Sunteți pe pagina 1din 52

jan/feb 2007

The publication for trading and investment professionals

Jack Schwager

Volatility Strategies:
Trading the VIX

Measuring Alpha
and Beta

HSBC Investments
Inside a Trend Following Fund

Markets Software Books

Outlook for Automated trading with Hedge Hogging with
FTSE 100 Barton Biggs
We are pleased to be holding our second annual European Conference in February. This
year includes two talks on automated trading systems - always a subject that generates
interest - plus coverage of more traditional trading techniques. With our talks on the global
markets outlook, we also hope to nail down the technical outlook for the FX and equity
markets this year given the uncertainly that has surrounded the dollar and US
stocks in recent months.

We look forward to seeing you there and hope you enjoy this issue of the magazine.

Matthew Clements, Editor.


FTSE 100 outlook
The technical view for UK stocks remains
broadly positive although, as Steven Wesiak of
ABN Amro explains, there are certain
levels that need to be watched for
possible signs of danger.

A Trend Following Fund

Charles Morris of HSBC Investments in London
explains the workings of his trend following
equity fund and how he selects the best
global stocks.

Jack Schwager
Jack Schwager is a well known figure in the
> 31
world of trading and analysis. He discusses his
trading approach and that of the traders he has
met writing his famous “Market Wizards” books.

© 2007 Global Markets Media Limited. All rights reserved. Neither this publication nor any part of it may be
reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical,
photocopying, recording or otherwise, without the prior permission of Global Markets Media Limited. While the
publisher believes that all information contained in this publication was correct at the time of going to press, they
cannot accept liability for any errors or omissions that may appear or loss
suffered directly or indirectly by any reader as a result of any advertisement, editorial, photographs or other
material published in The Technical Analyst. No statement in this publication is to be considered as a
recommendation or solicitation to buy or sell securities or to provide investment, tax or legal advice. Readers
should be aware that this publication is not intended to replace the need to obtain professional advice in
relation to any topic discussed.

January/February 2007 THE TECHNICAL ANALYST 1

Inside a trend following fund Monday blues

10 26


Editor: Matthew Clements
Managing Editor: Jim Biss
Consultant Editor: Trevor Neil
Advertising & subscriptions: MARKET VIEWS
Louiza Charalambous FTSE 100: positive trend to continue? 07
Marketing: Vanessa Green Euro dollar: bullish outlook ahead 08
Events: Adam Coole
Design & Production:
Paul Simpson & Thomas Prior
The Technical Analyst is published by Inside a trend following fund 10
Global Markets Media Ltd
Unit 201, Panther House, Spreading financial futures 15
38 Mount Pleasant, Trading the DAX open 19
London WC1X 0AN VIX strategies 22
Monday blues 26
Tel: +44 (0)20 7833 1441
Jack Schwager 31

Subscription rates (6 issues) SOFTWARE

UK: £160 per annum
Automated trading with Futures Betting 34
Rest of world: £185 per annum
Electronic pdf: £49 per annum
For information, please contact: BOOKS
Hedge Hogging with Barton Biggs 37

For information, please contact: AUTOMATED TRADING SYSTEMS
Measuring alpha and beta using Excel 39
A machine-learning method for automated trading 43
Art, design and typesetting by
all-Perception Ltd.
Printed by The Friary Press


January/February 2007 THE TECHNICAL ANALYST 3

Industry News

Reuters offer news DOW JONES

for automated trading ‘87.5% ACCURATE’
Reuters has launched Reuters mated trading systems. The product
NewsScope, a new product allowing enables computers to recognize and
its news output to be "read" by auto- process items in Reuters news stories
which can then be incorporated into
automated trading strategies.
NewsScope includes two products:
"Real-time" which lets users feed live
news content into automated trading
systems and respond to market-mov-
ing events as they occur. "News
Scope Archive" lets users replay sto-
ries as they unfold in the market
allowing them to back-test trading


BLACK BOX FX FUND Dow Jones Newswires has reported

Barclays Capital in London has weightings. According to Philippos that its Charting Europe technical
launched a new currency fund track- Kassimatis, Global Head of FX analysis column, written by Axel
ing their Intelligent Carry Index Structuring, "This is a black box sys- Rudolph, was 87.5% accurate in
(ICI) which targets the yields on the tem that removes any discretionary making market forecasts in
world's most liquid currencies. The element from trading. It is optimised November. They say that
ICI fund is run using an automated for maximum returns allowing for a GBP/USD was most accurately
strategy with trading rules that are 5%-volatility level of risk which forecast with cumulative gains of
based on the optimisation of a port- means the positions are left alone 260 ticks on the month based on
folio of the G10 currencies. It then throughout the month regardless of their calls. The average accuracy for
takes long and short positions once a what is happening in the markets". Dow Jones technical forecasts for
month based on these optimised 2006 was 73.4%.

