should be sought.
Library of Congress Cotalopinp-in-publication
Data:
Sweeney, John.
Maximum adverse excursion : analyzing price fluctuations far

trading management / John Sweeney.
em. ~ (Wiley traders advantage series)
P.
Includes index.
ISBN O-471-14152-6 (cloth : alk. paper,
1. Investment analysis-Mathematical
modela. 2. SccuritiesPrices-Mathematical models. 3. Risk manilgement-Mathematical
models. I. Title. Il. Series.
IIG4529.S93 1 9 9 6
332.640151-dc2
96.29431
Printed in the United States of America
10
THE TRADERS ADVANTAGE

SERIES PREFACE
The Traders Advantage Series is a new concept in publishing for

traders and analysts of futures, options, equity, and generally all
world economic markets. Books in the series present single ideas witb
only that background information needed to understand the content.
No long introductions, no definitions of the futures contract, clearing
house, and order entry: Focused.
The futures and options industry is no longer in its infancy. From
its role as an agricultural vehicle it has become the alter ego of the
most active world markets. The use of EFPs (exchange for physicals)
in currency markets makes the selection of physical or futures markets transparent, in the same way the futures markets evolved into
the official pricing vehicle for world grain. With a singe telephone call,
a trader or investment manager can hedge a stock portfolio, set a
crossrate, perform a swap, or buy the protection of an inflation index.
The classic regimes can no longer be clearly separated.
And this is just the beginning. Automated exchanges are penetrating traditional open outcry markets. Even now, from the time
the transaction is completed in the pit, everything else is electronic.
vi
THE TRADERS ADVANTAGE SERIES PREFACE
Program trading is the automated response to the analysis of 8. computerized ticker tape, and it is just the tip of the inevitable evolutionary process. Soon the executions will be computerized and then we
wont be able to call anyone to complain about a fill. Perhaps we wont
even have to place an order to get a fill.
Market literature has also evolved. Many of the books written on
trading are introductory. Even those intended for more advanced audiences often include a review of contract specifications and market
mechanics. There are very few books specifically targeted for the experienced and professional traders and analysts. The Traders Advantage Series changes all that.
This series presents contributions by established professionals
and exceptional research analysts. The authors highly specialized talents have been applied primarily to futures, cash, and equity markets
but are often generally applicable to price forecasting. Topics in the series include trading systems and individual techniques, but all are a
necessary part of the development process that is intrinsic to improving price forecasting and trading.
These works are creative, often state-of-the-art. They offer new
techniques, in-depth analysis of current trading methods, or innovative and enlightening ways of looking at still unsolved problems. The
ideas are explained in a clear, straightforward manner with frequent
examples and illustrations. Because they do not contain unnecessary
background material they are short and to the point. They require
careful reading, study, and consideration. In exchange, they contribute knowledge to help build an unparalleled understanding of all
areas of market analysis and forecasting.
Unless you are gifted with remarkable natural insight, then
trading successfully is the result of a great deal of work. For most of
those who can boast a history of profits, there were countless
days and long hours studying the way markets react to various government reports, its changes in volatility, the speed at which it
moves, its relationship to other markets, and periods of illiquidity
during the day. From an unlimited variety of patterns, successful traders are able to find some profitable sequence that could be
anticipated.
Perhaps the most important part of the insight achieved from
this effort is the understanding of risk. For every position in the mar-
vii
ket, there is a risk. Although price volatility is often used to express

risk, each trading style or systematic approach to the market has its
own specific pattern of exposure. A long-term trend following system,
that can hold a long position for months, has the potential for a much
larger profit or loss on that one trade than a day-trading method which
is in and out within six hours. The sequence of profits and losses, however, form a very different equity picture than individual trades.
When you define the rules for buying and selling, you must also understand the adverse price moves that are an integral part of that
trading method; you must study those risks and manage them. There
is no longevity without risk control.
John Swe&ey tactfully calls those large, distressing price moves
that we have seen in the past, and expect to see again, the maximum
adverse excursion. These equity drawdowns are the difference between
expected and actual performance-if the actual is much larger than
expected, you are not able to stay with your plan, no matter how sound
the theory seems. In a very realistic and practical evaluation of equity swings, Mr. Sweeney shows how a simple and thorough assessment of adverse price moves can yield very specific rules for risk
management and greatly improved results.
Every trading method has its own unique pattern of adverse
moves. These patterns, when studied in a systematic way, will show
the amount of capital needed to trade successfully, and even the specific stop-loss, or maximum risk, that should be taken for any one
trade. Once you see how it is done for one series of trades, it is easy to
apply to your own performance. Mr. Sweeney takes these patterns further and shows how betting methods, that is, changing the size of the
investment after profits and losses, can be designed to alter the returns and risk of a trading program.
While the author cannot possibly cover all the different trading methods, he gives the readers the tools and the understanding to
continue themselves. To help further, he has separated the development and teaching of the analytic methods, from the spreadsheet
code and numerous clear examples in the Appendixes. In this way he
has made the Appendixes into a useful workbook and an important
review.
You will find that the true value of John Sweeneys effort is in
the way he methodically organizes, displays, and evaluates the adverse
. ..
WI,
equity moves, as well as the favorable, profitable ones. It is this

straightforward, understandable approach that is most rewarding for
a reader. As he proceeds through the steps, we can see the conclusions
unfold. By the end of the book, you will be able to say that there is a
great deal to be gained by careful review and assessment of the adverm equity swings of your own trading approach.
PERRY J. KAUFMAN
Wells Riuer, Vermont
PREFACE
This is a book about losses, but dont let your eyes glaze over. Riskloss-is tied up with your fear, your profitability, and your career. For
the most part, this book is addressed to people with Investment Committees and Lines of Authority to deal with, people who end up explaining themselves if things go badly.
For those people, this book presents a novel approach to assessing the extent of risk and minimizing losses. In the world of futures,
the game is approximately zero-sum: for every winner theres a loser.
Stock traders have it easier because their pie is (currently) expanding
and, in many situations, everyone can win (or all go down together).
But, in a futures game (keep the word game in mind), things are
tighter. If I win, you lose. If you figure in commissions and slippage,
we both come up shorter than wed like.
From such a straightened trading environment, and from mathematical theory, we know going in to the competition that the key to
winning is minimizing the size of our largest loss. The problem, as
with all trading maxims, is: whats large? How can this be quantified? Thats where this book becomes novel.
I suggest you look at the results of your trading approach, consistently applied, to quantify things. If your traders eye is focused on
ix
PREFACE
the bark or the branches of the tree, you may have missed the forests
moving, or at least swaying in the breeze of the business cycle. Unless youve kept a detailed diary and read it, youve probably not noticed any particular consistency in the after-entry behavior of the
market. Each situation appears specific. Thats why you get paid the
big bucks: to be on top of the situation and trade smart.
Heres a thought for you, though: Shouldnt there be a consistent
pattern to what happens when you take action? What if, instead of gut
feelings, you could know objectively when to cut off a loser? What if,
instead of gut feelings, you could know when to put in a protective
stop? What if you could know objectively when to take profits? Well, if
all that were possible, all you would have to do is execute your scheme
properly.
From a management point of view, what if there were a way of assessing whether your traders were smart or lucky? What if there were
a way of consistently winning trading profits other than putting the
best and the brightest traders on the firing line until they burned out?
What if you could put up an objective performance benchmark, quantify
the amount of capital a trader needed (for his style and approach), and
assess the inevitable losses as normal or abnormal?
Now there is such a method, but it requires work: poking
into market behavior and using tools that arent easy to manipulate. This book looks at what happens when we make decisions on a
consistent basis-what happens to our trading positions as a result
of the markets behavior? It asks, If we consistently do this, what
does the market do? (This is different from asking, If we consistently do this, do we make a profit?) When this question is asked
today, the answer is I dont know or Its random or, worse, I
know its going to be a winner (since no one consciously takes
losers).
The answer closest to correct is I dont know. Think about it: if
your markets truly random, nobody should be trading. In fact, trading would be impossible because the next price would rarely be close
to the previous price-it might be anywhere.
Day to day, some speculative trades win, some lose, scnne go
nowhere. Were interested in the winners and we want to find and
eliminate the losers as soon as possible so as to keep our losses small.
Actually, if we know what we are doing, well find that winning and
PREFACE
xi
losing trades look different and the way they look can be the key to
success.
You might ask What do you mean how does a trade look? I
use the word look specifically to mean appear because Im going to
show you pictures of that behavior, pictures that are graphs which are
collections of data. Youre going to see the markets behavior in descriptive statistical terms rather than in price charts.
On these charts, youre going to see a line, the elusive edge that
traders seek and it will be a line all your own, from which profits can
flow with minimized, quantifiable risk.
Do the markets exhibit a consistent behavior? Can we possibly
adduce a consistency in what we see and experience? If so, what might
it be? Well, open your mind and read on.
It occurred to me as I prepared my manuscript, that this entire
book is really an exercise in exploratory statistics for traders. All the
things were used to seeing-charts, lines, indicators-have been replaced with graphs and tables; alien displays we traders dont use
much. The book is organized to show you the nuts and bolts of excursion analysis without delving into the theory.
I use examples, rather than generalities. Ive used spreadsheets
because most computerized traders have one available. There is more
elaborate software available for exploratory data analysis, but youd
probably need a degree in statistics plus considerable facility in programming to use it. (See the program Matlab from The Mathworks,
Inc., 24 Prime Park Way, Natick, MA 01760-1500. Phone 508-647-7000.)
Luckily, the data were examining is sufficiently sparse and simple that bonehead stat works fine. Still, it would be a great benefit for
you, if youre not numerically inclined, to pick up not only the technique of measuring excursion but also that of visualizing happenstance graphically. Within straightforward limitations that youll see
here, the data displays can be applied to many aspects of your trading
experience.
JOHN SWEENEY
CONTENTS
THE
CHAPTERS
IDEA
Experience
The Rule
The Data
CHARTERS
DEFININGMA~ADVERSEEXCURSION
Adverse Excursion, Favorable Excursion
10
Tweaking Stops
15
Sample Calculations
16
CHAPTER
21
DISPLAYINGMAF.
Aggregation
21
Frequency Diagrams
25
CHAPTERS
33
DEFININGPROFITBYBIN
Profit Tradeoffs
33
Profit Curves
34
Interpretation
37
CHAPTERS
IMPACTOFVOLATILITY
CHANGES
41
Tweaks
41
Summary
62
. ..
XIII
CONTENTS
xiv
CHAPTERS
Capital
RUNSEFFECTS
Conservation
63
63
Impact on a Particular Trading Tactic
64
Impact on Campaign Trading
74
Impact on Betting Strategies
83
Summary
84
CHAPTERS
WTINGALES
86
Simple Martingale
87
Complex Martingale
88
Martingales on Crude Oil
93
CHAPTERS
TFUDINGMANAGEMENT
Portfolio Impacts
99
99
Day-to-Day Trading
100
Elaborations
104
Conclusion
106
APPENDIX A COMPUTINGMAE
A~PENDIXB COMPUTING MAxFE
A~PENDIXC COMPUTING MINFE
APPENDMD GENERATINGAFREQUENCY
DISTRIBUTION
APPENDIX E MAE FOR SHORTS AND LONGS
A~PENDI~F COMPUTINGPROFIT CURVES
A~PEND~G RANGEANDVOLATILITY
APPENDS H RANGEEXCURSION
A~PENDIXI
MARTINGALES
A~PENDIXJ A~PL~NGMARTINGALESTO
TRADINGCAMPAIGNS
107
INDEX
157
112
116
119
123
127
133
138
145
151
1
THE IDEA
Imagine that you are a prehistoric hunter at an African waterhole.

Its dusk, dusty but cooling, the wind out of the north so that the animals, thousands of them, are easing up from the south along a wide,
gentle-sloped valley, seeking water. You and your mates are hidden in
the bushes to the south as well, but to the sides of the zebras advance.
Youre hungry and so is everyone at home. Roots, insects, a few
fruits, and berries arent cutting it this late in the summer. Youve
got to hunt and hunt successfully.
Hunting, however, is tough. Youre short of spears and arrows.
In addition, somebody usually gets hurt mixing it up with the animals; you use a lot of scarce energy; sometimes youre being hunted
while youre hunting; and half the time, you come up with nothing except exhaustion because you dont have a theory of zebras.
You dont know anything about zebra physiology, psychology, seasons or, for all you know thoughts, let alone their gods. You never know
exactly what the animals are going to do. Spooked from the north,
youve seen them wheel and go south along the valley or sweep in a
broad arc southward up the sides of the valley and then down (once
they even went over the crest!), or even stampede right over the top of
the hunters.
1
THE IDEA
Even so, youve got to hunt. Youre betting today that if the zebras are spooked simultaneously from the west, northwest, northeast, and east, theyll go straight back down the valley. Next, if
experience holds, theyll slow down after a few hundred yards if not
pursued and their tight running herd will spread out, right about
where the rest of your band has moved in and set up to spear a straggler from all sides.
Even if the hunt goes as planned, keeping the carcase out of the
mouths of lions, or hyenas and getting it back home will be tough work
in the dark. Still, you must hunt.
As a trader, your situation is a lot like the hunters, Whether you
have a team or are solo, you could use a theory of the market. Youve
probably got some ideas about what the zebra herd (the market) is
going to do. You know the season is dry and which way the wind is
blowing. You know generally where the herd heads when they break,
you know how far they like to run when they stampede, and you know
they will spread out when their fright dies down.
Youd like to make some money out of what you know. Are there
better ideas out there? Trading is an oral tradition, surprising in the
amount of money risked on fairly light formal credentials. As you
learn, you get lots of profundities (Keep your losses small, Dont
overtrade, etc.), lots of people with ideas, and books like this one but
a theory is something different.
In the scientific method, a theory is the result of observation
which leads to a provisional hypothesis of cause and effect, a hypothesis susceptible to testing. Testing will, with proper design, lead to
confirmation or rejection of the hypothesis. After confirmation, further hypothesizing and testing continues; after rejection, the hypothesis is reworked and retested. Finally, a theory can be formulated.
Only in relatively modern times have such processes been applied
to market behavior, that is, the behavior of groups in open, unre-
* For a look at markets in Laboratories, see work by Vernon Smith and his colleagues at the Economic Science Laboratory for Research and Education, University
of Arizona, McClelland Hall 116, Tucson, AZ, 85721. See also the new field of behavioral finance at web site http://www.sas.upenn.edu/-rrattgen/finps~.ht~l. At
this writing, only bits and pieces of research have popped up to indicate anomalies
in classical market theory. Far example, betas explanatory power for returns is
EXPERIENCE
stricted markets. As a result, our knowledge of how markets will act

or react is abysmal. We are in the first stage of the scientific method:
simple observation.
EXPERIENCE
Traders pick up experience while observing the market, but true experience comes from trading. Some keep notes mentally, some keep a
journal, some even keep a database.* A running discussion among
traders, economists, analysts, and the entire world also goes on, the result of which is the traders view: his outlook for the economy, his mar,ket and his tradable. Ideally, people record their views, their trades,
and their results. Mistakes and ~uccesses~ are recorded and, over
time, something is learned. Realistically though this work of recording is rarely done. Instead, there is an accretion of experience in a
traders head and a steady winnowing of losing traders.
Ditch that, I say, for statistically recorded results. Define your
trading rules objectively and see whether they yield results that can
help you define your actions operationally in the future. In other
words, does the market act, after your decision, consistently or not?
Since no one wins every trade, this is tough to tell. Still, it turns out
that, in at least one respect, a good set of trading rules generates a
classic set of responses by the market just as spooking a zebra herd
at a waterhole does: You can know from the markets behavior (like
the herds) roughly whats going to happen.
If the herd, instead of fleeing, runs right at you, your hopes are
dashed and you scramble out of the way. Instead of pursuing, you are
routed with, hopefully, the smallest possible injury. Its a question of
judgment. While youre in the act of spooking the herd, youre exposed.
At any instant, they can decide to flee or come at you. Youre dancing
on the traders edge, trying to decide if you should continue advancing
and yelling-or flee for your life.
questioned now, over reaction by market participants is acknowledged and rcsearchem are starting to attribute returns to market cap and market size or share.
* Chande, Tushar, $ecur-e (Chande Research and Trading, Pittsburgh, 19961. This
software tracks a traders actions and provides not only a trading .journal but a
checklist of factors to enter in the journal.
THE IDEA
So, too, with trading. The judgment comes in when you must continue the trade or get out. The market is moving around in front of
you and, like the zebra herd, is much bigger than you are. You must
judge when its decided to move favorably away from your entry or
right back over the top of it.
On the floor you can see the orders coming to the pit, hear the
noise, see the players screaming. Off the floor, you have the tape, your
own order flow, your phone, and the chart-information passed by the
recorder or the exchanges reporting system. Either way, youre laoking to see how close the herd is coming. Thats what youre tracking
and it turns out its a good indicator: past a certain point, they are
probably going to run over you; before that point, they are more likely
to flee properly.
THE RULE
Generally, good trades dont go too far against you while bad ones do.
Sometimes a winning trade could go strongly against you before turning right, but what generally happens? Whats usually the case?
It turns out that if your trading rules are consistent and can distinguish between good and bad trades, then, over many experiences,
you can measure how far good trades go bad and, usually, see at what
point a trade is mre likely to end badly than profitably. That is the
point at which you stop and/or reverse.
In this book, we will measure the price excursion from the point
of entry. Measuring things abstractly from the point of entry gets
away from the old news in the charts: support and resistance, value
points. It gives us a point of departure in a constantly changing sea.
In speculative trading, we only have our entry point and our exit
points, so this is a valid point of reference. We arent trading off a customers hedge and we dont see the order flow or the issue calendar or
the inventory. All these points of value arent relevant to the technical speculator anyway; he or she really nly has his price-take it or
leave it. Moreover, thats the point from which were judged. We may
as well focus on it.
In zebra terms, were going to see how close the herd cmes to us
before they shear off and head the other way-or decide to keep coming.
THE DATA
Well tweak this analysis with some fine points later in the book and
the general subject of using the technique in campaigning is dealt
with in Campaign Trading!, an earlier book, but here well make sure
the nuts and bolts of determining the breaking point of a trade are
covered completely.
THE DATA
One other basic point needs to be covered before we start. The data
for this exercise was developed for Campaign Trading! in mid-1995. It
includes Crude contracts from October 1983 through October 1994,
about eleven years of trading. The details of choosing and assembling
the data were covered there. This process is unique to futures trading
and equity or debt investors with long-lived tradeables can ignore the
issue of continuity.
To provide long continuous charts, the most active contract data
each month was put together with those before and after it in a data
series such that the interday price changes while jumping from one
contract month to the next were consistent. This process created the
actual price changes one would have experienced in rolling from one
contract to the next, but the values you may see here and there for
Crude probably wont be close to the actual values published. The results, shown in the charts in Campaign Trading!, are realistic chart
relationships and accurate day-to-day price changes.
I use daily data in my trading. I havent experimented with intraday data though I have used the concepts in this book with weekly
data.
2
DEFINING Mb
ADVERSE E XCURSION
Try to think of future prices from the vantage point of todays prices.
Imagine you are standing at a point looking forward toward a shifting
gray cloud of varying density, each miniscule dot representing a possible price occurrence. There are points of greater density and other
areas of near brightness. Looking directly forward, the mass is generally darkest but the cloud of possibilities shifts constantly as new information and new emotions enter the market and its participants.
There are areas of concentration and others of relative improbability.
Were interested in the edges of the cloud. If we translated the
haze of possibilities into tomorrows price bar, the edges would translate into the high and the low of the day, the points at which our stops
or limit orders would be last hit. We want to see if the shifts of the
haze are likely to hit our stops if we set them here or here or here.
The shifting-the movement of the price possibilities-is described statistically as a change from an expected value, an excursion
away from the darkest mars in front of us toward the outer edges of
likelihood.
7
IMPACT ON CAMPAIGN TRADING
81
Figure 6-l 1 Combining Trading Combinations. Melding the results of the figures
above for Drutschemark, Swiss Franc, Yen, and Cold generates this jagged curve.
Though the direction is upward, drawdowns seem to be serious. For a better look
at those, see Figure 6-l 2.
Figure 6-12 Drawdowns from Four Combinations. Somewhat easier to see are
the drawdowns, presented here as a percentage reduction from peak equity plus
initial trading capital.
DEFINING MAX ADVERSE EXCURSION
Figure
2-1
Standard Price Chart.
Time
Figure 2-2
Raw Excursion. Starting from the point of entry. price excursion is
measured as the gain or loss on the trade, not the price and exclusive of transactions costs.
Figure 2-3
Consistent Excursions. The ideal result
entry that behave consistently.
is a set of excursions from
10
Figure 2-4 Losers. In comparison to the upward trend of the winners, losers for
a given set of rules usually have a maximum upside, a shorter life span than winners, and a sharp terminating downfall.
Excursion, then, is just the change in price from our point of

entry, measured every bar. It could be weekly, daily, hourly, or on the
minute. Our interest is in whether there is some regularity in the excursion from entry, whether the position is long or short. If there is
some pattern, some regularity, we hope to exploit it while were in the
trade by discerning whether things are going properly or badly and, in
either case, what likely events are next.
ADVERSE
EXCURSION,
FAVORABLE
EXCURSION
When prices move against your trade, that is aduersity. From this comes
the term adverse excursion which is used to describe that price movement which goes against our favor during a trade. The abbreviation
used throughout the book is MAE-maximum adverse excursionwhich is an acronym for the worst that it gets while in a particular
position.
A key assumption for defining adverse or favorable price movement is the time frame. After all, if you wait long enough after an entry,
just about any price might pop up. To avoid this, your trading rules
must specify not only an entry, but also an exit. By doing this, you also
define a time horizon in which you can analyze price movement.
Conversely, there is favorable price movement. Maximum favorable excursion (MaxFE) and minimum favorable excursion (MinFE)
are discussed next.
ADVERSE EXCURSION, FAVORABLE EXCURSION
11
MawFE a n d MinFE
Adverse excursion is the greater of zero or the difference between your
entry price and the worst price experienced after entry but before the
trade is closed. If youre long:
M*%ng = MAXLO, (Entry price - Lowest subsequent low),
Previous value1
MAX here refers to the greater of zero or the absolute value of the
computed difference. If youre short, its:
MALTt = MAX[O, (Highest subsequent high - Entry price),
Previous valuel
Remember that zero is greater than a negative number. As for nmximum favorable excursion, if youre long, its:
MaxFELog
= MAX[O, (Highest subsequent high-Entry price),

Previous value1
and, for shorts:

MaxFEs,ort
= MAX[O, (Entry price - Lowest subsequent low),

Previous value1
For minimum favorable excursion, if youre long, calculate:

MinFELO# = Max[O, Highest subsequent low Entry price),
Previous value1
and, if youre short, calculate:
MinFEs,_, = MAX[O, (Entry price - Lowest subsequent high),
Previous duel
Examples of Maximum Adverse Excursion are shown in Figure
2-5. Maximum Favorable Excursion is illustrated in Figure 2-6 and
Minimum Favorable Excursion is illustrated in Figure 2-7.
12
MAE
MAE
Figure 2-5
(Top) Long MAE Example. Price rises from the entry on the opening
but just one day is far lower than the others between entry and exit. The low on
that day will be used for calculating the adverse excursion on the trade.
(Bottom) Short MAE Example. A short conw early in this example and suffers
through 35 points of adverse price movement before vindication. The absolute
value of the difference between the entry price and the MAE price will be the MAE
value for this trade.
ADVERSE EXCURSION, FAVORABLE EXCURSION
.-.-
._._.
-.-
. ..__..
.,-.-. -
.-.-.
13
-._17
Figure 2-6
(Top) Favorable Excursion When Long. Mid-Julys price just tops early
Mays during this extended mid-l 991 trade in NY Light Crude. The absolute value
of the difference between the MaxFE price and the entry will be the MaxFE.
(Bottom) Short MaxFE Example. Another early short gets as far as $48 per share before rebounding into the $505. The difference between the $52 entry price and the
$48 low will be the Maximum Favorable Excursion of $4,
14
Figure 2-7
(Top) Long MinFE Example. As the NASDAQ composite rises steadily

in early 1995, the Minimum Favorable Excursion rises along with it. MinFE oiten
turns out to he a better indicator of a successful trade than MaxFE.
(Bottom) Short MinFE Example. As Westinghouse declines below $28 in early I 994,
the Minimum Favorable Excursion prier level from a short declines with it. The absolute difference between the entry price and the MinFE price newt gets smaller.
TWEAKING STOPS
15
Ive never had too much trouble getting the idea of excursion across
when explained with examples. Most people can see it as a simplified
price chart connecting highs or lows from the point of entry. If theres
a difficulty, its in believing there could be any regularity in these
graphs of excursion over time. That there could be consistency in how
far they move from the point of entry also seems unlikely to people.
For some sets of entry/exit rules, there is no consistency. These
rules dont have the ability to distinguish between good and bad
trades. So, to that extent, people are right to be concerned. The only
way to determine this is to look and see. If there is no regularity, then
with your rules, you have little basis for trading. If there is, B potentially profitable strategy becomes possible.
In excursion analysis, were concerned with the extremes of
movement so we can analyze stop placement, limit entry, or optimal
exit. All these occur at the extremes of price ranges. Moreover, without intraday data, we have no way of knowing what goes on inside price
bars. The limitations of the typically available data force us to deal
with what we do know-even so we should keep in mind that highs
and lows are so thinly traded that the values reported can only be considered rough targets.
TWEAKING STOPS
A tweak is a small analytical adjustment which, while not central,
may offer benefits.
In analyzing stops, we usually look for just one violation of a
given price level, thinking thats where our stop will be triggered. YOU
could require one, two, three, or more violations if you had intraday
data. To satisfy this requirement, the levels at which these semi-stops
might be triggered would be lower (for highs) or higher (for lows), allowing tighter stops. Operationally, this would require being there
to monitor action tick by tick, something most traders will not want to
do, but it might be feasible for someone whos trading that way anyway
and is sufficiently automated.
For those aware of differing speeds with which Treasury Bonds
or some other tradeable move up and down, a further refinement
would be to separate the data by longs and shorts. I once checked
16
DEFINING
MAX
ADVERSE
EXCURSION
2-l
Calculating Long MAE. This position opened 7/11/94 on the
close at 18.94. Since it was long (the negative 18.94 refers to a cash outflow),
The MAE computations referred to the subsequent lows.
Table
Date
Open
High
Low
Close
-l/12/94
19.15
19.28
18.93
19.09
7/13/94
,19.08
19.26
18.81
T/14,94
19.06
19.13
7/15/94
19.00
19.01
Lang at
16.94
MAE
Compgtation M A E
-18.94
= MAX,0.(18.94
0.01
18.96
16.94
_~ MAX[0,(18.94
18.81),.011
0.13
16.93
19.04
-18.94
18.85
18.R5
-1R.94
18.93),0,
= MAX,O,~lR.94
- l&93),.131
= MAX,O,W.94
0.13
0.13
18.85~,.13,
7/18/94
18.62
16.64
18.45
18.64
-18.94
7/19,94
18.46
16.76
1x.43
18.75
-18.94
7/20194
18.77
18.80
18.62
18.76
-18.94
= MAXlO,(18.94
18.45),.131
= MAX,0,(18.94
- 18.433,.49,
= MAX,O,(18.94
0.49
0.51
0.51
- 18.62),.51,
7121194
18.66
18.93
18.64
18.87
-18.94
7/22/94
19.00
19.07
18.93
19.01
-16.94
7/25/94
18.89
18.95
18.78
18.65
-1R.94
7/26/94
18.84
18.90
18.69
18.78
-18.94
= MAXLOJ18.94
18.64),.511
= MAX[0,(18.94
18.93),.511
0.51
= MAX[0,(18.94
- 18.78~,.51,
= MAX,O,(lR.94
0.51
- 18.69,,.51,
0.51
0.51
Treasury Bonds in the late 1980s but found no noticeable difference

in the MAEs.
SAMPLE
CALCULATIONS
Were not dealing with involved mathematics here but to make sure
the technique is clear, Im including a tabular example as well. Many
traders are very uncertain about mathematics but there is very little
math to worry about with this technique. Were just comparing new
highs/lows to (1) previous highs/lows, (2) previous maximum values,*
@ The MAX function in a spreadsheet sclccts tho highest value from the three values
separated by a *,I within the square brackets 7 I.
Table 2-2 Calculating MAE When Short. This position opened 12/21/93 on
the close at 15.12 (a positive value referring to a cash inflow from the short).
The MAE computations referred to the subsequent highs.
Short at
Date
Ooen
Hish
Low Close
15.12
Computation
12/27/93
15.32
15.32
14.86
14.90
15.12
12/28/93
14.91
14.97
14.78
14.84
IS.12
= MAX[O,(15.32
15.12),01
= MAX[0,(14.97
12/29/93
14.96
15.15
14.88
15.13
15.12
15.07
15.18
14.89
14.92
15.12
l/3,94
14.96
15.34
14.96
15.29
15.12
l/5/94
15.23
15.55
15.43
15.99
15.13
15.54
0.20
0.20
- 15.12),.21
12/30/93
l/4/94
M A E
IS.39
15.93
MAX[O,(15.15
= MAXlO,(15.18
- 15.12),.21
= MAX[O,U5.34
15.12),.21
15.12
0.20
- 15.12),.2,
= MAX[O,(15.43
0.20
0.22
0.31
15.12),.221
0.00
MAX,0,(15.99
0.87
15.12L.311
Table 2-3 Calculating MaxFE When Long. MaxFE grows steadily as a

long position advances favorably. Longs compare the highest high to date
with the entry price.
Date
Long a*
15.88
Open High
Low Close
417194
15.95
15.97
15.66
15.69
-15.88
4/8/94
15.68
15.75
15.62
15.70
-15.88
4/11/94
15.70
16.05
15.68
15.96
~15.88
4/12/94
15.93
16.03
15.77
15.82
-15.68
Computation MsrFE
= MAX[O, (15.97
15.88),01
= MAX[O. (15.97
15.88),.091
= MAX,0,~16.05
- 15.88),.091
_ MAXFO,
(16.05
15.88),.171
4,13/94
15.86
15.96
15.76
15.90
-15.88
4/14/94
16.03
16.07
15.79
16.05
-15.68
4/15/94
15.92
16.40
15.85
16.36
-16.88
4/18/94
16.32
16.46
16.20
16.31
-15.88
1.6.04
16.04
-15.68
4/20/94
16.08
16.22
15.96
16.16
-15.88
4/21/94
16.10
16.37
16.04
16.36
-15.88
= MAXLO, (16.05
- 15.88~,.17,
= MAXIO,
(16.07
- 15.88),.171
= MAXCO, (16.40
X5.88),.191
= MAX[O,(l6.46
- 15.88~,.52,
= MAXIO, (16.46
- 15.88),.581
= MAXCO, (16.46
15.88),.581
= MAX(0. (16.46
15.88),.581
0.09
0.09
0.17
0.17
0.17
0.19
0.52
0.58
0.58
0.58
0.5R
17
DEFINING
18
MAX
ADVERSE
EXCURSION
and (3) our entry point. A series of prices and the associated MAE for
a long position is shown in Table 2-1.
Table 2-2 is an example of a computation of MAE when going
short.
Calculating maximum favorable excursion is similar. Again,
compare the subsequent highs and lows to the entry point and each
other using the formulae above. Table 2-3 is an example for a long position, calculating MaxFE. Table 2-4 is an example for short MaxFE
computation.
These extreme values-MAE, MaxFE, and MinFE-are not always positive. Sometimes, for example, there is no favorable excursion. In this case, the MAX function serves to limit the value to zero
(Table 2-5).
Table 2-4 Calculating MaxFE When Short. A short goes well at first but
then turns bad. This experience makes the point that MaxFE never gets
smaller.
Open
High
Low
Close
Short at
15.12
12/27/93
15.32
15.32
14.36
14.9
15.12
KU28193
14.91
14.97
14.78
14.34
15.12
U/29/93
14.96
15.15
14.88
15.13
15.12
12/30,93
15.07
15.E
14.89
14.92
15.12
l/3/94
14.96
15.34
14.96
15.29
15.12
Date
Computation
MaxFE
= MAXLO, (15.12
- 14.86),01
= MAXIO. (15.12
14.%,.261
= M A X I O . (15.12
- 14.78),.341
= MAXIO,
(15.12
14.78),.34,
= MAXLO, (15.12
14.78),.341
0.26
0.34
0.34
0.34
0.34
Table 2-5 Minimum Extreme Value. Here a short at the close of 3/28/94
goes so far awry that there is never any MaxFE. Not only do MaxFE, MinFE
and MAE never shrink, they never go below zero.
Date
Open
High
Low
Close
Shorter at
14.33
3128194
3129194
14.41
14.58
14.33
14.33
14.58
14.33
14.33
3/30/94
14.55
14.67
14.43
14.65
14.33
Computation
= MAX,0,~14.33
14.33),0,
= MAXl0,(14.33
14.43),01
MaxFE
0.00
0.00
19
SAMPLE CALCULATIONS
Table 2-6 Calculating MinFE When Long. Going long at 15.88 takes a long
time to bear fruit as MinFE takes eight days to rise above zero. For many
trades, there is no minimum favorable excursion.
Date
Long at
16.88
Open High
Low Close
4/l/94
15.95
15.97
15.66
15.69
-15.88
4,8/94
15.66
15.15
15.62
15.7
-15.38
411 l/94
15.7
16.05
15.66
15.96
-15.86
4/n/94
15.93
16.03
15.17
15.82
-15.63
4/13/94
15.86
15.96
15.16
15.9
-15.63
4,14/94
16.03
16.07
15.79
16.05
-15.88
4/15/94
15.92
16.4
15.85
16.36
-15.88
4/M/94
16.32
16.46
16.2
16.31
-15.86
409194
16.15
16.23
16.04
16.04
-15.66
4/20/94
16.03
16.22
15.96
16.16
~16.63
4/21/94
16.1
16.37
16.04
16.36
-15.68
4/22/94
16.47
16.76
16.33
16.74
-15.88
4/25/94
16.62
16.91
16.55
16.37
-15.88
4126194
16.61
16.69
16.53
16.59
~~,15.68
4/26/94
16.43
16.63
16.37
16.4
-15.66
4/29/94
16.38
16.65
16.25
16.63
-15.88
5/2/94
16.64
16.97
16.64
16.63
~15.33
5/3/94
16.16
16.8
16.51
16.64
-15.88
514194
16.68
16.68
16.55
16.51
-15.88
Computation
MinFE
= MAX10.(15.66
,.
15.38),01
= MAXCO,W5.62
15.88),01
= MAX10.Cl5.66
- 15.68),OI
= MAX,O,U5.77
- 15.68~,01
= MAX[0,(15.16
.. 15.88),01
= MAX[0,(16.19
- 16.88),01
= MAX,0,(15.85
- 15.88),01
= MAX,0,~16.20
- 15.88),01
= MAXl0,(16.04
- 15.88),01
= MAX[0,(15.96
- 15.88),01
= MAX[0,(16.04
15.88),01
= MAX[0,(16.33
- 15.88),01
= MAXL0,(16.55
15.88),01
= MAXCOJ16.53
15.88),01
= MAX,0,(16.37
- 15.86),0,
= MAX10.Cl6.65
,.
15.33),01
-~ MAX10.Cl6.91
- 15.kR),Ol
= MAX,0,(16.80
- 15.88),0,
= MAXL0,(16.83
15.88),01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.32
0.32
0.32
0.32
0.45
0.61
0.67
0.67
0.61
0.76
0.16
0.16
20
Table 2-7 Calculating MinFE When Short. Short goes bad after initial
surge! The MAX function serves to capture the initial favorable movement
and retain it as the ending MinFE value.
Open High
Low
Close
Short at
15.12
12/27/93
15.32
15.32
14.86
14.9
lS.12
12/28/93
14.91
14.97
14.78
14.84
15.12
12/29/93
14.96
15.15
14.88
15.13
15.12
12/30/93
15.07
15.18
14.89
14.92
15.12
l/3/94
14.96
15.34
14.96
1.5.29
15.12
l/4/94
15.23
15.43
15.13
15.39
15.12
Date
Computation
MinFE
= MAX,O, (15.12
15.321, 0 1
_ MAX,O, (15.12
0.00
= MAXIO, ClS.12
- lS.lS~,.lS,
= MAXIO,
(15.12
15.18),.151
= MAXLO, (15.12
15.34),.15,
= MAX,O, (15.12
- 15.43~,.15,
0.15
l&97,, 0.001
0.15
0.15
0.15
0.15
Minimum Favorable Excursion tracks the least favorable price

excursion from our entry. It compares the lows to the entry if we are
long or the entry to the highs if we are short (Table Z-6).
Table 2-7 is an example of a short position from 15.12 and the
computation of minimum favorable excursion, MinFE.
These examples help to compute the MAE, MaxFE, or MinFE for
any one trade. For exemplary Excel code, see Appendices A, B, and C,
respectively.
3
DISPLAYING MAEI
AGGREGATION
In Chapter 2, I showed how to measure MAE, MaxFE, and MinFE. I
included some sample Excel code in the appendices and, for such a
simple concept, it generated a lot of spaghetti code. This chapter
deals with the next problem in using MAE: assembling the collected
measurements and displaying them in a way that makes sense and
contributes to concrete decisions on where to put stops.
Personally, I like to see pictures. I can inspect tables of results
but Im more comfortable with a picture of the results than with a
table of results. Consider what would be the best way to show the long
list of MAE measurements.
As a rule of thumb, having thirty or more trades that are losers
and thirty or more that are winners should provide enough data to
have reasonable confidence in the results. The list might begin as
shown in Table 3-1.
Just these few items represent a lot of information, but picking it
out of several hundred lines is problematic. The models in Appendices
A, B, and C have the structure to eventually show day-by-day (1) time
in trade; (2) MAE, MaxFE, and MinFE; (3) trade profitability and
21
22
DISPLAYING MAE
Table 3-l Collecting MAE Data. For each trade, MAE is measured and
recorded along with the net profit or loss from the trade. In this data, commission and slippage are omitted but the analyst can easily factor this into
the profit/loss computation.
Date of
Entrv
Entry
Price*
Date of
Exit
Price
Profit or
LOSS
6/23183
-31.04
-31.2
30.96
-31.17
30.69
29.98
30.09
29.03
29.36
-29.26
-29.18
-30.62
-30.49
711183
31.18
31.96
-31.22
31.21
-30.34
-30.31
-28.14
-29.6
.14
.76
-.26
.04
.35
-.33
1.35
T/8/83
WW83
9/22/83
9/28/83
10/26/83
11/9/83
l/2/84
Z/10/84
Z/15/84
Z/24/84
3/29/84
4/13/84
Exit
8/16/83
g/14/83
9/23/83
10117183
11/4/83
12/21/83
l/16/84
2/14/84
Z/16/84
3116184
4/12/84
4/16/84
MAE
.a7
.a1
.46
0.00
0.00
.34
0.00
-.57
.12
-29.24
29.34
30.13
30.5
30.47
.64
.03
.08
.45
0.00
-.02
.02
.06
-.12
25
*A long is represented by a negative price-the cash outflow in taking a long position.
(4) account equity, all of which will he used later. For now, this is the
question: Is there any difference between winners and losers, any difference we can use while were in the trade?
To get at this, separate the results by winners and losers as
shown:
Win
.14
.76
.04
.35
1.35
.12
.08
.45
Winning
.07
.Ol
.oo
.oo
.oo
.03
.oo
.06
MAE
Loss
-.26
-.33
-.57
-.12
-.02
Losing
MAE
.46
.34
.64
.25
.02
23
AGGREGATION
Tweaking Data Slicing

An enterprising analyst will immediately think of separating the trades
by longs and shorts or by duration of trade. Ive done this on the few
trading rule sets I use, to little advantage. Still it seems to me that this
could be worthwhile.
Inspecting these numbers, notice that the winning trades have

very small MAEs while the losers tend to have larger numbers. This
is the seminal observation about MAE and it verifies the experience
of centuries of trading. The anomalous characteristic of the list is
that there are more winners than losers, something Ive rarely found
in trading systems. To display the information in these lists, turn to
the graphics portion of your spreadsheet. Assemble the two columns
like this:
Profit
MAE
0.14
0.76
0.04
0.35
1.35
0.12
0.08
0.45
-0.26
-0.33
-0.57
-0.12
-0.02
0.07
0.01
0
0
0
0.03
0
0.06
0.46
0.34
0.64
0.25
0.02
When selected and charted, they should resemble Figure 3-1.

Generally, a rough linear relationship between the size of the loss
and the size of MAE should exist while the MAE for winners will be
24
DISPLAYING MAE
0.8
.
I .
3:12- 4-m-W
1
-1
0.6
-0.5
0.5
1.5
Profit or Loss
3-l
MAE vs. Profit/Loss. This analytical chart highlights the distinction
between the MAE for winners and that for losers. The winners, to the right of the
vertical axis along the horizontal axis, have MAEs less than .I while the losers scatm
tered to the left of the vertical axis, have MAEs greater than .l (save one).
F,igure
ielatively small. If not, the market may have had a bout of disfunction or your rules may be unable to distinguish between winners and
lOSSI%.
Stop and think for a second about how youre seeing trading data
now. Instead of a summary table of wins and losses, Sharpe ratios,
drawdowns, results for shorts and longs, and so forth, youre seeing a
picture of your actual experience with, in this case, thirteen trades.
Youre also seeing all the market action from the viewpoint of the
trade entry, not from arrows on a price chart. Isnt it striking that,
from this viewpoint, there is some regularity to the market action?
Seen this way, statistically, from your point of entry, might there be
other regularities?
Putting up a chart like this serves to find outliers that may be
real or artifacts. In our example, all appears normal: the size of the
maximum adverse excursion rises as the size of the loss rises plus the
maximum adverse excursion for winners stays relatively low no matter the size of the win.
Unfortunately, even winning trades can go bad a little bit. Looking at Figure 3-1, we see that winning trades might have a maximum
adverse excursion of up to .l, which happens to be ten ticks. What if
we knew for certain that any trade that went more than ten ticks bad
FREQUENCY DIAGRAMS
25
would not be a winner? If it were not to be a winner, it would necessarily become a loser, right?
If we knew that, then, right in the middle of the trade, wed have
valuable new information about what to do. Wed see it go fifteen
ticks in the wrong direction and wed know we had a loser on our
hands. Trading experience would tell us to get out while the loss was
small.
Perhaps youre even more decisive. When the trade is put on, you
put a stop at eleven ticks. Youre ready to say, Dont call me for a decision, just get out if its to become a loser. Youd have automated the
process of keeping your trading losses small.
Now, lets be realistic. We dont know for certain that a trade that
goes eleven ticks wrong is definitely going to be a loser. From our sample of thirteen trades, we just have an estimate of that likelihood. I
wont bore you with the mathematics of the statistical estimate. Instead, just look at the picture. You can see where the winners MAEs
cluster and how bad they get. From the picture, you can see how
things go. Just keep in mind that unusual things happen in a market
subject to countless random shocks. The picture enables you to estimate roughly where the good news stops and the bad news begins.
FREQUENCY DIAGRAMS
There is another type of picture of these numbers that gives even more
detail and, later, will make better decisions possible. This type of picture is called a frequency diagram. If you are familiar with them, you
may choose to skim (or skip) this section.
As the number of trades increases, diagrams like Figure 3-l become a little rough for picking stop points and also for perceiving
whether there is a distinction between MAEs for winners and losers.
To get around this, categorize the data by the size of the MAE. (That
is, by the size of the potential stop. We take the trouble to measure
MAE so we can find a reasonable stop and/or reverse point consistent
with our trading rules experience.)
For a first cut, just make the categories equal to .1 or ten ticks.
All the trades that have MAEs from 0 to .l, inclusive, will get lumped
together. Then all those with MAEs greater than .l and up to .2 will
26
DISPLAYING
MAE
be lumped together. Then those greater than .2 and up to .3, and so

forth. Now the data looks like:
o=
.l <
.2 -c
3 <
.4 <
.5 c
.6 <
.7 <
M A E M A E MAE M A E M A E M A E MAE MAE
Winners
LOSLXS
<=.l
<=.2
<=.3
<=.4
<=.5
8
1
<=.6
<=.7
< = . a
The data are shown graphically in Figure 3-2. This kind of display
is called a frequency diagram because it shows how often (that is,
how frequently) trades fall into different specific categories which
makes the distinction between the distributions of winners and
losers quite plain.
Figure 3-2 is stark compared to the typical result with sixty or
more trades but the distinction between the two sets of trades should
be clear. You hope to find a sharp cutoff like this one for the winners
but, if thats not possible, at least a distinct difference in the shapes
of the distributions.
7 1:
0.1
0.2
l-l: 0.4l-l: 0.5rl;
0.3
0.6
l-l:
0.7
0.0
Figure 3-2 Trades vs. MAE. Converting the data to graphics highlights the distinction between the MAE distribution oi winning trades and that of losing trades.
FREQUENCY DIAGRAMS
27
What you typically find is that the distribution of losers has a

peak value somewhat to the right of the peak value of the winners.
Another feature is that the distribution of losing MAEs has a long tail
to the right; hopefully, but not always, the distribution of winners
does not have a tail at all.
Ideally, the distribution of winners should have a huge cluster of
trades in the very first category or bin with many fewer trades in
the next few bins, then none at all in the bins to the right of the chart.
It would look something like this Figure 3-3.
Were just fortunate that this sample distribution came out so
neatly. The real world doesnt always work this way. For a taste of this,
Figure 3-3 is the distribution from 3,069 days of trading data in a
crude oil contract. (Thats over twelve years.)
The two distributions are still clearly different. The distribution of losing MAEs goes far off to the right which indicates that
many losing trades had adverse price movements. After many years
of trading with a consistent set of rules, only three winning trades
30
25
I
11 Winners
-.- LOsarS
y)
.
. rq\.,=> .---I..
,
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Maximum Adverse Excursion

Figure 3-3
Trade Frequency vs. MAE. An idealized version of the two distributions shows the winners with a sharp peak to the left and the losers with a peak
somewhat to the right and a long (ail.
DISPLAYING MAE
20
ever had adverse excursions of more than .45, a remarkable thought

since .45 in crude is a loss of only $450 (The details of this trading
system, a simple moving average system, are in Campaign Trading!)
The point of all this effort is to look at something like Figure 3-4 and
come to a trading decision, a decision that has nothing to do with
entry or exit but only with controlling losses. You want to find a place
to stop a trade thats going bad. The stop point shouldnt be too far
away nor should it be too close to the entry point. Rather than look at
retracements, wave counts, support/resistance levels, percentage price
moves, arbitrary money management points, parabolics, or any of the
other ways to pick wrong points, here you look at the experience displayed in Figure 3-4.
80
W i n n e r s
n
70
-.-
Losers
10
0
l-l,
0.3
mm<=>.
0.6
0.9
1.2
.-.-.
1.5
1.8
2.1
MAE in Trading Points

Figure 3-4 Exemplary MAE Distributions. The distinction after 252 trades is
still clear. The distribution of winners does have a single trade each at .9, 1.05,
and 1.5.
29
FREQUENCY DIAGRAMS
As a first cut (subject to a closer look in Chapter 41, you just eyeball the graph and estimate where its no longer worth your while to
stay in. In Figure 3-4, points that suggest themselves are at .15, .30,
and .45 away from the entry price. (These are the values on the horizontal axis.) Each represent a wider stop you could set.
Looking at Figure 3-5, the dropoff in number of wins as adverse
excursion grows is striking. Hardly any winners have trades go
against them more than .45, but there are three:
1
0.15
0.3
0.45
0.6
0.75
0.9
1.05
1.2
1.35
1.5
1.65
1.8
1.95
2.1
Totals
80
17
4
0
0
1
1
0
0
1
0
0
0
0
104
21
44
28
28
7
10
2
3
1
0
0
1
2
1
148
On the other hand, 55 losers go wrong by more than .45 from the
entry point. Therefore, as a first cut, setting the stop at .46 would cut
off 55 losers and convert three winners to losers. Setting the stop at
.31 would cut off 83 losers and convert seven winners to losers. The
most aggressive strategy would be setting the stop at only .16 where
wed have 80 winners and 172 losses. (You have all 148 losers plus
youd convert 17 + 4 + 1 fl + 1 = 24 winners into losers for a total
of 172.)
30
DISPLAYING MAE
80
60
Possible
Stops
10
MAE in Trading Points