New UK mobile quotes from MarketSource

MarketSource, a global provider of exchanges. Markets available are
mobile financial quotes, has launched futures, stocks, commodities, credit
a new service for UK traders. markets and FX. Financial charts,
Available via purpose built pagers market moving news, and price limit
and mobile phones, MarketSource's alerts are also included. For more
products, G3 and iSpy, are powered information, visit:
by live data feeds from global trading

4 THE TECHNICAL ANALYST January/February 2007


European Conference 2007

Vintners Hall, London EC4 - 7th February 2007
Strategies for trading the global markets

The Technical Analyst magazine is proud to present its 2nd Annual European
Conference for traders and investment managers. This year’s event brings
together the very best domestic and international experts to speak on a wide
range of important strategies in the world of technical trading. Including talks on
mechanical trading plus a panel discussion taking pre-submitted questions from
delegates, this is an essential event for Europe’s trading and investment

Who should attend: Topics covered:

+ Traders + Global markets outlook
Delegate fee:
+ Fund managers + Elliott Wave strategies £395 + VAT
+ Hedge funds + Trend following
+ Market analysts + Mechanical trading
+ Risk managers + Ichimoku charts To book online, please visit:
+ Brokers + TA question time

Speakers include:

Robin Griffiths John Noyce Tom Denham Trevor Neil Jeremy du Plessis Adrian Hughes Florian Leder
Rathbones Citigroup Elliott Wave Int. Betagroup Updata Societe Generale Amplitude Capital
+44 (0)20 7833 1441
Market Views


by Steven Wesiak

he uptrend that started back in
2003 is still robust and showing
no sign of an imminent rever-
sal. Objective indicators such as RSI,
MACD and momentum also give no
hint of any significant decline. That
said, levels are pushing into an area
where the going could get tough, lead-
ing to a choppy market phase. That is
namely the former head-and-shoulders,
or rounding top. Volatility could start
increasing at 6360.30 as the market tries
to make its way up through that thick completion of a head-and-shoulders peak of 6360.30. Slipping below
resistance zone, which extends to the bottom in the weekly chart. The pat- 5985.20 could tempt sellers to knock
6950.60 all time high. tern itself implies an eventual rise to out the last bottom at 5467.40. If that
A look at the 3 month volatility chart 5.38%. The timing though is difficult to fails to fend off a bearish assault then
(Figure 1) shows a possible up tic in the nail down. This assessment only ends if the uptrend would be severely compro-
works after a two year period of relative yields head back under 4.44%, which at mised. A decline as far as 5077.60 can
calmness. One external factor that the moment does not look likely. be expected before any new base starts
could eventually start weighing on The first sign that equities are correct- to build.
prices is the expanding up trend in ing lower comes with a break below However, the absence of any top
yields in Europe and the USA as well as 5985.20, which could occur if the mar- forming in the FTSE, as well as an
Japan. For example, gilt yields show the ket is unable to penetrate the nearby absence of any clear reversal signals
from objective indicators keeps the
focus on an extension of the up trend.
Once the market makes its way up
through the 6360.30 to the 6950/60
resistance zone, projected possible tops
at 7243.58 and 6436.75 come into
focus. The projections are based on a
Fibonacci extension of the up trend
but are not tested resistance levels.
Therefore they are but a guide to where
potentials tops could form.
In sum, the uptrend remains robust
but the chance for a correction lower
grows as the market pushes into the
former reversal pattern. Also, rising
yields could eventually dampen the
enthusiasm for equities.

Steven Wesiak is technical analyst

with ABN Amro.
Figure 1.

January/February 2007 THE TECHNICAL ANALYST 7

Market Views

by David Linton

ne of the great advantages of aspect is we have the cloud projected ment will be very risky as anything
Ichimoku charts is that you into the future giving us forward counter trend is likely to be very small
can get an instant picture of knowledge of where the trend is head- and short lived. We have found key
what is happening on multiple time- ing. support on the daily chart cloud base,
frames at a glance. Ichimoku resolves Certain time horizons may not match right where we would expect to, so we
your time horizons by looking at your own, but they can provide you should run to a 'higher high' in the mid
monthly, weekly, daily and hourly charts with very useful information. For 1.30s to test the 2005 high. Longer
as we see here in Figure 1 for instance, the monthly chart of term these charts tell us that the 1.40s
Eurodollar. I always look at all four Eurodollar may seem too long term but is a very possible area for Eurodollar to
time horizons on one screen for any it tells us that shorting the dollar in line move to. Remember to keep an eye on
instrument as a minimum whilst quar- with the bigger trend is likely to be all four time horizons simultaneously
terly and five minute charts can add to more profitable in the year or two for better timing of your trades
the picture even further. ahead. Meanwhile the hourly chart can
The key aspect of Ichimoku charts is be great for timing your entry if you
that of the price line fully crossing the have a more medium term view and
cloud from one side to the other as the can help you with a keener exit too. David Linton CFTe, MSTA is Chief
early signal. The lagging line (light blue So what does this screen tell as at the Executive of Updata plc which
line) is the confirming signal. The price moment? It's bullish for Eurodollar on provides technical software to users
may have more to do to be certain of a all four time horizons. Even taking a of the Bloomberg terminal. He can
full trend change. Another unique quick short position in this environ- be reached at

THE 1.40S IS A

Figure 1

8 THE TECHNICAL ANALYST January/February 2007

Energy Futures and OTC Markets. One Screen. ICE.

Introducing: NYBOT® Soft Commodity Contracts on ICE.

Commodities traders and risk managers rely on ICE.

ICE Futures’ suite of energy contracts, coupled with ICE’s
robust OTC energy marketplace, offers tighter spreads
and better fills – while increasing market transparency
and efficiency. So find out for yourself why traders
worldwide rely on ICE.

To learn more, contact

or call +1 (646) 733 5000 or +44 (0) 20 7265 3782.

Visit us at

Trading on the NYBOT is governed by specific rules as set forth by the Exchange. Those rules are subject to change. Contact a licensed broker for additional
information including commissions, fees and margin requirements.


10 THE TECHNICAL ANALYST January/February 2007


Charles Morris is a fund manager at HSBC Investments
in London where he heads a three man team running
absolute return strategy and the HSBC Global Trend
Fund. The fund, launched in May 2005, aims to
generate returns above the MSCI World Index.
He explains the strategies behind the fund’s success.

TA: What are the main strategies of low, sell high and add to winners. majority are average and only a few sur-
the Trend Fund? prise or disappoint. We need to identify
TA: How has the fund performed the strongest stocks and accelerate their
CM: We have around $2bn invested in since its launch? weighting in the portfolio and take
the absolute return service which com- money away very quickly from the los-
prises global equities, bonds and hedge CM: Up until May 2006 we had a great ing stocks. The challenge is to identify
funds. We have much less directional run beating the index by 15% or so and the best relative strength stocks and for
than a typical balanced portfolio. The it was relatively easy to make money. this we use a more quantitative, or
HSBC Global Trend Fund is our core However, since then it has performed mathematical, approach in our model.
equity strategy. We set this up in order poorly. Although many stocks in the
to pursue market leadership and it's index have performed ok, breadth has TA: What are your specific buy
basically a technical fund verging on been high and there have been fewer signals?
quants. By this I mean we look at charts real winners and trend following suc-
in a more numerical way instead of just cess depends on winners and in partic- CM: We use several techniques which
patterns. The objective of the fund is to ular, fat tails. The best thing to have include price highs, DMI crossovers,
beat the index; we make the decision of done since May last year was own the relative highs and others. There is a
which stock to own and how much we FTSE and ignore everything else. strong emphasis on technical analysis
should buy. For this we use relative here.
strength, timeliness and volatility. The TA: You mentioned quants. What is
overriding fact is relative strength. the quantitative element of the TA: How is low volatility, as indicat-
Firstly, we try and identify long term fund? ed by the VIX index for example,
trends using a model with stocks of a affecting the fund? If the S&P can
capitalisation above $3bn. We then look CM: We are increasing our emphasis still trend whilst exhibiting low
for all stocks with long term relative on money management rules which are volatility does this present a prob-
strength which we measure over a four very much on the quants side. I would lem for you?
year period. This is done by adjusting call it 'competitive allocation'. What we
everything into dollars and dividing by are trying to do is add to winners. If CM: I think it's certainly a problem for
the world index in dollars. This gives us you buy 10 stocks, you have a concen- trend followers. You can have low
a currency neutral ratio onto which we trated portfolio but as you add many volatility at the index level but at the
apply a moving average to rank the more names, say one or two hundred, stock level it's a different story. Here we
stocks. This helps ensure that we buy then you begin to see how the vast need stocks that are significantly →

January/February 2007 THE TECHNICAL ANALYST 11


outperforming the index, even if the

index is trending smoothly so to speak;
we won't get many of these in a low
volatility climate. Another problem is PROFITABLE IN EARLY BULL MARKETS.
that existing themes such as commodi-
ties and emerging markets are tired so DURING LATE STAGES OF A BULL MARKET…
it's difficult to see an area of market THEY DON’T REALLY MAKE MONEY.”
action that will pay active managers well
in the near future. Most active man-
agers are affected with trend followers TA: What stop losses do you use? from 50% at the beginning of 2005.
that are most heavily so. On the other Because trend following has started to
hand, if you are a relative value or arbi- CM: Rather than have hard and fast under perform, that must be telling us
trage based hedge fund then this situa- stop losses we tend to look at position that we are in a mature bull market.
tion won't worry you too much. sizing. For example we have just That said, certain indicators are sup-
reduced our holding of China stocks portive and there is no feeling of exu-
TA: As a trend follower, how do you and have been reducing oil for some berance in the market and valuations
recognize a market top? time. Otherwise, the ultimate stop loss are reasonable. There is also reason to
is the medium term low but our aim be cautious about areas that have done
CM: We look at sentiment, buying cli- would be to reduce exposure ahead of very well over the past 3 years or so
maxes and Economist magazine covers. this. We don't use percentage stop loss- which include emerging markets and
es because they tend to take you out of more significantly, the low interest plays
TA: Are techniques such as Elliott some volatile markets that perhaps you such as real estate and infrastructure.
Wave of any value given the longer should be in. For example, if you have Growth stocks are cheap and may still
term nature of your outlook? a 7% stop loss, then this a quite a lot for have some way to go and I am warming
a market that's trading in a range but its to tech.
CM: With Elliott Waves, the classic 1- nothing for something that's in a strong
5/ABC pattern happens all the time in trend. TA: What other ways can you extract
speculative bull markets; there's no value whilst trend following is under
doubt about that. Moreover, the longer TA: Are we near the top of the cur- performing?
term you look the more accurate they rent bull market in equities and
may be. However, it doesn't really fit commodities do you think? CM: There are several alternatives we
with the strategy of the fund. In peri- are looking at. These include volatility
ods of consolidation and bear markets CM: Actions speak louder than words. trades - buying volatility - because it
then it's a lot less clear. At the moment we are down to 22% in can't stay this low for ever. Rather than
equities in the absolute return service VIX options we will buy dispersion
baskets. When the stock market is ris-
ing with high breadth, and alpha strate-
gies aren't working, the way to own
equities other than via index funds is by
using structured products since options
are so cheap. The capital protection
reduces risks, and the low costs enable
high participation to the upside should
the bull continue.