3-5
Possible Stops. Once an MAE diagram is drawn, its used to pick
points to place stops. In this particular situation, there seems to be little advantage to holding on past .45 where there are far more losing trades than winning
trades.
Figure
Theres clearly a tradeoff (which discussed in Chapter 41, but the

cut at .45 is straightforward. It implies that to limit losses to, say, 2%
of trading capital, you need:
Capital needed = (.45
100)
$10/.02 = $22,500
to trade this single risky strategy. (It doesnt mean that the whole
$22,500 would be tied up supporting a single trade.) In this way, experience gives you a relatively precise estimate of your capital requirements and your stop-loss point, items that have been vague 01
without estimate in the past.
If those three winners are stopped out and converted to losers,
what would that have cost? Would it have been better to stop out at .3?
FREQUENCY DIAGRAMS
31
This subject is covered in Chapter 4. For now, there are some details of
this technique to clean up.
Picking Bin Size
For one thing, why should the categories we use be .15 or .30 or any
particular number? Computationally, theres little to recommend any
specific choice because todays spreadsheets can generate frequency
diagrams very conveniently. The details of calculating them and generating the charts in this chapter are discussed in Appendix D, Generating a Frequency Distribution.
The factors that control your bin size relate to your loss control.
The point of making MAE charts is to find a point at which profits
will be maximized with stops or reversals. Therefore, you want to construct the categories to help you find that point most accurately.
As a beginning point, there is your trading capital. Only a certain portion of that should be risked on any one speculative position
and that portion is generally quite small, about 2%. At this level, your
risk of ruin is quite small so the question is: Can you trade at this
level of loss? Is there a stop point that is less than 2% of your trading
capital? As a first cut, then, the category size should be no greater
than 2% of your trading capital. Otherwise, you wont be able to see
on the graph a stop point of that size or less.
On the hopeful thought that the stop point might be less, even
considerably less, than the 2% level you could make the bin size half
that amount or even one quarter that amount.
On the right hand side of the X-axis of your MAE chart, there is
the trade with the worst adverse excursion. If its huge and youve selected a very small bin size, youll end up with lots of empty bins between those on the far left and those on the far right of the graph.
From an analytical standpoint, you can graph just those on the left
side that interest you (those with small adverse excursions), but from
the standpoint of a presentation, such a large gap diminishes the impact of the information in the bars to the left.
The practical limitation on this fine division is often the amount
of data you have-the number of trades. For strategies that dont trade
often, you may have trouble coming up with thirty winning examples
and thirty losing examples, let alone dividing the thirty into ten or
32
DISPLAYING MAE
twenty sparsely populated bins. Theres no reason it should always

work but my rule of thumb using a bin size of 1% of trading capital
(that is, half the 2% loss limit) usually generates a workable display.
If it turns out poorly, just change the bin values in your spreadsheet
and recalculate.
It will have struck stock traders that I havent distinguished between

short and long trades and they have a point. Trading futures, I have
rarely seen much difference between behavior in up and down markets with the exception of the bond market where the mathematics of
yield makes a slight difference in the speed at which the market
moves. Stocks seem to have a complete variety of price movement in
both directions but Im less familiar with them.
A trader suspecting a difference between the behavior of a short
position in a stock and that of a long can keep separate records. In Appendix E, MAE for Shorts and Longs, Ive shown exemplary coding
for tracking the MAE for shorts and longs along with an example of generating the frequency distributions used for graphs. Once youve generated the graphs for shorts and longs, you just make a judgment on the
stop point for each and compare the two judgments. Whether they are
significantly different or not is, naturally, yet another judgment.
4
D E F I N I N G P ROFIT
BY
PROFIT
BIN
TRADEOFFS
Its all very well to see the numbers of trades that occurred in a given
bin but, at some point, youll want to know what the impact on profitability is as you widen or narrow stops. This usually comes up when
youre trying to pick a stop point and you want to know if theres much
of a profit difference between one level and another.
Perhaps two adjacent bins have nearly the same frequency of
wins and losses-youre curious if theres much difference in profit
when you set the stop at one bins level or the others, though you
wouldnt expect it. More often, the frequency curve for winners slopes
down nicely as that for losers rises. At what stop point do the profit
curves cross? Were there one or two inordinately sized wins or losses
distorting the curves? In other words, is the experience in your data
reliable and should you make trading rules using it? These are all
good questions, but lets define profit curues first.
33
34
DEFINING PROFIT BY BIN
PROFIT CURVES
In Chapter 3, I showed how to construct frequency diagrams or
curves. Curves refers to the smoothed shape of the line between the
plotted points.
Now, instead of plotting number of trades, you want to compute
the profit or loss on each trade within each category (or bin) and plot
the mm of those profits and losses for each bin. This seems straightforward, but there are some nuances to discuss after Ive shown the
prOCe*S.
To begin, youll have recorded for each trade its net profit or loss,
including commissions. Comparing these figures to MAEs and your
bin sizes, separate the trades into appropriate bins and total the profits.* For example, this table of trades:
Profit/Loss
-0.16
0.03
-0.55
0.16
0.07
-0.05
PO.21
-0.28
-0.16
-0.13
0.5
-0.06
0.02
MAE
0.18
0.11
0.55
0.00
0.03
0.22
0.21
0.36
0.22
0.23
0.00
0.13
0.25
* I me profits in the general tense here. Losses are negative profits.
35
PROFIT CURVES
becomes this:
Winners
Bin
0.0
.ll
.21
.31
.41
.51
Profit/Loss
.l
.2
.3
.4
.5
.6
= 0.16 + 0.07 + 0.5

= 0.03
= 0.02
LOSeI%
Bin
0.0
.ll
.21
.31
.41
.51
Profit/Loss
- .l
.2
- .3
- .4
- .5
- .6
= -0.16 - 0.06
= -0.05 - 0.21 - 0.16 - 0.13
= -0.28
= -0.55
whichcomputes as:
Winners
Bin
Profit/Loss
0.0 - .l
.ll ~ .2
.21 - .3
.31 - .4
.41 - .5
.51 - .6
0.73
0.03
0.02
Losers
Bin
0.0 - .l
.ll - .2
.21 .3
.31 - .4
.41 - .5
.51 - .6
Profit/Loss
-0.22
-0.55
-0.28
-0.55
36
DEFINING
PROFIT
BY
BIN
Losers again show far larger MAEs than winners. To compute profit
curves, segregate the trades by winners and losers, just as MAE
trades are, and further categorize them into bins defined by the size
of the adverse excursion.
This, in turn, displays graphically as shown in Figure 4-1.
Seen in this fashion, the tradeoff between setting stop/reversals
at .21 or .31 is clearer. If this were all the trades available as evidence,
youd see theres little to be gained from putting stops as these higher
values. (This is actually the last few trades in an eleven-year sequence
of trades.) The losses taken at that level, both in number and size, outweigh the few wins kept by allowing the wider stops. Youd want a stop
at ;ll and no doubt about it. To make this point clearer when dealing
with more trades than this, I usually plot both winners and losers
profits/losses as absolute values (Figure 4-2). A summary display may
be even better for some people. Figure 4-3 sums the losses and wins
in each bin to give a single curve.
Figure 4-l
Exemplary Profit Curve. Shown graphically, the losers losses vs. MAE
are even more distinct. To aid in picking stops and reversal points, the losws absolute values (the positive value) are usually plotted as in Figure 4-2.
INTERPRETATION
37
Figure 4-2
Absolute Value Profit Curves. To easily compare the size of the wins
by bin with the size of the IOSS~S by bins, it may be easier to plot the absolute
value of the losses. The format chosen for this display should he one you find
easiest to interpret, since your judgment will be on the line when start trading this
information.
Excel code for calculating all this on an automated basis, using

the data format for the previous appendices, is in Appendix F. Its important to do this on an automated basis because too many errors
creep in when doing it manually. Especially when you get enough
trades to be statistically significant, youre handling a lot of data and
the manual workload becomes burdensome.
Appendix F also shows how to generate three-dimensional charts
that, in complex situations, may be helpful in picking stop/ reverse
points.
INTERPRETATION
Normally, sample data wont be as clean as that shown here. The
curves for the winners will probably overlap that for the losers far
DEFINING PROFIT BY BIN
Figure 4-3
Total Profits. Summing the gains or losses from hoth winners and
loser, hy hin clearly shows the point at which adverse price movement suggests a
change in stance. Not only would stopping wt at .21 he healthy, hut there may he
gains from reversing.
more than they did in this exposition. When this happens, you will
make a judgment based on the overall trend of each curve and by visual estimation of the optimal crossing. Given the roughness of the
numbers, your visual judgment is wholly proper and probably superior
to mathematical algorithms.
Keep in mind that for these numbers to really work out you need
enough occurrences to generate reliable estimates. If theres a significant difference between the frequency curves of the trades and the
profit curves of the trades, the usual reason is that the number of
trades falls off rapidly for the winners as stops are widened, thereby
making the few remaining trades a greater influence.
Said differently, profit curves are very sensitive to the size of individual wins or losses. This is their limitation since one big win on a
trade with a large MAE can make your curve look like Pikes Peak in
the middle of the prairie. You need enough trades to make sense in
the low MAE ranges and you should take the profits shown in the large
MAE bins with a grain of salt. Typically, in the large MAE bins, you
are looking at the results with virtually no impact from stops.
INTERPRETATION
39
Sometimes, widening or tightening stops produces profit curves

that seem very counterintuitive. Generally, as you widen stops, (1) winners that had been stopped out at a loss before become winners again,
(2) losers that had been stopped out as a lesser loss become bigger losses
or are stopped out at a more expensive loss. The net effect is dependent
both on the frequency of each type of trade and the size of the losses
they experience.
As you tighten stops, opposite effects occur. More winners are
stopped out instead of going on to become winners and more losers are
stopped out earlier to become smaller losers. In both situations, the ef,fects are countervailing and we find the net impact by inspecting profit
graphs. Whenever resulting graphs seem odd, inspect the tabular data
for the trades that produced counterintuitive effects. Youll soon become
familiar with these effects and be able to factor them in to your judgment about stop placement.
Whatever your judgment, you are certainly better informed
using these procedures than if you were setting stops by arbitrary
money management rules that bear no relation to actual experience.
5
IMPACT 0F
V OLATILITY C HANGES
TWEAKS
With the basic concept of adverse excursion described in both text and
example, turn now to slight adjustments in the concepts that might
provide better results and, in any event, satisfy curiosity. In my days
as an analyst, we always called these adjustments tweaks to indicate
that, while they were not critical to success, they did slightly shift
the odds in the proper directions. The next chapters are devoted to
tweaking the use of adverse excursion in trading.
The first tweak relates to range volatility (as distinct from classical volatility). Then theres the effect of runs on the use MAE stops.
Last, I take up issues of betting strategies.
Range or Volatility to Change Our Stops?
In focusing on price excursion, price ranges are the main concern.
Highs and lows, being extremes, tend to be very lightly traded compared to the central values of the trading day. Nevertheless, if they do
41
42
IMPACT
OF
VOLATILITY
CHANGES
touch ones stop, your carefully placed order will be set off. The question for this chapter is to what extent volatility affects range which
might, in turn, affect the adverse excursion measurements and the
stop or reverse points that MAE suggests.
The intuitive idea is that, as intraday trading becomes more
volatile than previously, ranges are likely to expand and stops become
more likely to be hit, particularly closely set stops like MAE stops.
Contrarily, if intraday volatility is lethargic, ranges will contract and
stops are less likely to be hit (Figure 5-l).
Range and Vohtility
There are many logical questions about this concept. First, do price
ranges expand or contract with classically defined volatility? On this
point there is some quick evidence. Experiences such as that in Figure
5-2 suggest that range is related to classically defined volatility,
though not rigorously. Figure 5-2 is a mean-adjusted comparison of
20.day volatility with the 20.day average range. The qualifier meanadjusted tells you that the means of the two distributions have been
Gain
or
LOSS
L
---------- MAE Stop
-
Levels
Missed stop
Adjusting MAE Stops. Stops might be adjusted as range volatility

Figure 5-l
changes during trading. The hope would be that trades that would otherwise be
profitable would not hit any expanded stops. Another possible benefit: Stops could
be set more snugly during periods of low range volatility.
TWEAKS
Figure 5-2 1995 New York Light Crude Range and Volatility. Though volatility
fluctuates to rclativcly grcatcr exrremes than daily range, the relationship is oiten
quite direct. In crude, range may be an effective proxy for volatility.
set at the same value by multiplying one series by an arbitrary adjustment factor. The result is a display that facilitates comparisons of
the changes in the two series, not an analytical construct useful for
establishing definitive relationships.
Figure 5-2 doesnt show overlapping lines but it does show
changes in each series happening in roughly the same direction in
reasonably comparable time periods. This is just one instance. Figure
5-3 is another example where changes in volatility and range might
show some relationship to each other. You would need to run something similar for the tradables ycm use to get a first cut at whether
this was a relationship you could use. It might be quite a study and
this isnt the place for a detailed treatise on the relationship of range
to classically defined volatility. I havent found treatments of this subject in the academic literature, but it has been a subject of proprietary
IMPACT OF VOLATILITY CHANGES
Figure 5-3 1990 New York Light Crude Range and Volatility. I picked 1990, a
year of spectacular price fluctuation in crude, as an example of range and volatilitv coincidence.
investigation by traders and options dealers.* Anecdotal evidence in

the form of charts in Appendix G suggest that range is a proxy worth
investigating.
However, after looking at charts like Figures 5-2 and 5-3, I concluded that range might expand with volatility, but not necessarily,or
coincidentally. Therefore, to adjust stops for changes in price volatility was problematic; range itself would be the more direct indicator.
The charts provided evidence that range did expand and contract materially, usually in concert with price volatility. Given that, the impact of range fluctuation was worth further study.
* Authors who have linked range and volatility include Cynthia

(pp. 432-436); Andrew Sterge S&C V 7:12 (pp. 438-441).
Kase S&C Vol 11:IO
TWEAKS
45
At the Point of Entry. The second question about range volatility is

whether, at the point of entry, youll even know of a change in range
volatility (or price volatility, for that matter). That is, will you know
that it has changed or is about to change and that, therefore, youll
need to adjust the stop point right at the moment of putting on the position. Trading rules that enter upon breakout from a trading range
or at the beginning of a trend will often come from a period of relatively quiet price fluctuation (Figure 5-4). The trading rule sees the
sudden movement up or down, but the ZO-day average range or volatility doesnt see the change as quickly due to the averaging process typ
ically used.
It was this inability to look ahead that stimulated the development of MAE. Instead of trying to look ahead, MAE analysis asks
What is the actual experience of adverse price movement given these
trading rules? and Given that experience, is there an operational
difference that can be used at the point of entry?
Taking that tack, you, the analyst and trader, need to examine
what happens to range after you enter, using your trading rules. Just
Figure 5-4 Range at Entry vs. All Kanges. The averages of these fwo distributions are identical: .32, but the distribution of all ranges is more peaked and
skewed. Looking at this trend trading in crude, ifs striking that entry was rarely
from the most frequent ranges, the low ones at .lS, but from ranges around .25.
46
as in MAE analysis, you may find nothing you can exploit. Your combination of tradable and trading rule may not identify periods of
change in range. If it does give reliable estimates of changes in range,
then you have the first prerequisite for adjusting your stops and reverse points.
For example, beginning on your point of entry just measure the
amount the daily range exceeds the 20-day average range on the entry
day. Another comparison that would smooth the data would be to compare the range each day to the moving 20-day average range. An example or spreadsheet of doing these is given in Appendix H and graphed
data from both exercises is summarized next.
Comparing to the Moving 20-Day Average
Though this varies with the tradable, in my experience, if you compare the daily ranges after entry to the moving ZO-day average during
the trade, crudes result is common: the range doesnt vary statistically after entry. The daily range after entry, for the set of rules used
here, is usually close to the moving average of the range. Its important to inspect your results visually as well as statistically, though.
Looking at Figure 5-5, youll see plotted the range after entry
less the moving 20.day average of range. The average value of this
distribution is zero, but simply by looking at it, you can see that its
skewed upward and its the skewed values that might hit OUT stops.
In Figure 5-5, only one value goes below -2 and that barely. In
other words, range rarely shrinks more than 20 ticks below the moving 20-day average range. On the other hand, range often expands
but it expands eppisodicaZ2.y.
Figure 5-5 shows only the occurrence that ranges expand by more
than +.2. Its difficult to see when that occurs in the life of the trade.
Transforming Figure 5-5 to Figure 5-6 shows that there is typically
only one, occasionally two, days when the range expands by more than
+.2. That is, to emphasize, range doesnt usually expand after entry
and stay expanded. (This is specific to the trading vehicle and the trading rules. Your specific situation may give different results.) There is
an occasional blip upward but it is not held. It is that blip that hits
stops. (The shrinkage minimum of -.2 shouldnt be taken as absolute.
Later, more data will show that it can range as large as p.6.)
47
TWEAKS
Day of Trade
Figure 5-5
Range after Entry. Subtracting the 20.day range from the range on
each day after entry often produces a plot like this. Most values for range are between -2 and + .2 of the value on the day of entry. Also, while range rarely shrinks
more than -.2 (range is limited to zero), it often expands rrwre than +.2.
Day of Trade
Figure 5-6
Range Expansion. Most trades dont show expanded range after entry
on a consistent basis. By plotting range less range at entry, you can see that if range
is to expand more than .2, it will do so episodically.
48
What about the difference between winners and losers? Is there

any appreciable difference in the behavior of ranges? As with MAE,
separate the two range excursions and plot to get something on the
order of Figure 5-7.
Keep in mind that Figure 5-7 plots range after entry less the
moving 20-day average of range; its a difference not the actual range.
Most of the values are less than .2 and most of them are around or
,slightly above zero. The episodic nature of the range expansion is
there again: only one trade sustained a range expansion greater than
.2 for more than one day. (That trade was the one winner whose range
expansion was greater than .2 for five days.)
The average value of all the range expansions is slightly different. For losers, its .00808, not significantly different than zero. For
winners, its .02602, which might be barely significant depending on
your chosen level. (Keeping in mind that the measurements only go
out to two digits, this citation is an example of statistical humor.)
,Parametric statistics dont capture the trading situation is displayed
as illustrated in Figure 5-8.
In Figure 5-7, the mean is highlighted at zero with some skew
toward the upside as suggested by Figure 5-6. The actual stats in
Table 5-1, though, are not exceptional given the episodic nature of the
range expansions.
Fortunately, for this set of data, if expansion is to occur, it will
generally occur with winners. Add to that our knowledge from MAE
work that winners dont generally mcwe very far against the entry.
Though this isnt always the case, the directional bias of winners is favorable (naturally!), so the bias of the range expansion we see in winner8 is generally (not always) favorable. Were it unfavorable, wed see
a greater amount of adverse movement.
Once again, your data might show different behavior.
Comparing to the ZO-Day
Average Range on the Date of Entry
Range at Entry vs. Range after Entry. Theres another way to look at it.
The data in Table 5-l compare the ranges after trade entry to the 20day simple moving average of the range during the trade. The virtue
of this is smoothing, but the future average range isnt a value you
know going into the trade. You do know the value of the average range
49
TWEAKS
Winning
Trades
Losing
Trades
Figure 5-7
(Top) Winners Range Expansion. With more examples, range can
contract dramatically (one instance went to -.h) but rarely. More common is expansion going above the .2 level seen in Figure 5-6.
(Bottom) Losers Range Expansion. Not only do losers last just a short time, but
their range expansion is less than that of winners. Range contraction also appears
more common in this limited sample.
50
Figure 5-8
Range Expansion vs. Moving 20.Day Average of Range Winners and
Losers Range Expansion Both Average Nearly Zero. Though theres a slight tendency to grow in both distributions. skew is just 1 .l for losers and 2.5 for winners.
The greater frequency of winning examples sterns from the fact that winning trades
last longer than losing trades. (Note the scale shift on the X-axis.)
Table 5-l Range Expansion for Winning and Losing Trades. The distributions described here are those in Figure 5-l.
n (Number of Days
in Trades)
Max
Min
Average
Std. Deviation
Skew
Kurtosis
Losers 139
0.68
-0.43
-0.02
0.15
1.03
4.27
Winners 374
1.43
-0.60
0.03
0.21
2.54
12.4R
51
TWEAKS
on the day of entry. The summary of an analysis done in Appendix H

compares (hy subtraction) ranges after entry to the 20-day simple
moving average of the range on the day ofentry (Table 5-2).
Just looking at the statistical summary in Table 5-2, virtually
the same raw data for Table 5-1 produce a different picture. Losers
contract range about the same amount and winners more than double
their range expansion from .03 to .08. The maximums for winners rise
as well while the minimums for winners decline. The normalized statistics, standard deviation, skew and kurtosis, dont change significantly. The scale of the expansion? Its about .1X+/.32 = 25% of the
20-day average on the day of entry for this sample.
What are the shapes of the distributions of range expansions for
winners and losers in this case? They are shown in Figure 5-Q and
its clear that the distributions are different. While the distribution
for losers peaks in the -0.1 range (that is, expansions between -0.1
and -0.29991, that for winners peaks in the range between 0 and
-.099999, experiencing some range contraction too. Despite this, the
distribution for winning trades shows a very long, substantial tail to
the right as well which pulls the mean over to .08.
Its clear that, in this sample-a sampling youd do on your own
tradable/trading rules combination--losers dont become losers because they drastically expand their range; instead, their tendency is
Range after Entry vs. Z&Day Average Range at Entry.

Comparing this table to Table 5-1, the distinction between winners
and losers is that winners expand range by 8 ticks c.08 trading points)
while losers slightly contract their range. The very high standard deviation for both winners and losers stems from the broad tails of both
distributions.
Table 5-2
n (Number of Days
in Trades)
Losers 139
M&X
Min
Average
Std. Deviation
Skew
Kurtosis
0.73
-0.44
-,-0.03
0.16
0.97
3.75
Winners
1.79
-0.39
0.08
0.25
2.58
10.58
374
52
Figure 5-9 Range Expansion after Entry. In contrast to Figure 5-8, comparing
range after entry to range at entry produces markedly different between winners
and losers. On average, losers shrink their range m.03 and winners expand their
range t.08. (Note the scale shift on the X-axis.)
to shrink their range, if anything. Winners, on the other hand, tend

to expand their range after entry even though they have many instances of range contraction. Both distributions put almost exactly
50% of their occurrences between .l and -.l, the means being pulled
left or right by the tails. If their movement werent favorable, this
would place stops at some risk.
Are these two effects significant enough to affect MAE stops?
To estimate this, compare the expansions to your MAE stop levels. For
this data the MAE stops were generally .31 to .51. At these levels, the
losers average contraction of three ticks and expansions up to, say,
30 ticks would become contractions of 1 to 2 ticks and expansions of
up to 15 ticks if the ranges expand or contract evenly (that is, on both
ends of the range). These are well within the existing MAE stops.
Winners may be at some risk though. As it is, winners can experience occasional range expansions up to 90 ticks. If a winners range
TWEAKS
53
Figure 5-10 Episodic Range Expansion in Winners. When measured from the
range at the point of entry, winning trades in this sample had episodes of range expansion at many times during the trade. This figure compares to Figure 5-7 (Top).
actually contracts, theres no problem but what about the 50% of the
times when it expands?* Does the winners advance offset its range expansion? Recall, too, that range expansion is episodic. Do these episodes
occur near entry when the stop is close or later in the trade? As to this
second question, Figure 5-10, which plots range expansion by day-oftrade for the sample trades, shows that ranges of winners can expand
at any time.
As to the first question, whether ranges expand faster than the
advance of the trade, Ive no evidence of this. Assuming ranges expand on both ends (the daily high and low) equally, an expansion of
.04 (= .08/2) is not threatening to an MAE stop in the .3 to .5 range.
Comparing advances on a winning trade to episodes of range expansion during the trade will typically show the trades profit growth
Actually, in this sample, 53%.