TA: Do you look at market cycles?

CM: To some extent cycle theory and

such like is very well known now so it's
difficult to see how one can derive any
value added from it. Also, it's not that
relevant when it comes to stock selec-
Figure 1 - The attached chart shows the top quintile of high relative strength in the UK in blue tion.
and the fifth in red. Black is the market. This justifies trend following which works better in some
markets than others. UK and Europe are better than most. - Charles Morris

12 THE TECHNICAL ANALYST January/February 2007




TA: What about market timing?

Someone trading FX on a daily
basis is obviously going to have a
different approach than a long term
equity fund manager.

CM: The ideal trade is to buy a low

volatility stock with good long term rel-
ative strength on the dip. If you can do
that consistently, you will do well. The
problem often as far as fund manage-
ment is concerned is that technical
based market timing is a very tricky
area. Breakout investing tends to be
profitable in early bull markets. During
late stages of a bull market, which is
where we are at the moment, they don't
really make money. It has been proven
time and again that the most effective
strategy is asset allocation followed by
stock picking and then market timing.
If a market breaks out from a long
term trend, is this market timing or

TA: Would you also describe your

fund as partly automated?

CM: I'd call it a grey box automated

strategy. We are transparent about the
basic model and how it generates sig-
nals. We have also done a lot of back-
testing of strategies. However, we don't
have a fully automated fund and we still
have some discretion in what we do.

January/February 2007 THE TECHNICAL ANALYST 13

30 & 31 May 2007

Day 1 (May 30):

Day 2 (May 31):
+44 (0)20 7833 1441


by Stephen Aikin

Exchange traded financial futures need little introduction. They are well-used, highly liquid
instruments, offering regulation, standardisation, transparency and removal of counter-party
risk. They are usually the first point of liquidity for a fund or institution looking to establish
or remove an outright interest rate or equity position, but they also have another dimension
which is less frequented by the larger investor. This is the spread market, a relative value mar-
ketplace, where it is the relationship between two or more instruments that counts, not the out-
right direction.
Intra contracts month. Such spreads are a melting pot of open-interest
Financial futures spreads have two broad categories, intra- dynamics and short-term interest rate or repo rate influences.
contract and inter-contract, both being based on the two However, a larger, longer-lived intra-contract spread mar-
popular asset classes of interest rates and equities. Out of ket exists in the STIR futures markets. These are futures on
these two classes, interest rates offer the highest number ofshort-term interest rates and are arguably the largest markets
trading permutations and have the substantial benefit of a in the world by nominal value. It is by no means unusual for
mathematical dependency, something that is often forgotten the two largest contracts, the Eurodollar and Euribor, to
by equity traders. Intra-contract spreads are the simplest totrade in excess of one trillion of dollars or euros each day.
understand and follow. Perhaps the most common intra-con- Most financial futures only have one active delivery contract
tract spread is the bond future or index futures calendar - the front month but STIR futures can have up to forty, as
in the case of the Eurodollar. This means there are an enor-
spread; a short lived but highly active trade based around the
roll over from an "expiring" contract into the "new" front mous number of spread permutations within the futures
complex which can be traded
independently or relatively as a
butterfly or condor spread.
All of these different intra-con-
tract spreads are quoted as sepa-
rate instruments by the various
exchanges and so have separate
price histories to the outright
component futures. For example,
The Euribor June 2008 (M8),
September 2008 (U8) 3-month
calendar spread (M8U8) is quot-
ed as a separate instrument.
Many quote vendors and chart-
ing services have been slow to
introduce these price histories
much to the chagrin of techni-
cally based traders, preferring
just to offer the spread facility of
Figure 1. Euribor M8U8(BLACK) spread expressed as both the spread strategy and the displaying the chart as the simple
differential between the M8 and U8contracts (RED) - (Hourly) July to Dec 2006. Charts differential between the two
by Reuters Metastock futures contracts. Whilst →

January/February 2007 THE TECHNICAL ANALYST 15


this works fine for highly liquid nearer month contracts, a Schatz future, representing a single spot on the yield curve
problems can arise when charting spreads deeper into the against the two years of cash flows associated with the bond
forty-contract month complex. Although liquid, quite often future. Both trades are popular but the Euribor/Schatz has
these further dated contracts are traded less than the nearer the benefit that the majority of curvature is often contained
months and so the "last trade" might be minutes old rather within the first two year of the term structure, for the simple
than reflecting the current bid and offer levels. The combina- reason that it is difficult to forecast interest rate expectations
tion of two of these "last trades" into a spread can create much beyond this. However, popularity does not confer ease
chart levels that never existed and certainly never traded. This and these spreads can be complex things. Not only are there
might not sound like a big deal but given that some back risk factors such as curve movement and the credit spread to
month calendar spreads might only have a yearly range of ten consider, but also advanced pricing influences such as con-
basis points, then misleading "last prices" can extend this vexity, which can become a significant factor for longer dura-
range by twenty percent or more, making any charting stud- tions and in times of interest rate volatility.
ies invalid. These spreads are also further complicated by different tick
Figure 1 shows the Euribor M8U8 spread displayed in both sizes and, as in the case of Euribor/Schatz, a lack of margin
formats. It is immediately apparent that there is significant offset. The technical trader first needs to be able to chart
noise when the spread is shown as a differential, whereas the these spreads in a format that incorporates the different tick
more accurate spread strategy depicts a much smoother and values and duration neutral hedge ratios, whilst clearly show-
accurate study reflecting actual trades not price differences. It ing the effects on the P&L of a directional movement. One
is also apparent that the spread strategy is much more the best methods is to display the spread as a price spread
inclined towards technical studies than the differential. Every which incorporates the prices of the relevant instruments,
trader has their own favourite indicators but on charts like their quantities as determined by the hedge ratio and the
these, simple horizontal support and resistance lines work respective tick values per basis point.
well. The two blue lines shown work well with the block style
chart data that tend to characterise 3-month spreads. This is expressed as:
Inter-contract spreads offer a myriad of trading opportuni-
ties and the ability to adapt the trade to the technical indica- (Pbf x T x Qbf) - (Ps x T x Qs)
tor, rather than the other way round. As mentioned earlier, Where :
the common mathematical dependency of STIR and bond
futures on basis point movement permits a duration mix and Pbf is the price of the bond future
match approach to yield results that are much more techni- T is the tick value per basis point
cally flexible. Qbf is the number of contracts used
Ps is the price of the STIR futures.
The curve trade Qs is the number of contracts used
One of the most popular types of interest rate trades must
the curve trade. This is normally done as either duration neu- Figure 2 illustrates the visual representation of price spreads.
tral spreads, for example between the Schatz (2-year), Bobl The top graph depicts a Euribor/Schatz spread (Schatz Z6
(5-year) and Bund (10-year) futures or a STIR future versus versus Euribor U8 in a 1.4:1 ratio). Although the value of the

16 THE TECHNICAL ANALYST January/February 2007


Financial theory dictates that futures
strips, when spread against bonds,
need to be weighted to compensate
for the fact that nearer dated bond
cash flows are worth more since they
benefit from the effects of reinvest-
ment. This means that the nearer
dated coupon payments of bonds are
more important than further dated
ones and this is compensated for by a
higher weighting assigned to nearer
dated Euribor contracts. This is a rela-
tively easy mathematical procedure or
Bloomberg can do it on page TED
<GO>. However, this still leaves an
even more unmanageable spread con-
sisting of Schatz versus 8 weighted
Figure 2. The Schatz Z6/Euribor U8 price spread (BLACK) and an optimised Schatz
Euribor contracts but this can be
Z6/ Euribor 2-leg strip (RED) - (30 minute) October to Dec 2006. Charts by
reduced to as little as two weighted
Reuters Metastock
contracts by a process of optimisation.
Since most Euribor delivery contracts
spread is large, it has the benefit of behaving like an intra- are hugely correlated to each other, one can easy substitute
contract spread. An increase in the value of the spread will for another. There are several ways of optimising the strip.
lead to a profit for a long spread position - that is a long posi- Some like Principal Components Analysis (PCA) are highly
tion in the bond and a short position in the Euribor and vice mathematical and others are very simple, such as tweaking
versa. Furthermore, the difference between the purchase and trail Euribor weightings on any charting package that incor-
sale price of the spread will indicate the overall profit of loss porates composite strategies. Given that it is known that the
on the trade. nearer dated futures will be weighted higher than the further
This spread, a two-year cash flow versus a single 3-month dated ones, reduces time spent in this "trail and error"
forward rate, exhibits a lot of movement, representing the approach significantly.
combinations of both curve and credit spread shifts. Some The bottom panel in Figure 2 shows an optimised
traders might like such amounts of variance but often such Schatz/Euribor spread. Instead of being a duration neutral
moves are sudden, unpredictable and large and applying tech- spread between Schatz Z6 and Euribor U8, it is now shown
nical indicators to try to interpret these movements can be a as being a spread between Schatz Z6 and a weighted combi-
fruitless exercise. Any indicators that might appear to work nation of Euribor H7 and U8 in a 10:8 weighting (versus 25
might necessitate the trader being in drawdown on a position Schatz). It can be clearly seen how much less volatile this
merely due to the variance involved. However, instead of try- spread is compared to the upper black version and yet no
ing to find an indicator or technical technique to tame this more contracts were used, just redistributed. Once again,
curve spread, why not tame the spread to fit an appropriate every trader has their own favourite indicator but it is visual-
indicator? In this case, this can be done by modifying the ly apparent that popular studies such as line indictors, mov-
spread so that both sides become similar representations of ing averages or oscillators are more applicable to the opti-
the curve. mised spread than the plain one.
At present, the variance of the spread stems from the curve Hopefully, this article might encourage the reader to con-
(and credit) movement due to the fact that a 2-year cash flow sider the financial futures spread market not only as a mar-
is being matched with a 3-month forward period. If the ketplace offering many different trading permutations, but
Schatz were spread against a Euribor bundle (eight sequential also as a unique facility to adapt the trade to the indicator
delivery months) then much of this variance would disap- rather than the other way around. Variance can be increased
pear. However, bundles are not quite the solution they might or decreased to order, in line with a trader's own style, some-
appear. Not only are they a little clumsy to manage, they are thing not readily available in the outright markets.
not an entirely accurate price match due to differences in
duration matching, stub exposure and the fact that a bundle Stephen Aikin is a proprietary trader, specialising in rel-
entails a equal number of contracts per delivery month. ative value trading and author of "Trading STIR
futures", published by Harriman House (Nov 2006)