The data for your tradable and trading rules may not show the same result. This is
just an exemplary analysis.
*
54
Figure 5-11
Range Expansion vs. Trade Advance. To see whether range expansion in winning trades is faster than the rate of the winners advance, the two are
plotted for inspection for several trades. Minimum Profit is, for longs, low minus
entry and for shorts, its entry-high. There was no evidence that range expansion
outpaced trade advance generally.
TWEAKS
55
outstripping any range expansion (Figure 5-11) but I know of no way

of summarizing that experience.
Though your analysis might show stronger range expansion, the
evidence in crude oil, for these trading rules, is that range expansion
is a negligible influence. At most, youd widen stops by half of the average range expansion or four ticks. This exemplary analysis and the
worksheets in Appendix H would be the structure youd apply to your
own trading situation.
M A E versus
Range at Entry
Its an intriguing idea: expanding the stops as volatility or range

volatility grow in our tradable. It makes intuitive sense, but you need
to check it out for your tradable. One way to do that is to compare maximum adverse excursion to the range at entry. If winners adverse excursion is large compared to range, there will be grounds for
expanding stops; otherwise, not.
For this example, I compared the 20-day average range on the day
of entry to the maximum adverse excursion experienced during a
trade for 252 trades in crude during 1983 to 1994. Thinking there
might be a quantifiable relationship, I divided MAE (in trading
points) into the 20-day average range (in trading points). The resulting distribution is in Figure 5-12 and shows, for this extensive data,
distinct differences.
Part of this distinction is that already made: Winning trades
dont go far against us, so their adverse excursion will be small and
their ratio to range at entry will be smaller than that of losers. Does
this difference tell us how much, if any, to expand our stops as range
grows? Since losers will hit the MAE stop anyway, lets look at the
distribution for winners
Looking at Figure 5-13, we can see that MAE is heavily concentrated in the ratios below 1.0. That is, MAE is generally much less
than daily range. Average winners MAE for this data was .ll while
the average range was .32. Average ratio was .35. The ratios to the
right of 1.0 on the horizontal axis amount to only seven instances. If,
as a first approximation, half the daily range goes in the direction of
an MAE stop, it would take a ratio of 2.0 (MAE twice as large as
56
range) to suggest expanding stops. Only one trades MAE exceeded 2.0
though.
Throwing out the values of zero (for all the trades with no adverse
excursion) and computing the mean ratio of the remaining trades, the
average ratio turns out to be .52. In other words, the average adverse excursion during a winning trade that has arty adverse excursion is about
half the size of the range at entry, though there are episodes of much
greater adverse excursion, as noted on the figure. Again, even this conservative basis doesnt suggest the need to expand stops.
No Problem?
The third question about the idea of expanding or contracting stops
based on expanding range is whether the information we have already
handles that. After all, the original MAE stops and reversals were
TWEAKS
57
Figure 5-13 Winners MAE vs. Range at Entry. Recalling that the ratio of 1 .O
places MAE and range at equality, winners in crude show barely enough adverse
movement above 1 .O to justify any expansion of stops.
developed without checking for volatility. The values came from experience in all conditions of volatility. Therefore, the first assumption
would be that the impact of volatility is already in the MAE diatrihutions. Sharply distinguished distributions like Figure 5-14 show very
few winners that would be cut off by stops set too tightly. On the other
hand, distributions with extended tails of winners might be candidates
for some adjustments (Figure 5-15).
Just taking Figure 5-15 as an example of the real world, expanding the stop from, say, .45 to 1.05 appears impractical. It would
absorb far more losing trades than winners. Nor is the situation in
Figure 5-15 unusual. Readers of Campaign Trading! will have seen
many distributions where, just as the winning trades tail off, the
number of losing trades picks up and this just as the adverse excursion
is rising steadily, forcing the acceptance of larger losses as the stop is
widened.
To boot, knowing as we now do (for crude) that range expansion
is episodic, if we did observe extraordinary ranges before a trade entry
Figure 5-14 Different Distributions. Expanding an MAE stop from .3 to .45 hecause range volatility had heightened would expose profits to far nwre losing trades
while possibly allowing only four morr winners.
we might suspect that didnt mean much. One episode of widened

range would not mean that more episodes were on the way.
Still, if it is possible to identify periods of extraordinary range
volatility, would that have any impact on the stops to be used? Presumably, youd know that daily ranges exceeded some normal value
and positions taken during such a period would be equipped with
wider stops. Possible sources for such volatility include news, 8888onality, approaching expiration of a trading contract, or general surges
in trading interest.
Normality. Most traders will have an idea of what a normal range is.
Defining it objectively, though, is a little tricky so Ill step through a
process that gives you an experiential idea of normal range. The normal range may turn out to be narrow and well-defined-in which
case you would know if ranges you were seeing were exceptional-or
it may turn out to be quite broad, too broad to effectively modify stop
settings.
TWEAKS
59
Figure 5-15
Mixed Distributions. In contrast to Figure 5-12, experience the sows

winners with large adverse excursions may indicate a tradable whose range volatility could affect MAC stops.
For example, Figure 5-16 shows the distribution of daily average ranges for eleven years of crude trading. Since the range cannot
go below zero, we end up with a distribution which is not normal as
statisticians define normal. Indeed there were seven trading days with
ranges beyond 200 ticks (2.0 on the chart), days I forbore charting. In
a situation like this, what constitutes a normal range and what
would constitute a range out of the ordinary?
Looking at the data in Figure 5-16 slightly differently, Ive plotted the cumulative percentage distribution in Figure 5-17. Nonstatisticians should recall from math class that in a normal distribution,
a mean plus one standard deviation will include 67% of all occurences, all events. Add another standard deviation and youve got 95%
of all occurrences-95% of occurrences being a rough idea of a normal
range of experience. Figure 5-17 connects the range sizes along the
bottom axis with the percent of all occurrences on the vertical axis so
that we can ask what ranges are included in 95% of our experience
with crude ranges.
60
Figure 5-16
Distribution of Daily Ranges. From late 1983 until October, 1994,
crude experienced the ranges shown here. The median for the distribution t.28) is
notsignificantly different from the mean c.32) but there is necessarily a long tail to
the upside that skews the perception of normal range.
Figure 5-17
Cumulative Percentage Distribution of Ranges. The distribution of
ranges for crude is concentrated so sharply that 64% of all occurrences are between zero and .32. the mean of the distribution.
61
TWEAKS
Looking at Figure 5-17, the high concentration of ranges between 20 and 30 ticks causes nearly a full standard deviations worth
of events (64%) to fall at or below the mean at .32. Recalling that
normally 67% or so of all occurrences will be within one standard
deviation of the mean (both above and below), it turns out that 67%
of all occurrences in this distribution fall between .Ol and .34, just
two ticks above the mean range. That is to say, 67% of all ranges will
have a value between .Ol and .34. This is good concentration but if
you wanted normal to include the more common standard of 95% of
events, youd have to include days with ranges up to .75 or twice the
size of the average range. Where in this range do you start widening
stops? Well, if 95% is the normal range of experience, youd consider
widening stops if you had seen daily ranges exceeding 75 ticks C.75
trading points). If, despite the logic above, you want wider stops, the
general rule is: widen stops when youre seeing daily ranges outside
95% of your past experience. How much youd widen the stop would
depend on the range expansion youd measured. In the crude data,
there were 13 average ranges above .75 with no discernible relationship either to MAE or size of win or loss.
Range at
Entry
ProfitOF
Loss
MAE
0.77
1.07
0.87
1.14
1.10
1.07
1.05
0.86
0.82
0.88
0.83
0.71
0.74
-0.32
0.87
-0.87
-1.64
0.77
-2.05
3.72
5.06
-0.70
0.48
-1.81
-1.27
0.15
0.32
1.38
0.87
1.92
0.90
2.05
0.00
0.00
1.14
0.00
1.81
1.67
0.13
62
SUMMARY
Expanding stops, particularly MAE stops, when range volatility rises
is an intuitively appealing idea. Using excursion analysis, you can
check whether it makes sense for your combination of tradable and
trading rules.
Begin by summarizing your experience with the tradables
ranges using the cumulative distribution of ranges as in Figure 5-17.
This will give you a sense of what the nwmal range is. Should you
exceed that, go to step two.
Step two is to check the range expansion for both winners and
losers. In the crude oil example in this chapter, it turned out that the
range of losers actually contracted, almost imperceptibly while that
of winners expanded. In any case of range expansion (Table 5-Z),
widen MAE stops by half the estimated range expansion.
In this example, I was surprised that crude oil showed range expansion up to 25% of its average range. My experience, certainly not
exhaustive, is that range expansion is much more a perceived phenomenon than one thats measurable. Perhaps because range volatility and price volatility are episodic, human perception remembers
most vividly the exceptional instances rather than the general rule.
6
RUNS EFFECTS
CAPITAL
CONSERVATION
When using excursion analysis to develop information about adverse

movement, one of the key benefits is getting a handle on how much
money you need to trade the tradable/trading rule combination youre
considering and whether you should trade it at all. Using adverse excursion information does three things:
1. Eliminates large losses. For the most part; huge gaps in pricing
can still jump over your stops. Options users will have son118 protection even from this event.
2. Tells you how much capital you need to trade the combination.
The combination of tradable, trading rules, and loss management
you use.
3. Tells you where to put your stops on individual trades.
All of these things are crucial and, in retail trading, are often set by
seat-of-the-pants guesstimates. Even commercial trading and money
63
64
RUNS EFFECTS
managers often have barely any defensible reasons for their losscutting tactics.
Using adverse excursion information, you can manage individual trades effectively and build up the ability to trade a variety of
market modes, such as trending or ranging, and a variety of tactics,
such as add-on trades, counter-trades, and reversals. The result is you
are trading far more actively than you would at any one time in a single trading system.
Why is this important? What could happen is that, despite controlling individual losses, you could end up with a string of trades that
impact your capital so severely as to stop your trading. When this happens, there is no chance of recovery and youve suffered the catastrophic
,106s that MAE analysis was designed to prevent in the first place.
In terms of trading management, a string of consecutive losses
(a run of losses) has a direct impact on any individual trading tactic and on two management plans, the actual trading campaign itself
and a subsidiary tactic which is not necessarily being used: a betting
strategy.
IMPACT ON A PARTICULAR TRADING TACTIC

The most discernible impact of a run of losses is the drawdown in account equity while using one particular trading tactic. Since one tactic is all that most retail traders and many commercial traders use,
their attention is quickly brought to any string of losses. This discouraging event causes many to abandon systematic trading rules that
have tested well and turn out well again, despite the run of losses.
Experience is the best teacher for each tactic and most trading
report summaries include an accumulation of drawdown (equity reduction, whether measured on an open-and-closed trades or closedtrades-only basis). Its impossible to say if traders using MAE are
more likely or less likely to have trades hit by stops than traders
using other means, so their susceptibility to runs is probably no different than others. MAE users can look at drawdown stats from standard software (using MAE stops) as a rough measure of their capital
exposure. However, these worst cases arent the best measure. You
65
need to measure the drawdowns and inspect their distribution for

information.
Having looked at those losses, you must decide if they are actionable and, if so, at what point. In other words, you cut off an individual trades losses at 2% of trading capital but do you cut off runs
losses? This depends somewhat on the frequency of runs generated by
a trading combination. If your trading rules were such that you experienced lots of losses and only a few gains, the chances for a run of
losers would be higher than a combination that produced mostly winners. Thus, to make the estimate, you must examine the record. The
details of the example that follows are developed in Appendix I.
Frequency of Runs
The first thing is to define a drawdown and see if it exists as defined.

By graphing account equity from a single trading tactic as in Figure
6-l you can quickly see where prolonged periods of equity drawdown
occurred. Figure 6-1 is the account equity from a single trading
Figure 6-1 Account Equity from Trend Entry Tactic. Lots of small losses and a
few huge wins typify trend trading. Herr MAE stops have served to keep losses
small but there are long series of losses to be endured.
66
RUNS
EFFECTS
Figure 6-2 Equity Ketracement

Levels. Levels A through G are all retrac~ment
levels caused by runs of losses, what is usually called drawdown. Adverse runs
,size is measured from the previous peak in equity.
tactic: entry on a dual moving average system using crude oil as the
trading vehicle. The stop used was .31 or 31 ticks, about $310, ignoring commissions. The result is typical for trending systems: a few
incredible gains and lots of small losses. Its almost ideal for the
study of adverse runs.
Looking at the periods of equity drawdown, most losses are
roughly offset by small gains. However, in periods 700 to 900 as well as
later periods around 2000, 2500, and 2900 there were extended periods of losses that slowly wore away account equity. You might pick out
different periods that I should have included and that brings up the
issue of definition. Whats to be a run and whats to be a significant
run against which we should guard?
Visually inspecting charts like Figures 6-2 and 6-3 repeatedly
suggests one practical answer: An adverse run is any sequence of
losses or gains (sic) resulting in equity falling below a previous peak
and continuing until a new high in. equity is attained. To measure an
* This system was described in detail in CampaLgrz
Trading!
67
Figure 6-3
Drawdown Compared to Capital Plus Winnings. A run of lk~sses builds
drawdown up to 8% of initial trading capital plus winnings. Comparing equity reduction to capital plus winnings is only correct on a one contract or no additional shares basis. For a mote conservative
view, see Figure h-4. Points A through
C are those referenced in Figure 6-2.
adverse run, simply keep a tally of the previous peak equity and the
current account equity, computing the percentage reduction on a daily
basis. Plotting the result makes judgment of the level of significance
and labeling the charts by inspection easier (Figure 6-3).
Figure 6-3 plots the percentage reduction from peak equity plus
initial trading capital of the trading combinations various adverse
runs. Pictured in this fashion, drawdown from adverse runs may show
some consistency, consistency you could use to as8e.w at what level to
become concerned about a string of trading losses from a specific trading tactic. At a minimum, each peak in Figure 6-3 defines a drawdown in Figure 6-2.
In this figure, material peaks are those of 4% or greater equity
reductions subsequent to new highs in equity. That is, a new high in
equity ends the adverse run and that new high appears in this figure
as a spot on the X-axis with no columns above it. Three peaks (1, 2,
and 3) show drops below the 4% level before going on to their ultimate
68
RUNS
EFFECTS
high but arent noted because the drawdown had not yet ended. In
other words, equity continued to decline after that interim peak was
established, despite .a temporary reduction in drawdown.
Figures like this make assessments of significance easier. However, if you want a completely objective standard, go with the original
definition and count all the peaks interrupted by periods of higher
profitability. Doing this highlights that, in a tractable trading combination, most of the drawdowns from adverse runs of trades will be
fairly small.
Computation
The precise computation of Percent Reduction in Trading Capital
Plus Winnings should be clear from Figure 6-3. Since we know the
MAE stop ($310) and we know our 2% rule for the MAE stop, we can
compute the trading capital required to support this trading combination. That is:
$3101.02 = 515,500
If you were trading two or more contracts or blocks of shares at the
$310 stop, your available trading capital would be, for example:
2 x $310/.02 = 531,000
It is to $15,500 plus any accumulated equity that Figure 6-3 compares the drawdown experienced by this trading combination:
(Maximum equity ~ Todays equity)/($15,500
+ Maximum equity)
Figure 6-3 certainly shows some consistency. In eleven years of

trading this combination, seven major drawdowns occurred. Though
too few to be statistically significant, these occurrences may give a
rough idea of the frequency of adverse runs. Size of runs is also
roughly calculable: between 4% and 8% of capital plus winnings before
going on to new highs in profitability. These adverse run effects
should be bearable for most well capitalized traders but there is a caution: results are dependent on when the run occurred. Had the largest
69
adverse run started on day one of the campaign, thered be a peak of

percentage reduction of:
(42.78 - 38.87 1115.5 = 25%
instead of 6.5% at point G. This argues that, instead of comparing the
adverse run to the trading tactics equity plus the initial trading capital, it should be compared to the trading capital alone.
A More Conservative Standard
A winning trader has the luxury of banking the winnings and continuing to trade on the initial capital committed to the trading tactic.
Alternatively, he can hold the winnings against future, probably inevitable, storms of adversity. Or, he can up his commitment to the particular tactic by increasing the shares or contracts traded, relying on
his analyses of the MAE stop and runs to protect him. I advocate the
third path. For those who pursue the first or second path, drawdowns
should be compared to initial trading capital which is the subject of
Figure 6-4.
Figure 6-4 shows more serious problems with adverse runs in
this experience with crude. The 25% figure mentioned earlier is there
to the right, but so are regular stabs at the 15% level. Larger adversities cannot be ruled out either. Indeed, in this worst case outlook,
adversities amounting to 5% have more than a 50% chance of reaching 10% to 15%. Still, none of them amount to the 40% reduction in
trading capital which is the most popular rule of thumb for suspending trading. If winnings are considered as well (Figure 6-3), none of
the drawdowns remotely approaches 40%. That is, despite a very good
likelihood of seeing an adverse run of 10% to 15% of initial trading
capital, this combinations experience is that losses from such runs
will be recovered and equity will move to new highs before seeing another adverse run of similar or greater magnitude.
/mmediate Disaster
Finally, even eleven years experience is only suggestive, not exhaustive of all possibilities. A single 40% reduction could happen right off
70
RUNS EFFECTS
Figure 6-4
Drawdowns of Initial Trading Capital. A more arduous comparison

than that 01 Figure b-3 is to compare the drawdown
from adverse runs of trades
with initial trading capital, ignoring any winnings. On this basis, the experience
with crude shows the possibility of 25% drawdown
though most are in the 10%
t 15% range.
the bat, putting you out of business. However, this small sample of experience suggests the possibility of a huge adverse run is small. There
is no instance of 40% drawdown before returning to higher profitability. The probability of even being in one of the samples 200-day
drawdown states is about 45% and, once that hurdle is surmounted,
the possibility of its being a 25% drawdown is one in seventeen. That
works out to less than 3% chance at any one time for the worst case on
day one. Actual probabilities may be even lower as shown next.
The likely extent of drawdown is suggested by the distribution of
occurrences of drawdowns in the historical data as shown in Table 6-l
and Figure 6-5.
Looking at Figure 6-5, its apparent that most drawdowns are
small, even in relation to initial trading capital (vs. trading capital
plus trading profits) and 70% are 10% or less. If you are looking at this
kind of concentration, its reassuring evidence that your trading combination is workable. That 94% of the time its drawdowns dont exceed
18% is also workable. Plus, its worth keeping in mind that drawdown
71
Table 6-l Frequency of Size of Drawdowns. From the sample

data that produced Figures 6-l to 6-3, this summary of every
drawdown between periods of rising profitability was prepared.
This sort of table gives you the relationship between the size of
the drawdowns in your data and their probability, here summarized in Figure 6-5.
Trading
Cspital
0%
2
4
6
8
10
12
14
16
18
20
22
24
26
OOClllY?IlC.%
0
3
5
3
1
0
1
1
1
1
0
0
0
1
Probability
0%
18
29
18
6
0
6
6
6
fi
0
0
0
6
Cumulative
Probability
0
0.18
0.47
0.65
0.7,
0.71
0.76
0.82
0.88
0.94
0.94
0.94
0.94
1
is defined here as the entire period between a peak in trading equity

and the next new high in equity. That means that even drawdowns
closely spaced together were separated by new highs in profits from
the trading combination. That is, the equity curve was still moving
up. If there were no break in the drawdown, youd have a trading combination that was a steady loser or a loser despite occasional wins, and
drawdowns would be continuous, not episodic.
Actually, given the probability of being in a drawdown (that is,
below the most recent peak in trading profits) almost half the time
(46% of the time) and the size/probability relationship in Figure 6-5,
whats the realistic probability of starting off with a disastrous sequence? I estimate that with displays like Figure 6-6.
Figure 6-6 shows the actual experience with this trading combination in the white columns. On the right is the 25% drawdown we first
noticed in Figure 6-4. Given that one occurrence, I earlier calculated
.45 X .06 = .03 or 3% for running into a disaster on day one. However,
72
RUNS EFFECTS
Figure 6-5,
Probability of Drawdown. The fourth column in Table 6-l translates

into this picture of the relationship between probability and size of drawdown
from
adverse runs. This limited experience with the exemplary trading combination suggests that 70% of the drawdowns will be 10% or less before the combination
reaches new highs in profitability.
its apparent from the figure, that there are fewer and fewer drawdowns
as we go out further on the x-axis. To estimate the likelihood of a killer
drawdown, I fitted through these actual events an exponential (or
growth) curve as a proxy for the actual distribution we cannot know.
If we read from it the estimated probabilities, the expected probability
for the 25% drawdown is vanishingly small. The probability of a 40%
drawdown is surely infinitesimal. This estimate leaves us little excuse
to avoid using this particular trading combination.
By displaying the drawdowns experienced in your system testing
and applying this straightforward inspection of the results, you can
make reasonable estimates of the likelihood of getting hammered right
out of the gate.
Significance
The exemplary data used here are of a worst case. This is a trading
system with only 30% winners so lots of time is spent in drawdown
Percent
Drawdown
of
Initial
73
Capital
Figure 6-6 Actual and Expected Probability of Drawdown. An exponential curve

fired to the declining frequency of drawdowns suggesrs vdnishingly small probabilities for disastrous 40% drawdowns.
periods. The last analysis assumed that, even though the trading combination was profitable, none of the winnings would be used to support
future trading; all drawdowns would be from the initial trading capital. Moreover, drawdowns themselves were (and should be) defined
comprehensively and conservatively as the worst reduction in trading
equity before a new high in trading equity is set.
Nevertheless, so effective was the use of MAE stops to minimize
losses that even lots of losses did not prevent the trader from being
around when large winners showed up. Also, the MAE stops prevented
any disastrous single loss that would have destroyed the trader.
The loss control from MAE stops also minimized the impact of
adverse runs of losses. Figure 6-5 summarizes the actual experience
of adverse runs showing how unlikely they would be to break the
traders bank. The significance of losses from adverse runs depends
on their distribution. When experience indicates, for instance, that
74
RUNS EFFECTS
(1) most adverse runs will cause losses of less than lo%, (2) 95% of
the losses will be less than 18% of initial trading capital, and (3) the
probability of a trade-stopping 40% drawdown on the most conservative basis is extremely low, heart should be taken and the trading plan
executed faithfully.
If examination shows that the probability of hitting the 40%
drawdown is high, its time for revisions. However, that doesnt seem
to be common with MAE trading. Consider the probabilities that must
occur with even so dismal a system as my 30%.winner example. Using
its most common drawdown of 4% (from Figure 6-6) and assuming
there were no interval of high profits between drawdowns,* it would
take ten of these in a row to hit 40% total drawdown. This event would
have a probability of .3 I0 = .0000059. A 10% drawdown, using the
same assumptions, would have, at the worst, a .O@ = .OOOOl probability. In fact, the evidence collected here shows that lots of small, consecutive losses result in drawdowns that are generally manageable,
frequently sizable, and never catastrophic.
Your historical testing may have quite different results and you
may set different limits. A 3% chance, a chance that may actually be
much lower, of a 25% drawdown moderated by MAE stops is manageable.

Campaigning is the overall plan by which the trading manager approaches the market. A campaign is made up of strategies and tactics.
For example, strategies could include trading trending and ranging
market modes. Tactics could include entry trades, add-on trades, countertrading, and reversal trades. Each one is separately capitalized
based on their expected loss on individual trades. Normally, I have advocated risking no more than 2% of total trading capital on any one
* This has the effect of turning drawdowns as defined here into independent events
so that it would never occur that a drawdawn would never be succeeded by another
drawdown without an interval of higher profitability. Thus, the fear of succeeding
drawdowns is met in these statistics, unless your trading equity actually is beaded
consistently downward and there is no interval of higher profitability.
75
trade but, if, for example, add-on trades during a trend had progressed
well you might have on at one time:
1. The basic trend entry trade
2. Two or three additional trend add-on trades
3. An add-on day trade. Since this is a day trade, you might avoid
your broker seeing a margin requirement but you still have the
risk of loss and must account for it by an absorption of capital.
4. A countertrend trade
Ive even seen a situation where a well-developed trend triggered
a (brief) ranging trade without ending the trades associated with the
trend. In that case, you could add a fifth exposed trade.
In such a situation, you will have some comfort that your trend
trades have progressed well (since you now have several of them that
were triggered by earlier successes) and add-on trades usually have a
very high percentage of success. Since the returns to any individual
tactic are largely independent of those of another tactic, diversification
effects usually work in your favor by offsetting losses in one tactic
with wins in another.
Nevertheless, correlation between tactics should be checked where
its conceivable and it is possible that several could go bad coincidentally. Though you could check past experience on this one, the correlation of trading system/tradable combinations is usually pretty sporadic.
Where you do find relationships are those well-established among stock
industries, groups and sectors; indices; and futures related to underlying economic sectors (the rate complex, currencies, indices of many
kinds).
What you must check is whether, once a trading systems entries
and exits are used on a specific pair of tradables, the correlation
events survives and, after trade management (i.e., stops), any correlation of returns and drawdown experiences survives. If the trading
systems for the two (or more) related tradables are different, correlation of returns and drawdowns is even more problematic.
Moreover, using standard statistics for this comparison is inadequate. Since, in statistical terms, gains and losses occur episodically,
correlations may be very low when measured in the usual way. Its
76
RUNS EFFECTS
Figure 6-7
Equity for Two Trading Tactics. Since drawdowns and advances occur
episodically during trading campaigns, visual comparisons are more to the point in
rearching for relationships between those events in two different trading tactics.
Here, the equity curve for simple the trend trading of Figure 6-l is compared to
that for additional day-trades taken during trend trades.
better to inspect the behavior of the equity curve visually, comparing

the fluctuations of two combinations directly for sympathetic movement. Finding such movement, you then have the problem of determining if there exists a relationship and, if so, whether the
relationship is likely to be in force in the future.
For example, Figure 6-7 compares the equity lines for two different trading tactics. The Trend Equity line is the trend trading
first seen in Figure 6-l; the second is that for add-on day trades*
taken during the first periods trend trades. That the add-on trades
would be related to the trend trades is built in to their rationale, yet
looking at the two lines suggests that there is not necessarily a close
relationship. In fact, the correlation between the two is .26.
* The trading rule, outlined in Campaign Trading!, applies hut when, during a long
trend trade, the low of the day is below the 20.day average, closing the position on the
close of tho day. If short, sell if the high exceeds the 20.day average and buy it hack
on the close of the same day.
77
The episodic nature of the relationship of two trading combinations plays out well here. For roughly half of the 11 years shown here,
the add-on trades did not do well while the trend trading tactic advanced steadily in equity. Add-ons did not participate in the single
large gain around the 700th trade day nor did they participate in the
trend trades gains from many small trades taken during the first six
years. Then, in 1989, around the 1600th day, the tactic started to
work well with very few drawdowns. During that same period, the
trend equity line did advance, but had several periods of significant
drawdown.
Experiences like these-where the tradable is the same and
the trading tactics somewhat related-generally produce very little
relationship in equity impacts. I am tempted to say never because I
havent seen a solid relationship, but they still are possible. Nevertheless, I dont believe this concern is worth more than this visual inspection. By the time you have differences in the tradables, differences in
the trading tactics, and differences in the loss management techniques,
you have very little relationship between the equity curves. I certainly
cant prove they dont exist, but I couldnt find one to show you!
Correlated Equity Curves
Where you did find those strings of losses from different trading combinations coincided regularly, you deal with the issue by raising the
capital available, lowering the amount of loss acceptable on trades
(tough to do if youve selected the right MAE stop level to start with),
reducing the number of contracts used by each tactic (only possible if
your MAE analysis allows you to trade more than one contract within
the 2% of capital limit), or ceasing one of the two related tactics.
The best example of this is trading the same tactic on highly related tradables, for example, the DMark and the Swiss Franc. The
correlation in monthly returns between these two using, as an example,
the Donchian Rule, was reported to be on the order of .77 in 1993.
(Kestner, Lam. The Role of Diversification, Techrzical Analysis of
Stocks and Commodities, March, 1996.) in 1993. The simple thing to do
is trade the more liquid issue in greater size since youre getting the
same movement in both and diversification benefits are minimized if
you trade two vehicles with highly correlated equity curves.
78
RUNS
EFFECTS
Day in Trading Campaign

Figure 6-8 Trading Related Tradahles. When the DMark and Swiss are traded
with the same trading system, the numbers come cut slightly different but the relationship between the equity curves is apparent. Trading highly correlated vehi&
cles with the same tactics robs the trader of the benefits of diversification.
Take Figure 6-8 as an example. Here, both the Swiss Franc and
Deutschemark contracts are traded with the same dual moving average system used in Figure 6-1, albeit with 12- and 20-day parameters for the averages. The Swiss contract came out less favorably, but
both advances and declines were shared coincidentally. Indeed, the
correlation of the two equity curves is an astoundingly high .94. Here
is clearly a case where trading both doubles the downside.
Kestner reports the correlation of returns for Deutschemark and
Japanese Yen at .38 (Figure 6-9) and that for Deutschemark and Gold
at -.Ol (Figure 6-101, roughly half that of the Swiss Franc. The equivalent figures for the equity of the dual moving average system are 59
for the DM/JY and -.27 for DM/Gold. The DM/Y relation is clearly
less sympathetic than DM/Swiss and both head in the same general
upward direction.
79
Drawdowns seem barely related. Looking at DM/Gold, its hard to

see any relationship unless, as the correlation suggests, it is negative.
Combining, for purposes of exposition, the several curves gives
Figure 6-11 which, in turn, gives us the drawdown chart in Figure
6-12. This sobering presentation shows, for exposition purposes
only, drawdown effects that would certainly get any traders attention. Though the general direction is upward, golds huge fluctuations impose drawdowns on the total that are relatively large,
generally diminishing the normal diversification smoothing of returns. To boot, golds losses in the latter part of the period drag
down the overall results of the portfolio, only slightly offset by the
Yens gains.
* The more proper procedure would be to adjust the ratios of e~ntrxt~ to achieve a
consistent dollar ex~wure between all four combinations.
RUNS EFFECTS
Day
of Trading Campaign
Figure 6-10
Deutschemark and Gold. Trading gold with the same system as in

Figures 6-8 and 6-g actually generates a negative relationship, -.27. Trading a
pair such as this could potentially smooth overall returns ii both had positive results. (The two curves have been meanxentered for exposition.)
A look at Figure 6-12 to see what our expectation of disaster is

confirms that drawdown is frequent and steadily runs in the 10% to
15% range. Gains, while strong, are short-lived and most of the trading
period is spent in drawdown from previous peaks. Some of the drawdown periods last over a year and a half as hard-won equity is worn
away by various combinations individual losses. Though the chance
of hitting a 40% overall reduction in the capital committed to all
four trading combinations appears low, drawdown periods happen quite
frequently.
Since each trading combination is separately capitalized and
none individually come close to a trading cutoff (though Gold is working on it), the traders consolation is that overall gains for the Portfolio of four are greater than trading any one issue alone. At the close
81
Figure 6-l 1 Combining Trading Combinations. Melding the results of the figures
above for Drutschemark, Swiss Franc, Yen, and Cold generates this jagged curve.
Though the direction is upward, drawdowns seem to be serious. For a better look
at those, see Figure 6-l 2.
Figure 6-12 Drawdowns from Four Combinations. Somewhat easier to see are
the drawdowns, presented here as a percentage reduction from peak equity plus
initial trading capital.
82
RUNS EFFECTS
Table 8-2 Variability of Returns. DMark, Swiss, Yen, and the total portfolio achieve a K-ratio > 1 but fail to make the grade when the risk-free interest rate is included via the Sharpe ratio. Only the Yen, with its
last-minute surge, achieves profitability in excess of the risk free rate.
DMark
K-ratio
Std dev./avg.
Shsrpe
ratio
Swiss
2.13
0.48
1.43
0.51
-2.41
-3.22
Yen
1.42
0.50
-1.19
Gold
-0.51
0.50
-11.74
Total
Portfolio
1.60
0.31
-0.90
of the period, DMarks are up $23,500, Swiss $15,500, Yen $54,900,

and Gold $0 (Table 6-2).
Next you check the variability in relation to return with the
K-ratio and the Sharpe ratio which are in Table 6-2.* The K-ratio
test favors a steady advance of the equity curve and the author, Lars
Kestner, recommends a minimum value of one for a single trading combination. Save gold, all achieve this but the group does not achieve the
higher hurdle of 3.0 for a portfolio.
Also, despite overall profitability, these trading combinations are
generally disappointing when faced with the job of earning a return in
excess of the risk free rate. Variability of the returns in terms of the
average return is fairly low and reduced further in the total portfolio,
surprising given the steep drawdowns seen in the equity curve but
consistent with the larger total profitability. Its apparent that all the
loss management in the world cant make a humdrum system sufficiently profitable even if it can prevent disasters.
* Kestner, Lam, Measuring System Performance, Technical Analysis of Stocks and

Commodities (March, 1996) pp. 46-50. The K-ratio compares the slope of a regression
line of the combinations equity curve to its standard error to reward elosenes~ of fit
to the line of advance.
K-ratio = Slope of regression / [(Standard error of slope)

(Number of observations)*1
Sharpe, William, Mutual Fund Performance,Journal
ofBusiness (January, 1966L

pp. 119-138. The Sharpe ratio is the excem return divided by the standard deviation
of return.
Sharpe ratio = (Return Risk free rate)/Standard
deviation of return
IMPACT ON BETTING STRATEGIES
83
The bottom line for this analysis:

1. The tiny portfolio is susceptible to drawdowns somewhat higher
than the 4% to 8% seen for DMarks alone but nothing so high as
25%.
2. The potential for disastrous 40% reductions in equity appears
minimal.
3. Diversification benefits for this oddly-correlated group (Swiss
and DMarks highly correlated, Yen somewhat correlated, Gold
negatively correlated) appear only in the reduction of worst case
drawdowns of equity from the 25% level to the 10% level.
4. Profitability vs. the risk free rate is inadequate so its must be
inadequate in comparison to other risky strategies (say, a stock
index fund).
IMPACT
ON
BETTING
STRATEGIES
Betting strategies alter the size of the amount traded based on, commonly, the likelihood of success, pyramiding, the progression of
betting, and the amount of capital available. Of all these, the last
two have some theoretical and experiential support. Increasing or decreasing the number of shares or contracts based on capital available
is straightforward: the size of the play for an individual trading tactic is simply limited by the 2% rule, the loss being defined by an MAE
stop. As your capital grows (hopefully), 2% of it grows apace and the
size of your commitments grows with it.
Its the use of progressions thats more difficult. An individual
trading tactic certainly produces a series of wins and losses (outcomes) that resembles a betting series and a properly prepared progression will eventually win. The trouble with progressions is that a
run of losses can extend the trading beyond your capability to play.
In Chapter 7, I give some examples (and references) for using progressions in trading and show how your experience with runs of losses
can impact this particular betting strategy.
84
RUNS EFFECTS
SUMMARY
Practical inspection of the day-to-day performance of individual trading combinations and that of the group of combinations as a whole will
give a good indication of the potential for disastrous hits. Asset allocation schemes can be used by quantitatively-oriented trading management to objectively assemble portfolios of trading combinations (or
traders, for that matter) but those after a graphic depiction of their potential to lose may use the approach outlined here. Repeated experience, as shown in the graphics, may lend some intuitive understanding
of the nature of the drawdowns faced day-to-day that a smoothly assembled efficient frontier does not.
7
MARTINGALES
If the runs information means anything, it means there will be adverse runs. Runs of winners are fine, but well be faced with losers
certainly, especially as Ive defined runs for trading purposes in this
book. If you recall that definition is that an adverse run lasts until a
new high in equity is achieved, even if there are some intervening
wins. Is there some means of taking advantage of this situation?
Given enough money, almost anything can be overcome, even adverse runs. With adequate capital (i.e., LOTS of capital), you can adjust the size of your bets to suit a variety of win/loss sequences. An
extraordinary amount of capital is needed because you dont know how
long the adversity will persist. The trick is estimating that length and
size probabilistically and arranging your capital to withstand it. Fortunately, with MAE stops, loss size is controlled and the major remaining question is how long adversity will go on. In the previous
section, we went over the estimation of adversity from runs by examining the experience both for a single trading combination and for a
group of trading combinations.
85
86
MARTINGALES
There is another tactic though. Ferguson, Eliason, and Pelletier* have dealt with the subject of handling a series of trades
through betting strategies. The technique, a martingale, adjusts
the bet size by fixed rules following a win or loss. In Fergusons first
article, he demonstrated the positive impact this approach can
have on the typical trend following trading system, turning it from
a loser to a winner. The approach was so apt I applied it to the humdrum system discussed in this book. That system is a small winner
in DMarks, Swiss, and Yen, but a loser in gold (or, at best, a small
winner).
Martingales are best suited for professional traders and trading
organizations who are not only well-capitalized but also thoroughly
disciplined. It can be a disastrous strategy from which to exit in midstream so, given that quitting a strategy prematurely is common
among individual traders, its not something unseasoned traders
should undertake.
Moreover, if applied to a series of trades that take place over two
,0r three years, even a institutional trader may have difficulty implementing it simply because normal turnover on the trading floor will
mean he or she wont be there in two or three years. In a wellorganized and disciplined institution with adequate position tracking
facilities, run control or loss management techniques can be sue
cessfully implemented over longer periods of time. Even there, though,
it helps if the trading combination used trades rather frequently so
that traders and management can see the technique workin,&! and get
away from the notion that they are sitting on a loss which can only be
requited at some distant time in the future.
One factor in trading that favors a martingale is that winning
trades are normally larger than losing trades. Although this isnt the
* Ferguson,
James W., Martingales, Stocks & Commodities, V. 852 (Seattle, 1938)
pp. 66-59, and Reverse Martingales V. 8:3 (Seattle, 19881 pp. 105-108.
Eliason, Peter Tactical Stock Trading, Stocks & Commodities, V. 7:3 (Seattle,
1988) pp. 69-72.
Pelletier, Robert, Martingale Money Management, Stocks & Commodities,
. 63 @eattIe, 1966) pp. 266-267.
87
SIMPLE MARTINGALE
case for every instance, traders normally pick systems where a 2 : 1 or

3: 1 ratio is maintained between the size of a win and the size of a
loss. Were this not so, trading a martingale would be very tedious so
you may go through ten trades only to eventually win an amount equal
to your average win.
Lastly, transactions costs are a big factor in using martingales.
Unlike casinos, exchanges and brokers charge for their services irrespective of outcomes. Transactions costs must be very low to keep
them from being a material detriment to the ultimate outcome. All
these factors play a part in the exemplary analysis shown in this
chapter.
Before going into the simulation using the crude data on which
this book focuses, heres a brief description of simple and complex martingales for those who arent familiar with them. The origins of the
name seem lost in time. It seems certain though that, contrary to legend, there was never anyone named Martingale. Ferguson felt the
term originated from the French describing an apparatus used to keep
a horse from swinging its head-a reference to the use of martingales
to check losses.
SIMPLE
MARTINGALE
A simple martingale consists of this rule: Double your bet (at even
odds) after each loss until you win. When you finally win, your winnings will amount to your original bet. In a trading situation, that
means that your expected win is the same size as your expected loss.
For example, check this sequence:
Trade
Bet
Win or
LOSS
Equity
1
2
3
4
$1,000
2,000
4,000
8,000
LOSS
LOSS
LOSS
Win
$(l,OOO)
(3,000)
(7,000)
1,000
88
MARTINGALES
Try a few sequences of your own. The bet size (the number of
shares or contracts) increases rapidly (from $8,000 to $16,000 to
$32,000, to $64,000 and upward) so that you run the chance of running out of money. MAE stops help with this by limiting the size of the
loss which, in trading terms, is really the size of your bet. If the same
sequence as above were played out with the stop of $310, things would
be more manageable for most traders, though the commitment rises
faster than most would like. Again. its assumed the win will be $310
just as will be the loss:
Win or
Trade
1
2
3
4
Bet
LOSS
Equity
310
620*
1,240
2,480
LOSS
LOSS
LOSS
Win
(310)
(930)
(2,170)
310
* Two contracts with a stop at 31 ticks from entry.
Whether this would be economic tactic depends on the frequency

of the trading combinations trading, the transactions costs (15 trades
at $19 costs costs $2851, the length of time it takes for the sequence of
trades to play out, on average, the returns on alternative trading techniques, and tax rates.
COMPLEX
MARTINGALE
As easy as the simple martingale is to understand, the complex martingales are not. There are many varieties of complex martingales but
all have one feature: They reduce the size of the additional bets that
must be made at the cost of extending the time you spend in the betting sequence.
To see a complex martingale (again, just one of many possible)
working consider what would happen if, in the above sequence, you
bet not double the previous bet but only one. unit more or less:
89
COMPLEX MARTINGALE
Trade
Bet
1
2
3
4
5
$ 310
620*
930
1,240
930
Win or
Loss
Loss/Gain
$ (310)
(620)
LOSS
Loss
LOSS
(930)
Win
Win
1,240
930
Equity
$
(310)
(930)
(1,860)
(620)
310
* Two contracts. Share traders could adjust the number of

shares, commksion costs allowing.
The result was the same but it took longer to get it. Again, you
should try a few of these sequences yourself. Just for a taste of a eomplex martingale the rules for the number of contracts to trade were,
after the first loss, to add the number of contracts traded in the loss
to the sequence of trades, sum the first and last number in the sequence, and trade that number next. If a trade is a winner, strike the
first and last numbers in the sequence and sum the remaining first
and last numbers to get the number to trade next. For example:
sequence
Number to
Trade
Win or
Trade
1
2
3
4
5
1
1,1*
1,1*,2
1,1*,2,3
1*,2
1
1+1=2
1+2=3
1+3=4
1+2=3
LOSS
LOSS
LOSS
Loss
Win
Win
* Marks the second 1 in the series.
You can readily imagine with all the combinations of wins and
losses the market can throw at you how complicated a series could be
generated. To really get a practical feel for this, you should play with
the martingale simulation given in Appendix J. Youll quickly see
that, even though martingales cc~me right in the end, youll not enjoy
a situation where the win size is equal to the loss size. Figure 7-l is
90
MARTINGALES
Figure 7-l
Distribution of Win Size. Computed for the trend following dual moving average system used throughout this book, the distribution highlights that selecting an appropriate wins size for martingale modeling is subject to wide latitude.
an example from Appendix J where you can also read an explanation

of the columns and their computations in Table 7-l.
Hitting six losses in a row is not that frequent but when it happens youll need a while to recover. In this simulation, the martingale
gradually wore away the $6,510 loss that had been built up at the start
of the series but it took it eleven more trades to get back to just $310
in winnings. Many traders who have prayed fervently to get back to
breakeven might think this a good result but would they have stayed
the course through 17 trades?
In Table 7-1, P(Win) = .5 and the ratio of the win size to the
loss size is just 1: 1. Given that you have a higher win ratio or a better win:loss ratio, things can be much better despite adverse runs
of losses. Table 7-2 shows an adverse run with P(win) = .4 and
wins:losses = 3.0.
I leave it to you to experiment with the model in the appendix
but it turns out that selecting a system with the right dollar win:loss
ratio is most important to managing a martingale, especially for a
trading combination that doesnt have a good probability of winning,
as most dont. Of the things you can influence in designing and soletting your trading combination, you may have little control over the
COMPLEX MARTINGALE
91
Table
7-1
Long Drawdown. Though martingales can always bring
one back to new highs in profitability, a truly adverse run of trades
absorbs both time and capital.
PCWIN)
$Win / $ Loss
0.5
A s s u m p t i o n s : W i n o=
Trade
Bet
LOSS
Contracts
Loss/Gain
LOSS
- 1
LO88
Loss
-2
2
3
LOSS
LOSS
LOSS
4
5
6
7
8
9
10
11
12
13
Win
Win
LOSS
Win
Win
LOSS
Win
LOSS
6
3
14
1s
16
17
LOSS
Win
Win
F&t!
False
F&e
18
19
20
Bet Size
$310
Equity Units
Equity
- 1
-3
-6
-3
-4
-5
-6
7
6
-6
6
6
-3
4
-2
-3
4
3
0
0
0
-10
-15
-21
-14
-8
-14
-8
-2
-5
-1
-3
-6
-2
1
1
1
1
Bet Table
(310)
- 4 5
IO
(930)
-36
-28
-27
-21
-20
-19
9
8
8
7
(1,860)
(3,100)
(4,650)
(6,510)
(4,340)
(2,480)
(4,340)
(2,480)
(620)
(1,550)
(310)
(930)
(1,860)
(620)
310
310
310
310
-15
-14
-11
-10
-9
-8
-6
-5
-4
-3
-2
-1
0
10
8
6
6
8
5
5
6
4
4
5
3
3
2
1
Table 7-2
Adverse Runs. A good win size compared to the loss size (the
MAE stop) can overcome a multitude of bad trades, even an adverse win probability. A martingale series is terminated when the equity after a trade
turns positive.
PCWIN)
0.4
Assumptions:
Trade
Bet
2
3
4
5
2
3
4
5
$ Win / $ Loss
3
Contracts
Loss/Gain
LOSS
- 1
- 1
5 (310)
LOSS
-2
-3
-4
15
-3
-6
-10
5
(930)
(1,860)
(3,100)
1.550
LOSS
LOSS
Win
Equity
Units
Bet Size
$310
Win or
Loss
Equity
MARTINGALES
92
ratio of wins and losses without imposing too many rules and overfitting. Small adjustments, however, often effect the size of you wins
and losses and, in MAE, you have an excellent tool for managing the
size of your losses.
Managing the size of the wins is tougher. The average win figure
is not necessarily a good one as the size of the wins is usually quite
skewed (see Figure 7-l).
Nor is the distribution of loss size normal. See Figure 7-2.
Thus, it isnt idle to inspect closely the distribution of the size of
wins and losses. When you actually trade, remember, you wont get
average wins or losses but values all over the ballpark, usually within
expected ranges. Thus, you must plan to go through many martingales
to come up with the average win and average loss, assuming prior experience holds true. Some experimenting with the models will give
you a feel for the levers available to you but the most common result
of low frequency trading is to cut things too finely: without a lot of
trades, the win size and loss size tend toward the median rather than
the mean. As a result, the ratio of win size to loss size moves closer to
1: 1 from the ideal 2: 1 or 3: 1. In the example above, using means we
Median = ~2,
Mean =
N = 148
Sire of Losses (Hundreds of Trading Points)

Figure 7-2
Distribution of Loss Sire. lust as wins arent normally distributed,
neither are losses. Consider, though, the combined distributions about the
breakeven point are near to normal with the exception of the winners long tail to
the right. Losses can be controlled with well-defined MAE stops.
MARTINGALES ON CRUDE OIL
93
have .76/.34 = 2 while if medians are used we have .36/.24 = 1.5. The
practical result is that, if youre going to use martingales you need to
use them for a lot of trades to get sufficient to reach your desired winsize, loss-size ratio.
In summary, martingales should be used with a trading tactic
that trades frequently (add-on day trades come to mind), where the
probability of profit is 40% or better and/or the ratio of win size to loss
size is better than 2: 1, to pick rough rules of thumb. These factors
shorten the length of the martingale which will cut down the number
of trades before conclusion which, in turn, reduces the cost of capital
invested in the tactic as well 8s commissions and slippage. Given that
the martingale only returns the size of the original bet times the win
size:loss size ratio, careful comparison, considering the amount of capital tied up (losses plus margin), to alternative speculations should be
made. Keeping in mind that transactions will be numerous, tax effects
must also be added to the mix. Dealers with the advantage on all these
issues will probably find martingales of most use.
MARTINGALES ON CRUDE 011
As an example, consider the crude oil trading described in Campaign