January/February 2007 THE TECHNICAL ANALYST 17


18 THE TECHNICAL ANALYST January/February 2007



by Mircea Dologa

How to interpret and trade gaps in the German stock market.

ome traders argue that the first ping back, the profit potential for these inception of a new trend. It represents
hour is the novice's and the last trades is particularly attractive. acceleration in the market. Even if
hour belongs to the professional. there is strong news it will not have
I disagree. The opening price carries Understanding the gap been fully discounted and the trend
with it the weight of all that has hap- An opening gap in the Dax can be should continue in the direction of the
pened overnight, whereas the closing caused by fundamentals generated out- news (see Figure 2).
price is not nearly as important. side of the Dax (such as international
Opening prices are a complex reaction economic and/or political news), Large Gaps - often over-extensions of
to a raft of news and market informa- developments in other markets since the market which tend to be followed
tion that has been released prior to the the Dax closed (such as EUR/USD, by a severe correction and a re-asser-
open, and it is the prepared and well Nikkei, US markets and oil), as well as tion of the pre-gap trend. This usually
informed trader that will be best able to overnight developments related to the happens when the gap measures over
profit from the market in this session. Dax itself (such as overnight news 80% of the daily ATR(21) (see Figure
In trading the opening, my preferred relating to Siemens for example). 3).
market and instrument is the Dax 30 Crucially, the size and location of a
Future. This is because the Dax is rela- gap is important to its interpretation Placing stops
tively volatile (the 21 day daily average and trading potential: One of the critical decisions in how to
true range is around 70 Dax points (c. trade large opening gaps is knowing
$2275) compared to only 11.5 for the Small Gaps (less than 10 Dax points) - where to place your stop-loss so that it
S&P 500 e-mini (c.$575), a 4 to 1 usually signal the continuation of the can account for the fact that the market
advantage). Furthermore, the Dax prior day's trend (see Figure 1). will often continue in the direction of
works more like clockwork than any the news for several minutes or hours,
other market I know. It tends to have Medium Gaps (less than 25 Dax without risking too much capital.
two daily consistent swings - in the points) - often a breakout gap, i.e. the My observation is that - more →
early morning and in the afternoon
(around 14.30 to 15.30 CET) - and
reversals tend to happen at quarter-,
half- or full hour intervals. The size of
the many intra-day swings is consistent-
ly around 15 points and several times a
month the Dax trader is treated to a
major intra-day trending move of more
than 80 points. Such consistent intra-
day trends simply do not exist on the
S&P 500.
The trade I particularly look for is
one that exploits an opening gap (i.e. an
opening price that is significantly high-
er or lower than the previous day's clos-
ing price) and the Dax offers many
textbook examples of these. Opening
gaps can expand as much as 200% and
as these large gaps tend to be quickly
filled, much like a rubber-band snap- Figure 1.

January/February 2007 THE TECHNICAL ANALYST 19


often than not - the effect of external

factors tends to wane by the end of the
morning session and then internal chart
factors that influenced the Dax prior to
the gap re-assert themselves, snapping
the Dax back to where it was before the
gap as if the gap had never happened.
Unfortunately, such an observation
won't necessarily help you set your
stop-loss level when the trade is
entered. (In fact, given the difficulties
involved with this part of the trade I
prefer to give myself two bites of the
cherry by splitting my allotted capital
into two halves, i.e. if the first trade is
stopped out I go in again at a newer
level or conversely I double up when
Figure 2. the trade goes in my direction).
Certainly, an understanding of what
has happened overnight (or over the
weekend) will give you a clue as to the
“THE TRADE I PARTICULARLY LOOK FOR likely strength and direction of any
expected gap and its subsequent follow
IS ONE THAT EXPLOITS AN OPENING through, and therefore be able to help
GAP…AN OPENING PRICE THAT IS you set a suitable stop level. Also, there
are many technical indicators and tools
SIGNIFICANTLY HIGHER OR LOWER for identifying when the market is over
extended and about to turn. My pre-
THAN THE PREVIOUS DAY'S ferred methods involve Andrew's
CLOSING PRICE.” Pitchfork, Elliott Wave, Fibonacci and
Gann, all of which should help identify
objective targets for the impulsive/cor-
rective swing and also the probable ter-
mination level of the trend.
I would also stress the importance of
looking at the pre-close trading zone (a
period of at least two hours before the
close). The study of the pre-close of
the prior day's trading zone will reveal
valuable information about the next
morning's moves and also about strong
resistance and support levels.