Trading! and elaborated here. To apply a martingale to this trading
campaign, I selected the complex martingale from Appendix J, elaborated for the longer loss strings experienced when only 30% of the
trades are winners. When no martingale was in progress and a win
was recorded, equity went up by the amount of the win or down by the
amount of the MAE stop if it were hit.* If a martingale wasnt in
progress and a loss came up, the martingale began with the recording of a loss.
To track the martingale in a trading situation means keeping
track of the counts for wins and losses, but also the actual amounts won
or lost on each trade. In this application to crude, the martingale was
*( For those who havent read Campaign Trading!, MAE stops are developed on trading combinations which have an entry and and exit defined, without consideration of
stops. Therefore, its possible for the combination to record a win but, later, after
imposition of a stop, to have that win turned to a lass because, during the trade, the
adverse excursion exceeded the MAE stoo.
94
MARTINGALES
ended whenever the martingales count of outstanding values was equal

to or greater than zero or the martingales actual dollar winnings were
equal to or greater than zero. That is, even if the martingales count
had not played itself out fully, achieving a new high in equity would
stop the trading series.
In Table 7-3, the first column shows the number of share blocks
or contracts traded, in this case, one for the first loss in the third
row, two in the 18th row and so on. The second column totals the
number of contracts or share blocks lost as the martingale progresses. In this excerpted martingale, it was down ten contracts before winning on a five-contract trade in the last row. Column 3 suns
the trading points lost or gained by the martingale as it progresses.
In crude, $310 = 31 trading points at $10 each, usually displayed as
.31. Normally, only the last value will be positive, but not always.
See below. Lastly, column four accumulates the trading6 wins and
losses to form the equity curve.
Under these conditions you may not end the martingale with a
profit! Its possible to have a combination of wins and losses which
end the martingale series but at the same time experience a combination of win sizes and loss sizes which end the martingale with a
loss. For example, if the first trade is a loss of .31, the next trade will
be two contracts or blocks of shares. If that trade wins .12, then you
have ended the martingale but your profit is (2 Contracts X 12
Points/Contract) - 31 Points = -7 Points. In this eleven year series,
this happened nine times, the losses ranging from 15 points to 78
points with, as we shall see, very little impact on the outcome. Here
are the actual points lost: 25, 78, 29, 62, 15, 7, 56, 38, 7. In eleven
years, 41 martingale series were counted.
In practice, all this works out to steeper equity drawdowns than
trading without using the martingale and extremely quick, explosive
recoveries to new highs. The martingale progression served to increase the volatility of returns and, in the process, generate profits
two or three times the size of the same series of trades without martingale money management. Figure 7-3 shows the two equity curves
together, the first the dual-moving average trend trading system and
the second the same system with martingale money management.
Figure 7-3 shows the results of combining MAE stops with martingale money management. The Trend Equity line is the dual moving
MARTINGALES ON CRUDE OIL
95
Table 7-3 Exemplary Martingale Worksheet. A trading martingale must track both the martingales counts of trades won or
lost but also the actual amount won or lost on each trade, here, by
trade in the third column and cumulatively in the fourth column.
(This is just the beginning of a long martingale.)
Number
Blocks
Won or
Lost
False
False
False
-1
-1
-1
-1
-1
-1
-1
-1
-1
1
~~1
-~I
-1
-1
-2
-2
-3
-3
-3
-3
-3
-4
-4
-4
5
Martingales
BI&S
Outstanding
F&e
False
F&e
-1
-1
-1
~~- 1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-1
-3
-3
~43
-6
-6
-6
-6
-10
-10
-10
-5
Msxtingde
Equity
False
False
False
-0.31
-0.31
-0.31
-0.31
-0.31
-0.31
~,~.0.31
-0.31
-0.31
-0.31
-0.31
-0.31
-0.31
-0.31
-0.93
-0.93
-1.86
'~ 1.86
-1.86
-1.86
-1.86
-2.58
-2.58
-2.58
-2.43
Trend and
Martingale
Equity
112.77
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
112.46
111.84
111.84
110.91
110.91
110.91
110.91
110.91
110.19
110.19
110.19
110.34
96
MARTINGALES
Figure 7-3
Explosive Results of Martingale Management. Managing the same series of trades with a martingale sysrem produces explosive results at the cost of increased volatility. Drawdown B had an l-1 1 win-loss streak before recovering and
Drawdown C went 3-20 before recovering it all on just one trade and going on to
new highs in profitability.
average system using an MAE stop of .31. The only thing that changed
in the other line was the use of the martingale system. The result was
to triple overall profits at the expense of increased volatility of returns and greater commitment of capital.
Ignoring for the moment the dramatic movement around C, notice
that at A and B, downturns in the Trend Equity line became major
dips in the martingale line. On the other hand,the slope of the martingale line is much steeper upward. This is because, absent extended
drawdowns, the 2: 1 win/loss ratio of this trading combination and the
MAE stops holding losses to .31 mean, together with the growth in
contracts as the martingale progressed, meant that just one win would
~vercmne all the previous losses.
To say it another way, MAE kept the losses small until a solid
win came along, a win which was amplified by the greater number of
contracts being traded when it popped up. Thus, the sharp upthrusts
MARTINGALES
ON
CRUDE
OIL
97
in the Trend Equity line were amplified in the martingale line. A big
win when youre running a martingale with a good dollar win/loss
ratio is a very big win indeed.
Along with the increased volatility and higher profit came higher
capital commitments. Though profits easily funded the drawdowns,
had the worst come first, a loss series of 3,000 points ($30,000 would
have been endured up front, something most managements could not
stomach). Additionally, the margin required to support these positions
lowers the return to capital. Figure 7-4 shows how the margin requirements shift with the invocation of the martingale sequences during drawdown periods.
Margin required rises dramatically in the last drawdown where
as many as 28 contracts must be supported for what started as a
single-contract trade. This is meant for large money or large lines of
credit. However, most of the time, requirements are much less and the
computation of return to capital will be more favorable.
Taken together, this exemplary analysis nicely demonstrates the
strengths and weaknesses of the martingale strategy. On the favorable
Days in Campaign
Figure 7-4 Martingale Margin Requirements. Margin pops dramatically as the
martingale sequences elongate. Margins are the columnar elements of the chart
while profits arc the line element. Note the differing scales.
98
MARTINGALES
tally, there is the significant increase in profitability, an increase

which would be even more dramatic with a system that hit, say, 50%
of its trades correctly or had a men-e consistent win size loss size
ratio. The unfavorable tally includes the great increase in volatility,
experienced as larger drawdowns and long (not longer) periods of adversity before winning through. There are no perfect solutions in trading, just tradeoffs between risk and return. Martingales are no
exception and analyses like these will help the trading manager make
the implementation decision.
8
TRADING MANAGEMENT
While we are certainly far from true control of trading activity for
relatively assured profits, its not too early to consider how that might
work. Its reasonable to think of this because we can put B measure to
losses which is part of giving us a standard for performance (the other
half being measures of gain which have been well studied). We actually have a picture of how a particular trading combination should
perform; we know, to some degree, what constitutes normal experiences and what doesnt; and, since we know what the losses should be,
we have the gain/loss picture more firmly focused.
PORTFOLIO
IMPACTS
Though Ive used the gain/loss terminology, portfolio managers will

immediately think of risk/return. Those who assemble portfolios efficiently, portfolios which include speculative trading combinations,
will recognize that the limitation of losses alters the risk profile of the
trading asset which, if done across the board, means a shift in the
composition and weightings of their portfolio. Trading combinations
using MAE stops generally should have lower variance of returns but,
99
100
TRADING MANAGEMENT
as seen in Chapter 7, combinations using progressive capital commitment schemes (betting progressions) may have both higher returns
and higher variance of returns. All of this necessitates a portfolio reassessment.
DAY-TO-DAY
TRADING
On the more practical day-to-day level, trading by system using MAE

techniques should be more manageable than trading until the system
taps out or its time for the end-of-the-year wrapup. After all, management is a specific discipline. Without going into all the ways its
been defined, list the specific things managers do. Managers:
l
Study an issue to reach a goal.
Prepare or plan a course of action to reach the goal.
Predict from their plan a desired result.
Direct the execution of their plan.
Monitor and/or measure deviations from the desired result,
Correct their plan
Managers are always in the loop of measuring results, seeing deviations from plan and making corrections to the plan (Figure 8-l).
Traders, in contrast, are usually wondering if theres been a change in
plan. I took a loss, they mull. Was it a serious loss? Is it telling me
something? Did I do something wrong? Has the situation changed? Did
everyone else do something wrong? Am I out of touch? What happened?
Traders usually think they have a plan when what they have is
an idea of how to get into a trade. Someone one level up whos managing traders usually hasnt any illusions that hes planning, directing, and controlling. He tries to make sure the cannons are firing
regularly and pointedly rather than rolling around aimlessly.
Instead, what if the trader could turn to his plan and check if the
loss was within expected size or if the overall frequency of losses was
within expected ranges? What if wins were similarly comparable to
DAY-TO-DAY TRADING
101
Figure 8-l
Control Loop. While managers achieve planned results by changing

their plan as deviations crop up, traders usually have no way of knowing if their iailures are normal or abnormal. Traders have no means of making corrections toward
a goal. Its impossible to manage trading without the means to asses5 the situation
objectively.
play? For a trader, a simple checklist would then tell him if he were
performing properly (Figure 8-2).
Heres another thought: a trading manager who knows objectively what his traders losses should look like (as most trading managers feel intuitively) has a tool for making changes. He can look at
losses first through the checklist above to identify problems with the
trader, problems with conditions, or problems with trading rules (Figure 8-3).
To make these assessments, either at the traders level or the
managers level, some standard is necessary. To say whether a trader
102
TRADING
MANAGEMEN
Figure 8-2 Trader Performance. Broad categories of judgment tell the trader or
his manager if he or she is performing to spec. There should he no distress for a loss
taken properly hut there should be distress for a win taken improperly.
DAY-TO-DAY TRADING
Figure 8-3
103
Management Performance. Managers, like traders, make mistakes.

When results are out of bounds, adverse excursion measurements provide objective means of assessing where the problem is.
104
TRADING MANAGEMENT
was early or late in getting out of a bad trade, you must know where
the stop level should have been. If you believe the stop varies with the
trade, the conditions of the market, or the ability of the trader you
have a lot of variables to deal with and none you can cite without arbitrariness. If, on the other hand, you set the stops based on the elements you can control-the entry point, your indicators, and your
rules-theres a good chance youll find a distinct stop (and/or reversal) point to which you can manage.
Notice I said manage. Knowing your *wrong point objectively
gives you a beacon when you drift off course. Correct to the target as
a: good manager would. Knowing your historical win profitability and
choosing your loss level you should be able to
l
Manage losses: size, frequency, runs. Isnt there a better term

than runs for the phenomenon of losses occurring in a row and
depleting capital?
Manage precious trading capital by determining what level is
needed for each set of rules.
Manage trading rules and systems by loss/gain targets.
Manage your traders, measuring by whether they performed the
system they were charged to perform.
Manage speculative portfolios allocating your capital, people, and
systems by better-defined loss/gain targets.
ELABORATIONS
Typical Path
This book has focused on adverse excursion because it is most important to manage losing properly. However, back in Chapter 2, definitions of two forms of favorable excursion were also given. These two
definitions serve when defining what could be called the typical path
of a trade. Like adverse excursion, excursion analysis of winning trades
begins with separating the winners from the losers, the winning behavior from the losing behavior. These excursions are then analyzed
ELABORATIONS
105
with an eye to finding operational regularity, behaviors that can be

used in the management of the winning trade. Some of this was
touched on in Campaign Trading! where the ideas of add-on trades
and reversal trades were described as well as the subject of moving
up stops. Just as the general question for adverse price movement is
How far against us do winners go? the symmetrical question for favorable excursion is often How little for us do winners go?
Redefining Winners and Losers
Another analytical area thats sometimes fruitful is questioning what

a winner and a loser is. In this book, its been simply defined as a profitable or unprofitable trade. However, there may well be a middle zone
between winning and losing, an area plus or minus n points from the
entry where youve got a draw. Perhaps its defined by transactions
costs, ,perhaps by required return (somewhere between losing money
outright and not making enough to justify this trade over some other).
In this approach, you separate trades into three groups and see if behavior after entry is markedly different.
Related Indicators
The conjoint events (trading signal:excursion

measurement) that
MAE analysis uses could have been other events. For example, if you
trade the S&P from short-term money flows or exchange rates from
short term rate differentials, it would be legitimate to compare price
excursion, favorable and adverse, from your signal in the underlying indicator. Nothing in the technique described here restricts the
trading rules to price-based numbers.
MAE uses price extremes, normally highs and lows, but its occasionally fruitful to use closes instead of extremes as the measure. Highs
and lows are thinly traded and, sometimes, not really traded at all
while closes are difficult to trade on a stop basis. However, where liquidity is bare, the close might be the most likely price to actually receive or take. Too, its excursions are generally less than those of the
106
TRADING MANAGEMENT
high or low so that you get less adversity to handle at the cost of increased intraday vigilance and some premium for efficient execution
at the close.
CONCLUSION
All these elaborations stem from knowing what losses should look
like. That knowledge gives you the tool to control losses 80 the wins
can do their work. There are more than enough traders and trading
rules to generate winning trades but there are few objective techniques for managing losses appropriately. This book described one
such practical technique: measuring maximum adverse excursion.
Appendix
COMPUTING MAE
Maximum Adverse Excursion is defined and explained in Chapter 2.

This Appendix provides Microsoft Excel 4.0 code* which can be used
as is or modified to fit your own needs.
The worksheet is structured in rows, each row representing a
unit of time, usually one day.
Begin by loading Date, Open, High, Low and Close in the first
five columns, columns A, B, C, D, and E.
Next, in column F, indicate the position outstanding at the close
of the day. This value is a + 1 for a long position, 0 for no position, and
L 1 for a short position.
Any value will serve as a logical switch that columns to the
right can use for computation. The position column can be computed
or simply put in manually. If elaborate computations must be made,
additional columns can be inserted before this column to handle them.
In this example, the values have simply been inserted.
Next, record the price of the position taken. In cell G2 enter an
initialization value of zero. In cell G3, enter
=IF(F2< >F3,IF(F3=1,-E3,IF(F3=-1,E3,O)),G2)
* Microsoft has been good to date about compatibility between various versions of
Excel. Later versions extant by the time you read this should be able to use this code.
108
APPENDIX A
and extend this formula down column G to the bottom of your list of
prices.
This formula captures the closing price as the entry price of the
position. Modify the formula to use another price if thats what you
wish.
This formula displays long positions as negative values, the negative representing a cash outflow. Short positions are shown as positive values, the positive representing a cash inflow.
Next, we prepare a column which (1) computes profit or loss on
the trade and (2) serves as a switch to tell the later MAE column that
the trade has closed.
Assure that cells H2 and 12 are blank as row three will refer to
them.
In cell H3, enter:
=IF(G3< >G2,IF(G:!>O,G2%E3,IF(G2<0,G2+E3)))
and extend the formula down to the bottom of your price list.
This formula computes the net profit or loss, exclusive of commissions and slippage, as of the close of the day when the signal in
column F changes. Modify the G2 ~ E3 or G2 + E3 expressions to
include commissions and/or slippage.
Finally, we compute MAE with an involved expression which (1)
will not compute MAE when there is no position except on the day
closing a position, (2) will not compute MAE on the opening day of the
position taken at the close, (3) will compute the proper MAE on a reversai, and (4) will compute MAE using the MAX expression in Chapter 2 which compares zero, the difference between the entry and the
worst price, and the previous MAE. The spreadsheet logic is tortuous
but enter in cell 13:
=IF($G3< >O,IF($G2< >O,IF($H2
=FALSE,1F($G2<0,MAX(0,-5G2-$D3,$12),MAX(0,5C3
~$G2,512)),1F($G2<0,MAX(0,-5G2-$D3),MAX(0,5C3
p$G2)))),IF($H3< >FALSE,IF($G2<0,MAX(O,-$G2
-5D3,512),MAX(O,5C3-5G2,512))))
and extend the formula down to the bottom of your price range.
109
COMPUTING MAE
Here are exemplary computations for longs and shorts against

which you can check your spreadsheet (Table A-l).
Long MAE Example

Table A-l Long MAE Example. This trade goes long (indicated by a 1) on
the close of 1117/90 at 30.95. Things immediately go wrong, the low the next
day reaching 30.82 for an MAE of 30.95 - 30.82 = .13. The trade ends on
the close of 11/12/90. The Entry Price of -30.95 is negative to indicate a cash
outflow.
Row
Profit
1
Date
2
3
4
5
6
Open
11/6/90
11/7/90
1118190
11/9/90
11/12/90
High
28.93
29.40
31.32
31.50
29.32
Low
Close
29.62 28.70
30.95 29.30
32.35 30.82
31.97 30.42
29.53 29.03
Position
29.51
30.95
32.02
30.59
29.31
1
1
1
0
Entry OF
Price Loss
MAE
-30.95
-30.95
-30.95
0
False
0.13
0.53
1.92
False
False
False
-1.64
Short MAE Example

Table A-2 Short MAE Example. MAE grows steadily as a short (indicated
by a -1) from 25.56 at the close of 7/12/85 goes steadily into the hole until
the trade is ended on the close of 7123185. No MAF. is computed for 7112185 because there is no pricing after the trade is entered on the close. On the last
day of the trade, 7/23/85, adverse excursion reaches .38 as the high goes to
25.94 before the close.
Profit
Entry
Date
or
Open High Low Close Position Price
Loss
MAR
7112185
7/15/85
7/16/&i
7/17/85
7/18/85
7/19/85
25.77
25.43
25.56
25.69
25.61
25.58
25.76
25.66
25.70
25.69
25.73
25.62
25.54
25.37
25.47
25.53
25.58
25.52
25.56
25.59
25.53
25.55
25.62
25.57
71221.35
7123185
25.48
25.82
25.77
25.94
25.38
25.73
25.66
25.93
False
0.10
0.14
0.14
0.17
0.17
0.21
0.38
-1
-1
-1
-1
-1
-1
-1
25.56
25.56
25.56
25.56
25.56
False
False
False
False
False
False
False
-0.37
25.56
25.56
110
APPENDIX A
Remember, MAE never goes lower than its previous value in a trade
and never goes below zero.
MIXED MAE EXAMPLE

You have an example to test your code going from long to short, here
are results for a long from 5/24/89, exiting 6/9/89, shorting 6/12/89, exiting 6/23/89 in Table A-3.
Table A-3 Mixed MAE Example. Your logic should be able to handle reverx=Js such as this and also even faster reversals shown in Figure A-4.
Entry
Date Open High Low Close Position Price
5/24/89
5/25/89
5/26/89
5/30/89
5131189
6/l/89
6/Z/89
615189
6/6/89
617189
618189
619189
6112189
6113189
6114189
6/15/89
6/16/89
6119189
6/20/89
6/21/89
6122189
6123189
17.80
18.15
18.08
18.31
18.36
18.27
18.24
19.04
19.08
18.84
18.40
18.56
17.93
17.64
17.66
17.46
17.63
17.38
17.58
17.63
18.11
18.20
18.31
18.50
18.33
18.40
18.42
18.33
18.77
19.17
19.12
19.08
18.66
18.58
18.01
17.74
17.73
17.93
17.63
17.62
17.74
18.08
18.26
18.49
17.78 18.28
18.13 18.21
18.07 18.31
18.18 18.38
18.22 18.27
18.20 18.31
18.18 18.75
18.91 19.13
18.81 19.11
18.36 18.41
18.24 18.57
18.20 18.23
17.58 17.61
17.39 17.41
17.22 17.57
17.45 17.74
17.38 17.51
17.25 17.57
17.52 17.58
17.57 18.01
17.99 18.10
18.18 18.48
1
1
1
1
1
1
1
1
1
1
1
0
-1
-1
-1
-1
-1
-1
-1
-1
-1
0
-18.28
-18.28
- 18.28
-18.28
-18.28
-18.28
-18.28
-18.28
-18.28
-18.28
- 18.28
0
17.61
17.61
17.61
17.61
17.61
17.61
17.61
17.61
17.61
0
Profit
or
Loss
MAE
False
False
False
False
False
False
False
False
False
False
False
-0.05
False
False
False
False
False
False
False
False
False
-0.87
False
0.15
0.21
0.21
0.21
0.21
0.21
0.21
0.21
0.21
0.21
0.21
False
0.13
0.13
0.32
0.32
0.32
0.32
0.47
0.65
0.88
COMPUTING MAE
111
MAE in fmmediafe Reversals

Finally, positions which go immediately from long to short (or vice
versa) test your logic harshly. Table A-4 is an example the code above
handled.
Table A-4 Immediate Reversal Example. Long at 21.60 on the close
10/6/92. Close and reverse on the close 10/21/92 at 21.44. Note the MAE =
21.6 - 21.37 = .23 on 10/21/92 is that for the previous long from 21.60 while
the MAE = 21.48 - 21.44 = .04 on 10/22/92 is that for the short taken at
the close on 10/21/92.
Profit
Date
1016192
1017192
10/s/92
10/g/92
10/12/92
10/13/92
10/14/92
10/15/92
10/16/92
10/19/92
10120/92
10/21/92
10/22/92
lo/23192
10/26/92
10127192
10/28/92
10/29/92
10/30/92
Entry
Open High Low Close Position Price
L:s
MAE
21.55
21.67
21.71
21.82
22.00
21.97
21.81
21.96
22.04
22.01
21.63
21.76
21.41
21.27
20.97
21.23
20.82
20.95
20.54
False
False
False
False
False
False
F&e
False
False
False
False
-0.16
F&e
False
False
False
False
False
False
False
0
0
0
0
0
0
0
0
0
0
0.23
0.04
0.04
0.04
0.04
0.04
0.04
0.04
21.63
21.70
21.78
22.14
22.04
22.05
21.89
22.06
22.08
22.03
21.92
21.77
21.48
21.29
21.14
21.24
21.01
20.98
20.65
21.53 21.60
21.65 21.67
21.70 21.73
21.82 22.09
21.97 22.02
21.81 21.81
21.77 21.81
21.95 22.03
21.92 21.96
21.86 21.86
21.79 21.83
21.37 21.44
21.18 21.19
21.05 21.07
20.97 21.14
20.93 20.94
20.69 21.00
20.63 20.69
20.46 20.61
1
1
1
1
1
1
1
1
1
1
1
-1
-1
-1
-1
-1
-1
-1
-1
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
I put all these computations (MAE, MaxFE, and MinFE) in the

sane spreadsheet, If you do this, youll need to adjust the column reference~ in the formulae to match your arrangement.
Appendix
COMPUTING MAxFE
Maximum Favorable Excursion is defined and explained in Chapter 2.

This appendix provides Microsoft Excel 4.0 code* which can be used
The worksheet is structured in rows, each row representing a
Begin by loading Date, Open, High, Low, and Close in the first
of the day. This value is a +l for a long position, 0 for no position, and
- 1 for a short position.
* Microsoft has been good to date about compatibility between variom version8 of
Excel. Later versions extant by the time you read this should be able to we thiv code.
112
COMPUTING MAxFE
113

initialization value of zero. In cell G3. enter
=IF(F2< >F3,IF(F3=1,-E3,IF(F3=-l,E3,0)),G2)
prices.
This formula captures the closing price as the entry price of
the position. Modify the formula to use another price if thats what
you wish.
This formula displays long positions as negative values, the negative representing a cash outflow. Short positions are shown as positive values, the positive representing a cash inflow.
Next, we prepare a column which (1) computes profit or loss
on the trade and (2) serves as a switch to tell the later MaxFE column
that the trade has closed.
Assure that cells H2 and 12 are blank as row three will refer to
them.
In cell H3, enter:
=IF(G3< >G2,IF(G2>O,G2%E3,IF(G2<O,G2+E3)))
column F changes. Modify the G2 - E3 or G2 + E3 expressions to
Finally, we compute MaxFE with an involved expression that (1)
will not compute MaxFE when there is no position except on the day
closing a position, (2) will not compute MaxFE on the opening day of
the position taken at the close, (3) will compute the proper MaxFE on
a reversal, and (4) will compute MaxFE using the MAX expression in
Chapter 2 which compares zero, the difference between the entry and
the worst price, and the previous MaxFE. The spreadsheet logic is tortuous but enter in cell 13:
114
APPENDIX B
=IF($G3< >O,IF($G2< >O,IF($H2

=FALSE,IF($G2<O,MAX(O,$C3+$G2,$12),MAX(O,$G2
-$D3,$12,,,IF($G2<O,MAX(O,$C3+$G2),MAX(O,$G3
-$D3N)),IF($H3<
>FALSE,IF(G2<O,MAX(O,CS
+G2,12),MAX(O,G2-D3,12)N
and extend the formula down to the bottom of your price range.
which you can check your spreadsheet (Table B-l).
Table B-2 is another, more complicated example for code testing.
Table B-l Exemplary MaxFE Computations. First long, then short but neither trade moves very far favorably. The negative -31.04 refers to the cash
outflow from a long position; the positive 31.2 to the cash inflow from a short
position.
A
D&e
2 6/28/87
3 6/29/87
4
5
6
7
8
9
10
11
12
13
6/30/87
7/l/87
7/2167
716187
l/7/87
7/B/87
719/87
7,12/87
7/13/87
7/14/87
14 7/15/87
15 7/16/87
Open High Low Close

31.06 31.12
30.95 31.08
30.99 31.04
31.15 31.21
31.19 31.19
31.11 31.13
30.95 31.03
30.99
30.67
30.97
31.15
31.08
31.03
30.75
30.66
F
Long (1)
or Short
31.02
31.04
31.04
31.19
31.18
31.13
31.03
31.15
31.00 31.15
31.12 31.20 31.12 31.20
31.19 31.34 31.19 31.33
-1
-1
31.37 31.44 31.37 31.39
-1
31.55 31.72 31.05 31.68 -1
31.66 31.70 31.57 31.67 -1
31.66 31.66 31.56 31.65 -1
II
Price P&L
n
False
-31.04
-31.04
F&e
False
-31.04
-31.04
False
0
0.09
0
False
False
0
31.20
False
31.20
False
F&e
31.20
31.20
F&e
31.20
False
31.20
False
Entry
MaxFB
F&e
0.00
0.11
0.17
0.17
False
False
False
0.01
0.01
0.15
0.15
0.15
COMPUTING MAxFE
115
T a b l e B-2 MaxFE During Reversal. A long from 21.6 reverses to a short on

the close of 10121/92. The MaxFE for 10/21/92 is that of the long position.
The MaxFE on 10/22/92 is that for the short from 21.44. The code supplied
above should handle even this situation.
A
P&L
MaxFE
F&e
F&W
False
False
False
False
False
False
False
False
False
-0.16
False
False
False
False
False
F&e
False
False
False
F&e
0.10
0.18
0.64
0.54
0.54
0.54
0.54
0.64
0.54
0.64
0.54
0.26
0.39
0.47
0.51
0.75
0.81
0.98
0.98
0.96
Long (1)
1
Date
2
3
4
5
6
7
8
9
10
11
12
13
14
16
16
17
18
19
20
21
22
10/6/!32
10/7,92
10/8/92
10/9/92
10/12/92
10/13/92
10,14,92
10/15/92
10/16/92
10/19/92
10/20/92
IO/21192
10122192
10/23/92
10126192
lo/27192
lo/28192
10,29,92
10,30/92
11/2,92
11/3/92
Open High Low Close

21.55
21.67
21.71
21.82
22.00
21.97
21.81
21.96
22.04
22.01
21.63
21.76
21.41
21.27
20.97
21.23
20.62
20.95
20.64
20.72
20.77
21.63
21.70
21.78
22.14
22.04
22.05
21.89
22.06
22.06
22.03
21.92
21.77
21.48
21.29
21.14
21.24
21.01
20.98
20.65
20.76
20.78
21.53
21.65
21.7
21.82
21.97
21.81
21.77
21.95
21.92
21.66
21.79
21.37
21.18
21.05
20.97
20.93
20.69
20.63
20.46
20.55
20.6
21.60
21.67
21.73
22.09
22.02
21.81
21.81
22.03
21.96
21.66
21.63
21.44
21.19
21.07
21.14
20.94
21.00
20.69
20.61
20.70
20.66
or Short
Entry
( -1)
1
1
1
1
1
1
1
1
1
1
1
-1
-~I
-1
-1
-1
- 1
- 1
-1
-1
-1
Price
-21.60
-21.60
-21.60
-21.60
~~-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
Appendix
COMPUTING MINFE
Minimum Favorable Excursion is defined and explained in Chapter 2.