Dr Mircea Dologa, MD, CTA is a

commodity-trading advisor who
founded a new teaching concept
for young and experienced traders
He can be contacted at:

Figure 3.

January/February 2007 THE TECHNICAL ANALYST 21



Krag Gregory discusses how VIX futures positions would have reacted
under different market conditions.

he VIX is a constant, 30-day ments. return was negative. To conduct this
benchmark of expected realized analysis we replicate one-month VIX
market volatility as measured by Being short VIX futures has paid futures and estimate the payoff at expi-
S&P 500 index option prices. VIX off: ration by going long $1 notional each
futures are standardized futures con- (1) On average over the last decade. calendar month from February 1996 to
tracts on forward 30 day implied Over the 127 months in our study, the August 2006.
volatilities. VIX futures have only been S&P 500 was up about 61% of the time
listed since 2004 and the short history with an average monthly return of VIX future - VIX spot basis can
has made it difficult to draw substantial 0.66%. Given the positive return envi- amplify gains or exacerbate
trading conclusions. Nevertheless, here ronment, short VIX futures strategies losses
we estimate VIX futures back to 1996 had an average gain of $0.28 per dollar We break down success rates and aver-
and quantify profitability under differ- notional invested. age profit or loss from long calendar-
ent market environments. We believe month VIX futures positions to identi-
VIX products provide an efficient way (2) In up markets. In months when fy historically successful trading strate-
to monetize directional volatility views the S&P 500 return was positive, the gies:
and as VIX options are driven by the average gain on a short VIX futures
underlying future, understanding the position was $1.6. • In months when the S&P 500 was
dynamics of the VIX future is crucial down over 3%. Long one-month
for successful trading. After backtesting (3) At good entry points. The VIX is VIX futures trades were profitable
trading strategies across different mar- mean reverting, and the largest gains 91% of the time with an average pay-
ket environments and volatility regimes often occur in months directly follow- off of $4.6 per dollar invested. When
over the past 10 years, we show that ing market volatility spikes. Short posi- the future was trading within 1/2
long volatility strategies have proven tions initiated when the VIX point of spot VIX at trade initiation,
very profitable ahead of market shocks, future was trading substantially above the success rate was 100% and the
and shorts have been more consistently spot VIX also proved profitable. average payout increased to $6.9.
profitable in moderate-to-up markets.
We modelled the relationship Being long VIX futures has paid • In months when the S&P 500 was
between trading profits and three pri- off: up over 3%. Long VIX futures trades
mary factors: market returns, forward (1) During major shocks over the had low success rates regardless of
expectations, and the level of market last decade. Long one-month VIX the futures basis at trade initiation.
volatility at trade initiation. The results futures strategies had an average payout VIX futures trades experienced losses
provide strong intuition into the key of 11 to 1 across months covering: the 77% of the time in strong markets
drivers of VIX futures strategies. When Asia crisis, Russian debt default, the with an average loss of $2.6.
combined with a market view, the tech bubble, 9/11, and the corporate
model could also be a valuable tool in accounting scandals of 2002. • Moderate return environments.
constructing VIX trades. We find mar- When market returns were between -
ket returns and the VIX level to be the (2) In down markets. The average 3% and 3%, long VIX futures were
most important drivers of VIX futures payout was $1.9 per dollar notional profitable only 33% of the time. The
strategies in extreme market environ- invested in months when the S&P 500 - futures basis at trade initiation was