This appendix provides Microsoft Excel 4.0 code* which can be used
The worksheet is structured in rows, each TOW representing a
Begin by loading Date, Open, High, Low and Close in the first
of the day. This value is a + 1 for a long position, 0 for no position, and
- 1 for a short position.
* Microsoft has been good to date about compatibility between various versions of
Excel. Later versions extant by the time you read this should be able to use this code.
116
COMPUTING MINFE
117

initialization value of zero. In cell G3, enter
=IF(F2< >F3,IF(F3=1,-E3,IF(F3=-l,E3,O)),G2)
prices.
This formula captures the closing price as the entry price of the
position. Modify the formula to use another price if thats what you
wish.
This formula displays long positions as negative values, the negative ,representing a cash outflow. Short positions are shown as positive values, the positive representing a cash inflow.
Next, we prepare a column which (1) computes profit or loss on
the trade and (2) serves as a switch to tell the later MinFE column
that thetrade has closed.
Assure that cells H2 and I2 are blank as row three will refer to
them. In cell H3, enter:
column F changes. Modify the G2 - E3 or G2 + E3 expressions to
Finally, we compute MinFE with an involved expression that (1)
will not compute MinFE when there is no position except on the day
closing a position, (2) will not compute MinFE on the opening day of
the position taken at the close, (3) will compute the proper MinFE on
a reversal, and (4) will compute MinFE using the MAX expression in
Chapter 2 which compares zero, the difference between the entry and
the worst price, and the previous MinFE. The spreadsheet logic is tortuous but enter in cell 13:
APPENDIX C
118
=IF($G3< >O,IF($G2< >O,IF($H2

=FALSE,IF($G2<O,MAX(O,$G2+$D3,$IZ),MAX(O,$G2
-C3,$12,),IF($G2<O,MAX(O,$G2+$D3),MAX(O,$G2
-C3)))),IF($H3< >FALSE,IF($G2<0,MAX(O,$G2
+$D3,$12),MAX(O,$G2--C3,$12W
and extend the formula down to the bottom of your data range.
which you can check your spreadsheet (Table C-l).
Exemplary MinFE Computations. First long (the -21.6 refers

to the cash outflow in a long position), then short, this is the most difficult
situation for the spreadsheet logic to handle.
Table C-l
Row
21
22
23
Long (1)
or Short
Entry Price P&L MinFE
Date Ooen High Low Close
( -1)
lo/W92
10/7/92
10,8,92
10/g/92
10,12,92
10,13,92
10/14,92
10/15/92
IO,16192
10/19/92
10,20,92
10,21/92
IO,22192
10/23/92
10,26,92
lo/27192
10,26,92
lo/29192
10/30/92
11/z/92
11,3,92
11,4,92
21.55
21.67
21.71
21.62
22.00
21.97
21.81
21.96
22.04
22.01
21.83
21.76
21.41
21.27
20.97
21.23
20.82
20.95
20.64
20.72
20.77
20.40
21.63
21.70
21.78
22.14
22.04
22.05
21.89
22.06
22.08
22.03
21.92
21.77
21.48
21.29
21.14
21.24
21.01
20.98
20.65
20.7fi
20.78
20.44
21.53
21.65
21.70
21.82
21.97
21.81
21.77
21.95
21.92
21.86
21.79
21.37
21.18
21.05
20.97
20.93
20.69
20.63
20.46
20.55
20.60
20.28
21.60
21.67
21.73
22.09
22.02
21.81
21.81
22.03
21.96
21.86
21.63
21.44
21.19
21.07
21.14
20.94
21.00
20.69
20.61
20.70
20.65
20.40
1
1
1
1
1
1
1
1
1
1
- 1
- 1
- 1
-1
-1
-1
-1
- 1
- 1
~~ 1
-1
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
-21.60
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
21.44
0.05
0.10
0.22
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.37
0.00
0.15
0.30
0.30
0.43
0.46
0.79
0.79
0.79
1.00
Appendix
GENERATING A
FREQUENCY DISTRIBUTION
Instructions for setting up a frequency distribution will vary by the

software youre using. These are generic instructions for Microsoft
Excel. Its important to master this technique because, otherwise,
generating the graphs shown in this book is extremely tedious.
For this example, a spreadsheet was prepared as in Appendix A
and two new columns, J and K, were added. The rows, as before, are each
days prices and computations. Therefore, the columns are as follows:
A
B
C
D
E
F
G
H
Date
Open
High
LOW
Close
Position (Long +l, Short -1)
Entry Price
Profit or loss on the trade at
ClOW
I
J
K
MAE
Winning MAE
Losing MAE
119
120
APPENDIX D
To prepare column J, go to cell 53, enter:

=IF(HS<
>FALSE,IF(HS>O,IS))
and extend the formula down to the end of your data array. In column
K, go to K3, enter
=IF(H3<
>FALSE,IF(H3<0,13))
and extend the formula down to the last line of your data. These two
formula record the MAE! if its a winner or loser, respectively. Otherwise, they enter a FALSE value which wont be picked up by the function creating the frequency distribution.
Now, go to the last line of your data array. The spreadsheet I used
for this book has 3,069 lines or rows, so all the formulae were extended
to row 3069. Just below the last line, in cell 53070 (or the equivalent
in your spreadsheet) enter:
= count( j3:j3069)
In cell K3070, enter:
= count(i3:i3069)
These two counts will serve as checksums on the frequency distribution. This is a good place to enter other statistical measures such as
average, median, skew, and kurtosis. If you know what these are,
youll know how to put them in! For the graphically oriented amongst
us, lets get on with making the graph. Make these entries:
Cell 13070 MAE
Cell 53070
Winners
Cell K3070
LOSSIT
Next we make the categories by telling the spreadsheet the size of each
bin. Enter:
Cell 13071
.15
GENERATING A FREQUENCY DISTRIBUTION
121
In cell 13072 enter:

= 13071+.15
and extend that cells formula from 13072 down to cell 13084 which
should produce a value of 2.1 in cell 13084. You should now have something that looks like this:
I
3070
MAE
Winners
LOSWS
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
0.15
0.30
0.45
0.60
0.75
0.90
1.05
1.20
1.35
1.50
1.65
1.80
1.95
2.10
To make the distributions, select cells 53071 through J3085* and

type (Dont ENTER!):
= frequency(j3:j3069,13071:13084)
Then, if youre brightly using a Mac, while holding down the ti key,
press ENTER. If youre using a Windows machine, press CTRL +
SHIFT + ENTER for the same result.
*You select through 53085
to allow a cell for
allow for the full range ofoccurrences.
overflow if the bins you specify dont
122
APPENDIX D
To make the distribution for losing MAE, select cells K3071

through K3085 and type:
= frequency( j3:j3069,13071:13084)
before pressing ti + ENTER or CTRL + SHIFT + ENTER as your
machine dictates. You know have the raw distributions to which I have
added checksums at the bottom:
I
3070
MAE
Winners
LOSIX%
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
0.15
0.30
0.45
0.60
0.75
0.90
1.05
1.20
1.35
1.50
1.65
1.80
1.95
2.10
80
17
4
0
0
1
1
0
0
1
0
0
0
0
0
104
21
44
28
28
7
10
2
3
1
0
0
1
2
1
0
148
Finally, to get to the graphics, highlight 13070 through K3084

and step through the graphic routine for your spreadsheet. For Excel
4.0 for the Mac, select Ct + N, then C, then return. That will give you
the basic graph which you can then customize as you like.
Appendix
m FOR
SHORTS AND LONGS
You would like to distinguish between the MAE distributions for short
positions and that for long positions, probably because you feel the adverse price movement in those two situations is different. If it were
different, youd adjust your stops and reverse point differently depending on the direction of your trade. This appendix shows you some
Excel code that you can use to construct the two charts youll need to
analyze this.
To begin with, lay out the data as described in Appendix D, to wit:
Column
Information in Column
A
B
C
D
E
F
G
H
I
Date
Open
High
LOW
Close
Position (Long +l, Short - 1)
Entry Price
Profit or loss on the trade at close
MAE
123
124
APPENDIX E
Use the same entries described in those Appendices to fill in the

data. (An example is below). Then, label column J as Short MAE and
in cell 53 enter:
These two columns will extract from the MAE column, column I, the
MAE values if the position is long or short and it will extract them on
a daily basis. Extend the formulae in row 3 down to the bottom of your
data range.
Next create four new columns as follows:
Winning Long Losing Long Winning Short
MAE
MAE
MAE
Long Short
MAE
=IF($HY< >
FALSE,IF($G2
<O,IF($H3>
=IF($H3< >
FALSE,IF($GZ
<O,IF($H&
=IF($H3< >
FALSE,IF($GZ
>O,IF($H3>
=IF($H3< >
FALSE,IF($G2
>O,IF($B3<
0,513)))
0,513)))
0,$13)))
0,513)))
These formulae will check (1) if a position has closed, (2) if it was a long
or a short (by looking in column G), and (3) if it was a winner or a loser.
The column with the formula that checks out will transfer the MAE
value into itself; otherwise, it will set the cell to FALSE, a reading
which is ignored by the summary statistical functions well use next.
Table E-l is an exemplary table against which you can check your code.
Next, you summarize the data in columns L through 0 just a8 it
was summarized in Appendix D. In cells L22 through 022 enter the
labels for the graph youre creating:
ROW
22 Winners, Long Losers, Long Winners, Short Losers, Short

In cells, K2.3 through K33 enter the sizes of the bins you want. In
this example, Ive arbitrarily set the bin size to .l. You can enter any
MAE FOR SHORTS AND LONGS
125
Table E-l Long/Short MAE Computation. This table of artificial prices

shows all four possible combinations of winning and losing while being long
or short. The code described in this appendix generates either a FALSE or the
MAE value for the correct situation in columns L. M. N. and 0.
value you wish and you can have more bins or fewer bins as you think
necessary. Here you end up with values from .l to 1.0.
Select the range from L23 to L34 and type-but dont hit
ENTER= frequency (13:121,k23:k33)
and, while holding down the 6 key, press ENTER. If youre using a
Windows machine, press CTRL + SHIFT + ENTER for the same result. The entire range is filled automatically by EXCEL; you dont
need to extend the formula down. Normally, youd have far more rows
than just three through twenty-one, so youd adjust 13:21 in the formula to match the range you want checked. If your range of bins were
more or less than K23:K33 youd adjust that as well.
APPENDIX E
126
Just so you can be sure, you follow the same procedure for
columns M, N, and 0. Each time, select the range from row 23 to row
34. Here are the formulae for each column:
0
= FREQUENCY
(M3:MZl,K23:K33)
= FREQUENCY
(N3:N21$23:K33)
= FREQUENCY
(03:021$23:K33)
After adding cells summing the distributions at the bottom of

the table, you should end up with something like this:
Table E-2 Frequency Distributions. The distribution of adverse

excursions from the sample table in Table E-l is very small,
there being only one example of each case.
ROW
Winners,
LOIW
22
23
24
2s
28
27
28
29
30
31
32
33
34
3s
0.1
0.2
0.3
0.3
0.4
9.5
0.8
0.7
0.8
0.9
1.0
Overflnw
Totals
1
0
0
0
0
0
0
0
0
0
0
0
1
Lone
N
Winners,
Short
L0SlXS,
Short
0
0
0
0
1
0
0
0
0
0
0
0
1
0
1
0
0
0
0
0
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
1
LO.%3~S,
It would do no good to draw a chart of this example since there

is only one of each case. I leave it to you to draw one chart of K22:M33
and another chart of K22:K33,N22:033.
Youd the compare the
stop points youd pick off each chart to see if there were a material
difference.
Appendix
COMPUTING
PROFIT CURVES
COMPUTATION
It isnt always possible to pick out a stop and/or reverse point just from
a frequency diagram. That only shows the number of trades that have
taken place in each MAE bin and, frequently, the curve for winners
will overlap the curve for losers. To sort this out, using the same bins,
compute the profit or loss from the winners and losers and plot the
two curves for comparison. Chapter 4 outlines this process generally
with tables. This appendix shows you scme Excel code that you can
use to construct the charts youll need to analyze this.
To begin with, lay out the data as described in previous appendices as follows:
127
Column
A
B
C
D
E
F
G
H
I
J
K
L
M
Date
Open
High
LOW
Close
Entry Price
MAE
Winning MAE
Losing MAE
MinFE*
MaxFE
* For exposition, MinFE and MaxFE were shown in column I. Now move them to columns L and M by rearranging columns.
See columns A through I of Table E-l. You should have something like
this:
128
COMPUTING PROFIT CURVES
129
The table to be created here will have the MAE bins values
across the top of the columns. I set them for this example at .ll, .21,
.31, .41, 51, and .61 which values will vary as you change the size of
the bins during your analysis. These would be your stops. The .ll
stop would allow adverse excursion up to .lO before being triggered,
for example.
The rows will be each day (or week, month or period). The result
in each cell of the table will be the end-of-the-day closed equity from
all the trades to date, assuming the trades were stopped at .Ol less
than the column heading (that is, stopped at .l for the first column, .2
for the second column and so on).
First put in the column headings as:
ROW
0.11
0.21
0.31
0.41
0.51
0.61
Then, in cell N3, enter:

=IF($H3< >FALSE,IF(N$l<
and extend this formula to the right to, in this example, cell 53. Then
extend N3 through S3 (in Excel notation N3:S3) down to the bottom
of your data range. The results for columns N through S are shown in
Table F-l.
To be sure you understand the calculations, you should work
through the impact of each of the four trades here. This will be important later when you analyze changing stops and wonder why the
results arent always intuitive.
The first trade closes on row 7 with an adverse excursion of just
.07. None of the stops are tripped and the profit booked to date is increased by the trades profit of .09. This is the simplest case possible.
Next, in row 14, a trade closes with a profit of .8 (80 ticks) and
an MAE of .ll. Looking at cell N14, you can see the impact of tripping the stop set at .ll. Instead of adding 80 points of profit as in cell
014, eleven points of loss is added from the stops being executed:
.09 - .ll = -.02.
130
APPENDIX F
Table F-l Profit Curves. Too tight a stop at .ll causes losses
in this sample data while looser stops are profitable. Row 21 will
be the summary-to-date of each stops impact.
ROW
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
11
18
19
20
21
0.11
0.21
0.31
0.41
0.51
0.61
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
-0.02
-0.02
-0.02
-0.02
-0.13
-0.13
-0.13
-0.24
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
0.89
0.89
0.89
0.69
0.66
0.68
0.66
0.47
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
0.89
0.89
0.89
0.89
0.58
0.58
0.58
0.27
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
0.89
0.89
0.89
0.89
0.59
0.59
0.59
0.18
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
0.89
0.89
0.89
0.89
0.59
0.59
0.59
0.17
0.00
0.00
0.00
0.00
0.09
0.09
0.09
0.09
0.09
0.09
0.09
0.89
0.69
0.89
0.89
0.59
0.59
0.59
0.17
In cell N18, a trade with a loss of .3 and an MAE of .35 causes

the stop at .ll to be tripped limiting the loss. The cell is changed as
-.02 - .11 = -.13 instead of losing 30 ticks c.3). Cell 018 is calculated as 39 - .21 = .68 from the stop set at 21. In Cell PlS, the stop
at .31 generates this calculation: 89 - .31 = 58. In Cell &IS, the
stop is not tripped and the loss is just .3, so the cell is calculated as
.89 - .3 = .59.
Row 21 is the last trade shown with a loss of .4 and an MAE of
.43. I leave it to you to verify the computations of the cells values.
COMPUTING PROFIT CURVES
131
Figure F-l Profit vs. Stop Level. A tight stop at .11 produces losses while profits peak with a stop of .21 using the sample data in this Appendix. Results arent
always so neat but the display is typical.
GRAPHIC DISPLAY
To display the data graphically, select N1:Sl and N21:S21.* Then
order up a chart of the selected range.+ Figure F-l is the result using
the numbers in Table F-l.
In Figure F-l, Nl:Sl are the stop levels specified in the worksheet, all set one tick higher than the limits of the bins used for
analysis.
Interpreting Figure F-l is straightforward since all we seek is
a point for placing a stop. This (limited) experience shows that a stop
at .21 would generate the most profit.
Occasionally, youll have two or three stop levels that are close to
each other in profitability at the end of a long run of trades. It may be
* See your spreadsheets manual for non-contiguous aelection. In
the 1 key when selecting additional ranges of cells.
Excel 4.0, hold down
In Excel 4.0, type +N, C, return. In Excel 5.0, select the Chart Wizard button and
follow
instructions.
132
APPENDIX
Figure F-Z Closed Profit or Loss Over Time. Using the three dimensional charting capability in a spreadsheet, the numbers in Table F-l generate information
about which stop levels perform best over the trading period.
that one performed better prior to the end of trading and youd like to
check that. You can plot the equity performance of all three stop levels over time using the 3D graphics functions in your spreadsheet.
Using Table F-l as the example, select N3:P21, generate a chart and,
if its not your default option, change the display to a three dimensional line drawing.*
Figure F-2 with its four trades doesnt tell you much more than
the ending results of Figure F-l but it does show that the lowest stop
generates the slowest rate of decline when the two-trade losing streak
hits toward the end of trading. This sort of inspection of a long series
of trades may be valuable to a trader picking stops for the future.
* Excel users can consult the Gallery in Excel 4.0 and the Chart Wizard in Excel
5.0.
Appendix
RANGE fxm VourrI~~Ty
Here are some examples from a variety of tradables showing the relationship of range to volatility. These graphs may give you some
trading ideas for tracking volatility using range as a proxy or for adjusting entry points on volatility-based trading systems. Naturally,
this book is concerned with the effects expanded ranges have on
stops.
Each graph (Figures G-l through G-4) has been prepared
using a standard computation of volatility for twenty days (to pick
a number). From the high-low range for each day, the moving,
twenty-day, simple average was computed. The two series were
mean-normalized by adjusting the volatilitys mean to that of the
ranges, giving a graphic where the coincident fluctuations of both
were highlighted.
133
134
Figure G-l
APPENDIX C
AT&T Range and Volatility. Though not consistent in their fluctuation, changes in range and volatility in AT&T during early 1995 oken coincided
but occasionally diverged over longer periods.
RANGE AND VOLATILITY
135
Coffee Volatility and Range. Coffee, a tradable of immense fluctuaFigure G-2

tions shows that while range can change as volatility changes, it can remain at
higher levels even as volatility recedes.
APPENDIX C
Figure G-3 June 1995 U.S. Treasury Bond Futures. Leads and lags in the changes
of range and volatility are apparent in this graph of the futures contracts activity,
Though only a short extract of time, it suggests that range expansion or contraction
may precede volatility expansion or contraction, a neat intuitive result.
RANGE AND VOLATILITY
137
Figure G-4
Dow Industrials Range and Volatility. While daily range (smoothed)
certainly fluctuates, the associated volatility nwves much more quickly to produce
sharper peaks and valleys than that of the smoothed range. The coincidence of
these two lines suggests the range/volatility relationship may be most applicable in
indices where the law of large numbers has great effect.
Appendix
RANGE EXCURSION
This appendix shows how youd check to see if range expands or contracts after your combination of tradable and trading rules calls for a
trade. This programs an exemplary spreadsheet to demonstrate one
way of approaching the question.
To begin with, lay out the data as described in previous appendices in Table H-l. Notice that Ive kept columns I through M for
MinFE and MaxFE, respectively, but in the tabular output below, Ive
suppressed their printing to save space here.
Devote the first 63 rows to data so that the trading system has
all the data needs on row 64. Then go to cell N60 and, beginning from
there, enter column headings as below in Figure H-l.
138
RANGE EXCURSION
139
Table H-l Spreadsheet Setup. To be consonant with

previous appendices, extend the spreadsheet youve already got rightward by adding columns for the range
computations.
Column
A
B
C
D
E
F
G
H
I
J
K
L
M
Date
Open
High
LOW
Close
Entry Price
MAE
Winning MAE
Losing MAE
MinFE
MaxFE
* For exposition, MinFE and MaxFE were shown in column I.

Now move them to columns L and M by rearranging columns.
After the headings are all entered, begin in Cell N 64 and make
the following entries, leaving cell T64 blank:
Cell
N64
064
P64
Q64
R64
S64
U64
Cell Entry
=ROUND((SUM(C45:C64)-SUM(D45:D64))/20,2)
=C64-D64
=IF(G64< >0,064-N64)
=IF(G64< >O,IF(G64< >G63,O,MAX(Q63,C64-D64)))
=IF(G64< >O,Q64-N64)
=IF($G64< >O,IF($F64< >$F63,$N64))
=IF(G64< >0,064-T64)
140
APPENDIX
Figure H-l
Column Headings. To prepare for the computation of range expansion, enter these headings in these cells. If you prefer to put headings on the top
row. thats fine, too. Formula entry will start on cell N64. (SMA20 is a mnemonic
for 20.day simple moving average.1
Next select N64:U65 and hit CMD-D to extend the formulas

down one line. Then go to cell T64 and enter:
Cell
T65 =IF($G65<
Cell Entry
>O,IF($F65< >$F64,$N65,T64))
Lastly, ycm extend to the bottom of your data rows. If that data
ended on row 300, for example, select N65:U300* and hit CMD-D.
Excel may give you a warning that it cannot proceed unless there is no
Undo. If so, accept the warning and proceed anyway. This should extend all formulas down to the bottom of your model creating the data
series for your inspection.
GRAPHICS
Because different versions of Excel have differing interfaces to specify graphics Ill restrict myself to getting the data into place for graphing as shown in the text.
The values computed above create strings of values during the
trades. Since there is no way to know where in the column they will
*You can select large areas
like this without clicking and dragging by going to the

upper leftmost cell, selecting Go To (with a CMD-G or from a menu) entering the
bottom rightmost cells number, and hitting SHIFT-RETURN.
RANGEEXCURSION
141
occur and you should graph each from the day the trade starts you
must tediously cut these strings, transpose and especially paste them
into a second worksheet where their statistical properties and graphics can be processed.
Therefore create two new worksheets, one for comparing ranges
after entry to a running 20-day simple moving average and a second
for comparing those ranges to the range exactly at entry. Take extra
precautions to keep these two spreadsheets separate.
Each spreadsheet will have two blocks of data. The first block will
be, roughly from A3:AZ36 and the second from, roughly, A38:AZ65.
These can be changed as you need. The number of rows is dependent on
the number of trades and the number of columns is dependent on how
many days the trades go on. If you dont want to transpose the values
from vertical, as they are in the model above to horizontal, these two
blocks should be arranged vertically rather than horizontally.
Begin with the headlines, examples of which are in Figure H-2
which shows the beginning of the block for Winners. I put the title for
Losers in cell B38. As an example, I use the comparison of range to the
moving 20-day average of range.
Once the headlines are set, return to the original model first constructed in this appendix (I11 refer to it as the range expansion
model). Go to your first trade which, for example, looks like Table H-2.
Next switch over to the Graphics Worksheet and select cell A5.
From the menu, select EDIT Paste Special. In the pop up window, select Paste Values and click the box for Transpose, then click OK (or hit
Return). The values themselves will be laid out as shown in Figure H-2.
Next, return to the range expansion model and go to your next
trade by proceeding down the column. As before, highlight the entire
series of values, copy the series, return to the graphics worksheet, and
Figure H-2 Titles for Graphics Worksheet. Take care to keep the worksheets for
the two different range expansion measures separate with embedded comments.
Data for the first trade be,e,ins on row five.
142
APPENDIX H
Table H-2 Selecting Values to Transpose. The bordered values should be

highlighted (selected) and copied.
Over and
Under
Daily
Over or
Range at
Range Days Under Max
Range
Range Entry, Life Range at
Entry
SMASO Range SMASO Ram% Exrransion at Entry of Trade
0.18
0.18
0.18
0.18
0.18
0.19
0.19
0.19
0.07
0.06
0.11
0.10
0.28
-0.11
-0.12
-0.07
FZ3lZX
False
0.19
0.08
False
-0.11
0.15
-~0.04
0.07
0.07
0.11
m~o.11
-0.11
-0.07
False
False
F&e
0
0.15
False
False
False
-0.19
-0.04
0.01
0."6
0.11
F&X
False
False
-0.11
-0.04
paste them specially, if the trade is a winner, into cell A6. If it is a

loser, paste it into cell A39. Continue this until all the measured range
expansions for the approach of comparing to the ZO-day average have
been transposed into the graphics worksheet. Use the same procedure
for comparing range expansions to the price at entry but use the other
graphics worksheet to paste your values.
To create the analytical graphics in Chapter 5 use the following
approaches. For the daily range excursions such as Figure 5-4 select
the full range of winning cells, create a chart and change the chart
type to XY (Scatter). You can select either just winners, just losers,
or both. (Use the command key while highlighting to select noncontiguous data points.) For line charts like Figures 5-5 to 5-6,
change the chart type to line.
I put the statistical summary functions at the bottom of each
worksheet. To create Figure 5-7, I installed frequency distributions in
cells A77:A91. These are a little tricky to enter so let me describe it
practically. In cell B77, put the title Losers and in C77 put Winners.Then,incellsA78:A90,put
-.7, p.5, -.3, -.1,0,.1,.3,.5,.7,.9,
1.1, 1.3, and 1.5. Next select cells B78:B91 (yes, thats 91) and type in
the following:
= FREQUENCY(B39:AVt%,A78:A90)
RANGEEXCURSION
143
Figure H-3
Frequency Distribution Expressions. These formulations summarize
the distribution of winners and losers range expansions in the graphics worksheet.
BUT, dont hit the return key. Instead, hold down the command button and hit the enter key. This will extend the frequency distribution,
using the categories in A78:A90 throughout the range and put any
overflow into cell B91. Note that this is sorting all the data points in
the range B39:AV65, the range where all the losering trades range expansions lie.
Figure H-4
Summary Statistics for Range Expansion. These formulas can be used

to generate the usual parametric stats for range expansion. (See Table 5-l .I Since
most of the value of range analysis comes via inspection, these are usually good for
spotlighting errors when they come up with some unusual values.
144
APPENDIX H
For the winners highlight C78:C91 and enter

= FREQUENCY(BB:AV35,A78:A90)
Your result should look like Figure H-3 if you turn Excels display options to Display Formulas.
From this point on, graphics are a snap. Select A77:C90, mmman a chart, change the chart type to column, and customize it to your
content.
I also put in a summary statistics section in these worksheet at
Cell A67. Figure H-4 shows the formulas used.
Appendix
MARTINGALES
This appendix gives the instructions for creating martingale models

in Excel. Begin with a blank spreadsheet, laying out the headings like
Figure I-l.
The first eight rows are shown here but only the first four involve headings. The computations beginning in row 5 will be taken
up in a minute.
Row 1 entries are text titles. Put them in the indicated columns.
In Row two, insert the value .5 in C2,l.O in D2 and $310 in F2. Complete the headings by typing in the text for TOW 4 in each indicated
column. Bet Table is in cell G4.
Figure l-l
Headings Layout. This screen capture shows the models headings.
Rows one through four are text or fixed values used in the calculations below.
145
146
APPENDIX I
The probability of a win can be varied later in the model; its set
to .5 here for convenience and because the resulting calculations dont
get out of hand. The value in D2, set to 1, is actually the ratio of the
average win (in dollars) to the average loss in dollars. This allows you
to adjust this crucial factor land 888 the impact on your equity and
loss drawdowns immediately. Last, F2 is set to $310, the MAE and avwage loss for the Crude data used in this book. This is a little harsh;
some trades could be losers without hitting the MAE stop. The model
itself deals mostly with trading units but this allows a convenient calculation of dollar impacts. You could make it worse by adding in slippage and commissions.
COMPUTATIONS
To enter the computations, row 5 is given the initial values where
needed and row 6 is given the formulations. Then, row 6 is just extended downward to row 24 by highlighting A5:F24. NOTICE that the
Bet Table in columns G and H is entered manually. Its values are not
formulas but specific values that Excel will look up depending on the
state of the equity units in column E. Ive created this table from a
complex martingale so that you can quickly look up your bet without going through the martingale logic. The martingale used is a
slightly modified version of that published by Bob Pelletier.*
Enter these values in row 5:
ColumnlRow
A5
B5
c5
D5
E5
F5
Value or Expression
1
1
=IF(B5=O,FALSE,IF(RANDO~C$2,LOSS,WIN))
_ IF~C5~LOSS.-B5.BSD$2)
=D5
=E5*F$2
* Pelletier, Robert C. Martingale Money Management Stocks and Commodities,

V. 7.3 (Seattle, 19881 pp. 69-72 and Money Management for Martingale Commodity
Traders The Journal of Commodit,y Trading, 4(Z).
MARTINGALES
147
Before entering expressions in rows six and below, build the

lookup table so you wont get any error messages when you enter your
formulas. The lookup table contains the bet size for most of the equity
balances youll run into, though not all. The models capacity can be
exceeded with long, involved runs of wins and losses, in which case
youll get an N/A error. Ignore it and recalculate. This simplification is preferable to the involved logic youd type in to get Excel to
fully simulate a complex martingale.
Enter this table manually beginning with Bet Table in cell G4.
Row
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Bet
Table
-45
~36
-28
-27
-21
-20
-19
-15
-14
-11
-10
-9
-8
-6
-5
-4
-3
-2
-1
0
10
9
8
8
7
10
8
6
6
8
5
5
6
4
4
5
3
3
2
1
To activate this table, the range it covers needs to be named.

Highlight G5:H24 and select Formula/Define Name.* In the dialogue
* Excel versions differ Bornewhat in defining names. Check your documentation to
get the specifics for your version.
148
APPENDIX I
that comes up, give the range the name Bet-Table using the lower
dash (SHIFT-Dash), not the normal dash on the upper right of your
keyboard. Click OK to close the popup window.
Next enter these values in row 6:
Column/Row
A6=
B6=
Value or Exmession
A5+1
IF(E5~0,OVLOOKUP(E5,Bet_Table,2))
The value in B6 will use the lookup table just entered

Now highlight C5:F24 and hit CMD-D. This should extend the
formulas down to row 24. Similarly, highlight A6:B24 and hit CMDD again. Your model should now look like Table I-l, although the values in the cells will be different.
EXPLANATIONS
Column A is a simple counter, for convenience only. Note that the

counter numbers and the models row numbers do NOT match.
Column B computes the bet size using the Bet Table which consists of all the values from G5 to H24. When a cell in Column B sees
that the Equity Units in Column E after the previous trade are zero
or negative, it looks through Column G of the Bet Table for the value
it sees in the Equity Units column. Then it picks the value adjacent to
that value in Column H. For example, trade 2 sees a 1 in the Equity
Units column. Going to the Bet Table, it finds ~1 near the bottom and
brings back the value of 2 that it sees in the adjacent column. Thus,
trade number 2 bets 2 contracts or blocks of shares.
Column C computes a random value between 0 and .9999. It compares that to the hurdle for winning set in cell C2 and decides whether
the trade is to be a win or a loss.
Column D takes the win/loss decision and computes the number
of units won or lost. If its a loss, its just the number of units bet. If
its a win, its the number of units times the ratio of win size to loss
size specified in cell D2.
MARTINGALES
149
Table I-l Completed Martingale Model. Endure eighteen trades to win

$310 while tying up 54,340 plus margin. Experiences like that minimize the
appeal of Martingales but al least they generally come out to the good.
PCWIN)
0.6
Assumptions: Win or
Trade
Bet
LOSS
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
1
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
Loss
Win
False
False
False
False
False
False
False
F&e
False
False
False
False
FASC
False
False
False
False
False
$ Win I $ Loss
1
contracts
Loss/Gain
Equity
Units
-1
2
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
-1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
Bet Size
$310
Equity Bet Table
$(310)
310
310
310
310
310
310
310
310
310
310
310
310
310
310
310
310
310
310
310
-45
-36
-28
-27
-21
-20
-~19
-~15
-14
-11
-10
-9
-8
-6
-5
--4
-3
-2
~1
0
10
9
8
8
7
10
8
6
6
8
5
5
6
4
4
5
3
3
2
1
Column E adds the win or loss in Column D to the previous value

in Column E, creating a running total of the loss and eventual win.
Column F, just for convenience converts the units won or lost to
dollars.
USE
This model can show you very quickly the good and bad points of the
martingale. Since it randomly generates wins and losses, all you need
to do to see another scenario is hit your Calculate Now command
150
APPENDIX I
(CMD =). Once youve become familiar with the base cam of 50% winning probability and $l:$l win to loss size, see whats important in
managing a martingale by tweaking the probability of winning and
the ratio of wins to losses.
For some analytical adventure, turn the win/loss ratio into a random variable mapped between 1 and 5 and skewed to the left between
1 and 3, a pretty typical trading situation.
Appendix
APPLYING MARTINGALES ~0
T RADING C AMPAIGNS
This appendix gives Excel code to allow comparison of the equity

curves for a trading combination with a simple MAE stop and the
same combination with a martingale applied to it. The equity curves
that result from this analysis can be converted to returns series and
analyzed for volatility and risk/reward ratios. Plotted as they stand,
though, they produce an excellent visual comparison which most will
find almost fully informative.
This example is set up slightly differently than those in the earlier appendices. The first 64 rows are given over to the data itself in
columns A through E. The columns, however, are the same as those
arranged in Appendix F through Column I, the MAE computation. If
you have this model haul it out and discard all columns after I, not
forgetting to save the revised model under a different name.
Next go to cell F63 and enter, left to right the following titles in
the columns as shown in Figure J-1.
Now, two initial values for these columns must be set. In cell J61
enter the value .31. Later you can change this to adjust the model to
151
152
APPENDIX
Figure J-1
Headings for Martingale Analysis. The models first five columns and
63 rows are given over to data to feed the trading rules. Titles for the analysis are
in cells Fh3:063.
Figure J-2
Bet Table for Crude. Equipped to handle adverse runs up to twelve
losses and 23.trade martingales. the Crude bet table is much longer than the cimulations bet table in Appendix I.
APPLYING MARTINGALES TO TRADING CAMPAIGNS
153
your needs. In cell 059 enter the title Margin in $; in 060, enter
$2000; and in 061, enter the formula * = 059/10 without the quotes.
This assumes your tradable trades at $10 per tick. Adjust as necessary to compute margin in trading points.
One other preparation is necessary before filling in the formulations for the day-to-day computations. That is to create the Bet Table
as in Appendix I. This table must be filled in manually because its
values are specific to the martingale I used. I began it at cell 662 with
the title Bet Table. Enter it into the cells Q62:R96 as shown in Figure J-2.
This table is so much larger than that of the simulation model
because Crude goes on longer runs!
Beforeleaving the bet table it must be defined to Excel. Highlight Q63:R96 and then select from the menu bar FORMULA/DEFINE NAME Enter the name Bet-Table (without the quotes)
and click OK. From now on, the model can refer to this entire table by
the name weve just given it.
Next, move to the following cells and fill in the formulations
specified:
Cell
Formulation
F64 =IF(E64>E52,IF(E64>E4,1,O),IF(E64<E4,-1,0))
G64 =IF(F63< >F64,IF(F64=1,-E64,IF(F64=-1,E64,O)),G63)
H64 FALSE
164 FALSE
564 0
K64 =IF(H64< >FALSE,IF(H64<0,IF(K63=FALSE,-1,
-VLOOKUP(L63,BET_TabIe,Z)),IF(K63
=FALSE,FALSE,VLOOKUP(L63,Bet_Table,Z))),IF(L63>
=O,FALSE,IF(M63>=O,FALSE,K63)))
L64 =IF(K64=FALSE,FALSE,IF(H64< >FALSE,K64
+L63,L63))
M64 =IF(L64=FALSE,FALSE,IF(H64<0,IF(J$61>164,
-H64*K64+M63,J$61*K64+M63),H64*K64+M63))
N640
064 =IF(G64< >O,IF(K64<0,0$61-K64*0$61,0$61
+K64*0$61),FALSE)
154
APPENDIX J
H64,164,564, and N64 all held initial values but, for the rest of
the rows need formulas. To avoid typing row 65 again select F64:065
and hit CMD-D. This will fill in row 65 with the formulas from 64
plus some initial values which youll now overwrite with specific formulas. Enter the following formulas in the cells indicated.
Cell
Formulation
H65 =IF(G65< >G64,IF(G64>0,G64-E65,IF(G64<O,G64+E65)))

I65 =IF($G65< >O,IF($G64< >O,IF($H64=FALSE,IF
($G64<0,MAX(O,-$G64-$D65,$164),MAX(O,$C65
-5G64,$164)),IF($C64<O,MAX(O,-$G64
mm$D65),MAX(O,$C65 ~$G64)))),1F($H65<
>
FALSE,IF($G64<O,MAX(O,m$G64
-$D65,$164,,MAX(O,$C655$G64,$164))))
565 =IF($H65< >FALSE,IF(J$61<$165,564-J$61,564
+$H65),J64)
N65 =IF(H65=FALSE,N64,IF(M65=FALSE,N64+565
-J64,M65-M64cN64))
Wow! Seeing the actual logic behind simple trading ideas is astonishing!
Now, checking to see the last row of your data, (lets say row
3000, for example) select F65:03000.* Then hit CMD-D. Excel may
complain that the selection is too large to do with an Undo, but bit
OK. Once it has copied itself, select Calculate Now (if you dont have
Automatic Calculation turned on) and your equity curves are deposited in columns J and N.
With the various charting sequences in various packages of
Excel, I forbear instructing on creating a chart but the most foolproof
way is to make line charts your default chart type, select and move
the two columns adjacent to each other, select the range you wish to
chart and invoke the chart wizard. In Excel 4.0 for the Mac, the last
efficient version of Excel, hit CMD-N, C, and return. Voila!
* An easy way to do this is put your cursor in cell F65, select GO TO and enter 03000
in the popup box. BEFORE you close the box, hold the shift key down and click OK.
The entire area will he selected for you.
APPLYING MARTINGALES TO TRADING CAMPAIGNS
155
EXPLANATION
Columns F through I have already been explained in earlier Appendices. The Trade column computes whether a dual-moving average
trade is on or off and in what direction. You can change this formulation to use the model for different trading rules. Column G computes
and holds the positions entry price, negative (for cash outflow) for
longs and positive for shorts. When the trade ends Column H computes
the win or loss. Column I computes the maximum adverse excursion
on each day of the trade and its final value.
Column J keeps a running total of the wins and losses, subject
to the .31 MAE stop, resulting in the equity curve of the trading combination. Each time there is a closure of a trade, Column J checks to
see if the adverse excursion exceeded the .31 level. If so, the trade is
marked a loss of .31 even if it was a winning trade originally. This accounts for the negative impact of MAE-or any-stop.
To get to the martingales equity curve, Column K waits until a
losing trade is recorded. Then it computes the number of contracts or
share blocks that were risked on the trade in units. These are the
numbers used to track the martingales sequence. Later as the martingale proceeds through more losses, it recomputes the number of
contracts to be risked by looking up the values in the Bet Table. Using
the bet table is preferable to encoding the elaborate martingale logic
within the limitations of Excel. The bet table holds valid amounts to
risk for the most common values in the martingale used here. If you
want to experiment with a different martingale, youll need to make
the Excel logic or create a similar table by going through various
win/loss sequences.
Column L is a running total of the contracts or share blocks that
have been won or lost. If this running total turns to zero or positive
the martingale is ended. Similarly, Column M keeps a running total
of the trading points gained or lost during the progression of the martingale. If this turns positive the martingale is ended, placing the
trader at a new high level of profitability.
Column N keeps a running total of the equity from the trend
trade as impacted by the Martingales results, in trading points. Winning and losing trend trades plus all the wins and losses of the martingale are summed here to create the martingales equity curve.
156
APPENDIX I
Lastly, Column 0 computes from the number of blocks being traded in

Column K the margin outstanding to support the Martingale. The per
contract margin is taken from cell 061 where it can be adjusted to reflect your tradable.
Though it would take some expertise with Excel to modify the
trading rules, it should be fairly easy for different date to be cut and
pasted into the first five columns. Remembering, then, to adjust the
stop value in J61 and the margin in points in 061, the model should
be ready to serve in seeing the martingales impact.
INDEX
Add-on trades, 74, 75, 76, 77, 105

Aggregation, 21-25
Betting strategies, 83-84, 86
Bin(s), 27, 31-32, 33-39
defining profit by, 33-39
size, 31-32
rule of thumb (1% of trading
capital), 32
Campaign trading, 74-83
applying martingales to (Excel
code; Appendix J), 151-156
impact of runs of losses on, 74-83
Campaign Trading!, 5, 8, 28, 57,
66, 76, 93, 105

Capital:
conservation, 63-64
loss limit (2% rule of thumb), 68,
74,84
Checklists:
management performance (Figure
S-31, 103
trader performance (Figure S-21,
102
Closes, using, 105-106

Control loop, trading management
(Figure S-l), 101
Countertrading, 74, 75
Crude oil trading, and martingales,
93-98
Data:
collecting (for MAE), 21-25
slicing, 23
Source for examples, 5
Distribution, normal, 59
Donchian Rule, 77
Drawdown, see Equity drawdown
Entry trades, 74, 75
Equity, account, and single trading
tactic, 64-68
Equity curves, correlated, 77-83
Equity drawdown, 64-68
computation, 68-69
rule of thumb for suspending
trading (40% reduction),
69
Equity lines, trading tactics, 76
157
158
INDEX
Excursion(s), price:
adverse/favorable, 10-15. See also
Maximum adverse excursion
(MAE)
key assumption (time frame),
10
consistent (Figure 2-31, 9
measuring from point of entry, 4,
8
Experience, trading, 3-4
Frequency diagrams/distributions,
25-32
bins, see Bin(s)
calculating/generating
(Appendix
D), 119-122
example (Figure 3-21, 26
and long/short, 32
Indicators, related, 105
K-ratio, 82
Losers/winners, see Winners/losers
LossCes), runs of, see Runs effects
Loss limit (2% rule of thumb), 68,
14,84
MAE, see Maximum adverse
excursion (MAE)
Management, see Trading
management
Management performance checklist
(Figure S-3). 103
Market behavior, and zebra
analogy, l-4
Martingales, 85-98
applying to trading campaigns
(Excel code; Appendix J),
151-156
complex, 88-93
creating model in Excel
(Appendix I), 145-150
on crude oil, 93-98
margin requirements, 97
simple, 87-88
simulation (Appendix J),
E-156
strengths and weaknesses,
97-98
MaxFE, see Maximum favorable
excursion (MaxFE)
Maximum adverse excursion
(MAE):
collecting data for, 21-25
sample list (Table 3-l), 22
defining, 7-20
displaying, 21-32
examples, short/long (Figure
2:5),12
exemplary Excel code (Appendix
A). 107-111
in fre&cydiagrams, 25-32
us. profit/loss (Figure 3-l), 24
us. range at entry, 55-56
sample calculations (tabular
examples), 16-20
long (Table 2-11, 16
short (Table 2-21, 17
for shorts and longs, 12, 32,
123-126
and trading decisions (placing a
stop or reversing), 28-31
and trading management, 99-106
winning trades/losing trades,
seminal observation about,
23
Maximum favorable excursion
(MaxFE), 10, 11-15
2-61.13
B). 112-115
sample calculations:
long (Table 2-3). 17
short (Table 2.-41, 18
MinFE, see Minimum favorable
excursion (MinFE)
INDEX
Minimum favorable excursion
(MinFE), 10, 11-15
Z-7), 14
Cl, 116-118
sample calculations:
long (Table 2-61, 19
short (Table 2-71, 20
Moving 20-day average, and range
volatility, 46-55
Path, typical (of a trade), 104-105
Point of entry/point of exit
(measuritig price excursion
from), 4
Price chart, standard (Figure 2-21,
9
Price excursion, see Excursion(s),
price
Price ranges, 41. See also
Range/range volatility
Profit curves. 34-37
absolute value (Figure 4-21, 37
computing (Appendix F), 127-132
example (Figure 4-l), 36
summary display (Figure 4-31,
38
and widening/tightening stops,
39
Profit/loss us. MAE (Figure 3-l),
24
Profit tradeoffs, 33
Progressions, 63, 84
Range/range volatility, 41-62,
133-137, 138-144
anecdotal evidence (Appendix G),
133-137
and classically defined volatility,
42-46
MAE us. range at entry, 55-56
and moving 20.day average,
46-48
159
and moving 20.day average on

date of entry, 48-55
normality, 58-61
at point of entry, 45-46
range at entry us. range after
entry, 48-55
range excursion (Appendix H),
138-144
range expansion, and widening
stops, 61, 62
summary, 62
Redefining winners and losers, 105
Reversal trades, 74, 105
Reversing, 28-31
Rules, see Trading rules, defining
Runs effects, 63-84
on betting strategies, 83-84
on campaign trading, 74
computation of percent reduction
in trading capital plus
winnings, 68
conservative standard, 69
disaster, estimating likelihood of,
69-72
frequency of rum, 65-68
on particular trading tactic,
64-74
tule of thumb for suspending
trading (40% reduction), 69
sirrnificance of losses from
adverse runs, 73
Scientific method, 2-3
Sharpe ratio, 82
stops:
placing, 28-31
and profit curves, 39
and range expansion, 61
and range volatility, 42, 55
tweaking, 15-16
Strategies/tactics, 74
Tactics, 64, 74
correlations between, 75
160
INDEX
Trader performance checklist

(Figure a-21, 102
Trading combinations, 77-83
Trading journal, importance of, 3-4
Trading management, 64, 99-106
campaign trading, 74-83
control loop (Figure E-l), 101
day-to-day trading, loo-104
elaborations, 104-106
closes, using, 105-106
indicators, related, 105
redefining winners and losers,
105
typical path, 104
portfolio impacts, 99-100
Trading rules, defining, 3, 4-5, 8
Trading strategies/tactics, 74-75
Transaction costs, and martingales,
87
Trend equity and add-on equity,
compared (Figure 6-71, 76
Trend trading, 76, 77

TweakCsYtweaking, 15-16,41-61
data slicing, 23
and range volatility, 41
Typical path, 104
Volatility changes, impact of,
41-62. See also Range/range
volatility
Winners/losers:
price excursion charts of, 8-10
and range expansion, 48-55
redefining, 105
seminal observation about MAE,
23
thirty or more (rule of thumb), 21
Zebra analogy of market behavior,
l-4

Maximum Adverse Excursion

Uploaded by

Copyright:

Available Formats

Maximum Adverse Excursion

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Maximum Adverse Excursion

Uploaded by

Copyright:

Available Formats

What is the book about?

What is the book about?

What trading concepts are discussed in the book?

What trading concepts are discussed in the book?

IMI AXIMUM

JOHN WILEY & SONS, INC.

This text is printed an acid-free paper.

Reproduction or translation of any part of this work beyond

Maximum adverse excursion : analyzing price fluctuations far

THE TRADERS ADVANTAGE

The Traders Advantage Series is a new concept in publishing for

THE TRADERS ADVANTAGE SERIES PREFACE

THE TRADERS ADVANTAGE SERIES PREFACE

ket, there is a risk. Although price volatility is often used to express

THE TRADERS ADVANTAGE SERIES PREFACE

equity moves, as well as the favorable, profitable ones. It is this

Adverse Excursion, Favorable Excursion

Impact on a Particular Trading Tactic

Impact on Campaign Trading

Impact on Betting Strategies

Martingales on Crude Oil

Imagine that you are a prehistoric hunter at an African waterhole.

stricted markets. As a result, our knowledge of how markets will act

IMPACT ON CAMPAIGN TRADING

DEFINING MAX ADVERSE EXCURSION

Standard Price Chart.

is a set of excursions from

DEFINING MAX ADVERSE EXCURSION

Excursion, then, is just the change in price from our point of

ADVERSE EXCURSION, FAVORABLE EXCURSION

= MAX[O, (Highest subsequent high-Entry price),

and, for shorts:

= MAX[O, (Entry price - Lowest subsequent low),

For minimum favorable excursion, if youre long, calculate:

DEFINING MAX ADVERSE EXCURSION

ADVERSE EXCURSION, FAVORABLE EXCURSION

DEFINING MAX ADVERSE EXCURSION

(Top) Long MinFE Example. As the NASDAQ composite rises steadily

Treasury Bonds in the late 1980s but found no noticeable difference

Table 2-3 Calculating MaxFE When Long. MaxFE grows steadily as a

DEFINING MAX ADVERSE EXCURSION

Minimum Favorable Excursion tracks the least favorable price

*A long is represented by a negative price-the cash outflow in taking a long position.

Tweaking Data Slicing

Inspecting these numbers, notice that the winning trades have

When selected and charted, they should resemble Figure 3-1.

be lumped together. Then those greater than .2 and up to .3, and so

l-l: 0.4l-l: 0.5rl;

What you typically find is that the distribution of losers has a

Maximum Adverse Excursion

ever had adverse excursions of more than .45, a remarkable thought

MAE in Trading Points

MAE in Trading Points

Theres clearly a tradeoff (which discussed in Chapter 41, but the

twenty sparsely populated bins. Theres no reason it should always

It will have struck stock traders that I havent distinguished between

DEFINING PROFIT BY BIN

* I me profits in the general tense here. Losses are negative profits.

= 0.16 + 0.07 + 0.5

Excel code for calculating all this on an automated basis, using