22 THE TECHNICAL ANALYST January/February 2007


often the difference between a prof- Highlight volatility regimes • Eight out of the top ten largest long
itable and unprofitable trade. Two periods were extremely successful VIX futures gains came during
for short futures strategies: (1) the months when the S&P 500 was
- future > VIX spot at trade initia- aftermath of the Russian debt default down 3% or more. The average
tion. Long positions initiated when from October 1998 to December increase in VIX spot was 7.7 vol
the VIX future was more than 1/2 1999 and (2) the current low volatility points or 35% with an average mar-
point above spot VIX were only regime from January 2003 to August ket move of -6.5%.
profitable 20% of the time with 2006. Together these periods account
average loss of $1.6. for a string of 48 out of 59 months • The top ten periods of VIX futures
(81%) when short futures strategies performance all had payouts greater
- future < VIX spot at trade initia- were successful. These regimes were than 5 to 1 with the period covering
tion. Long trades were profitable both periods with a steep, upward-slop- the Russian debt default (July 31 -
53% of the time with an average ing term structure of implied volatility August 31, 1998) topping the list
payoff of $1.0 per dollar notional and a VIX future trading well above with a 20 to 1 payout.
invested when the future was trading VIX spot (high basis) at trade initiation.
1/2 point or more below VIX spot. The future converged to a lower VIX • The average gain on a calendar-
spot at expiration for a gain on the month futures position was $8.5 per
Mind the market: VIX futures short. dollar notional invested in the future
respond to market returns versus $7.7 if one could have traded
Our ten-year back-test confirms that Long VIX futures over the top spot VIX. The difference being that
long futures strategies produced strong ten calendar-month declines in VIX futures were trading an average
results during the major market shocks the S&P 500 of -0.8 volatility points below VIX
over the last decade. The most success- We highlight the ten largest S&P 500 spot at trade initiation.
ful regime for short strategies was the calendar-month declines from
low volatility environment from February 1996 to August 2006 and the The best time to be short VIX
January 2003 to August 2006, when corresponding long VIX futures pay- futures: Top ten largest gains
selling one month VIX futures was offs on a $1 notional trade. for short strategies:
profitable in 34 out of 44 months The top ten largest gains on a one-
(77%) with an average payout at expira- • The market was down 8.5% on aver- month short VIX futures position or
tion of $1.22 per dollar invested. age, accompanied by an average corresponding loss on a long position.
increase in VIX spot of 6.6 vol The following two themes stand out:
Long strategies experienced points (+28%). Long one-month
high payouts in major market VIX futures strategies were • Each of the top ten largest gains
shocks successful in each of these ten cases, occurred in months where the S&P
Long one-month VIX futures strategies with an average payout of more than 500 return was strongly positive
experienced their strongest gains dur- 7 to 1. (average return: 5.8%; lowest return:
ing the major shocks over the last 1.6%) and risk expectations plum-
decade: the Asia crisis, Russian debt • Three of the largest moves include: meted. The average gain on a short
default, 9/11, and the corporate VIX futures position during these
accounting scandals of 2002. Large - Russian debt default (August 1998): months was $6.7 per dollar notional
market downturns were accompanied S&P 500 return = -14.6%; VIX invested (22.8%).
by a significant increase in volatility futures P/L = $19.9
with long futures strategies capturing • Many of the largest gains came
the rise. - Corporate account scandals immediately after months with the
(September 2002): S&P 500 return largest losses as the VIX mean
Short strategies were success- = -11.0%; VIX futures P/L = $5.7 reverted after a spike. Examples
ful in rebound markets and include October/November 1998
high-basis environments - Aftermath of 9/11 (September after the Russian debt default, and
The VIX is mean reverting, and many 2001): S&P 500 return = -8.2%; months following 2002 corporate
of the largest gains on short futures VIX futures P/L = $7.2 accounting scandals.
strategies occurred by going short in
the months directly after market volatil- The best time to be long VIX How we calculate our VIX future
ity spikes. futures: payout
Top ten largest gains for long strategies In our back-test, we: →

January/February 2007 THE TECHNICAL ANALYST 23


• Initiate a long one-month VIX • Expiration payouts are estimated for Bloomberg codes
future trade at the end of each calen- the 127 calendar months covering Real time market quotes can be
dar month (midmarket) February 1996 to August 2006. For obtained in Bloomberg using the codes
example, during the Russian debt below:
• Close the one-month position at the crisis (August 1998):
end of the next calendar month • Listed VIX futures: VXB INDEX
Assume $1 notional is invested in • We initiated a position on July 31, CT <go>
each VIX future trade. 1998 and estimated a 1-month VIX
future at 24.34. ?We closed our posi- • Front-month VIX future: UXA
• The long one-month VIX futures tion on August 31, 1998 with a spot INDEX DES <go>
payout at expiration is calculated as: VIX level of 44.28.
• VIX future settlements: VVS
$1*( VIX spot at expiration - 1-mmonth • Our payout at expiration was $19.94 INDEX DES <go>
VIX future at trade initiation ) per dollar notional invested in the
long 1-monthVIX future trade.
• We traded $1 notional each month $1*(44.28 - 24.34) = $19.94 per dol-
so that bid-ask spreads could easily lar notional invested.
be applied. Applying a mid-to-ask
spread of 0.1 vol points would
reduce our overall long P/L by Krag Gregory is head of index
$0.10. options research at Goldman Sachs.

24 THE TECHNICAL ANALYST January/February 2007

Generating Alpha in up, down and sideways markets

“We have developed one of the most powerful indicators to forecast

the stock market. Our Scoring System is crucial to helping investors
achieve superior returns in a fully hedged environment”

Thierry Laduguie, Director, Wave Matrix Ltd

To test run our Scoring System call us on 0870 873 8811
or email
or visit

Monday Blues
The worst day of the week for stocks?

26 THE TECHNICAL ANALYST January/February 2007


hree economics (or even reversing) when the previous
researchers from the UK Friday return is positive.
and South Korea have In absolute terms, and based on tra-
ditional regression analysis, mean
recently tested and confirmed the
returns for Mondays are in the region
Monday effect in daily stock of 0.12 to 0.18% worse than the aver-
returns. Using a different and age daily return for the rest of the week
(they argue) more appropriate sta- in respect of the NASDAQ, RUSSELL
tistical technique called Stochastic 2000, NIKKEI 225 and the pre-1988
Dominance, Young-Hun Cho, DJIA and pre-1988 S&P 500 (see Table
Oliver Linton and Yoon-Jae 1). In respect of the FTSE 100, mean
returns on a Monday have only been -
Whang from the Korea
0.02% worse, whereas the Monday
University, London School of effect seems not to have held at all for
Economics and Seoul National the large cap DJIA and S&P 500 since
University tested the idea that 1988.
Monday is the most miserable day
of the week for stock markets. But how can the effect be
Specifically, rather than look at averages There are broadly speaking four ways
and regressions which are fraught with of explaining the effect:
problems of non-normality and an
inadequate treatment of risk, they use a 1) Some claim it is simply an artefact of
technique borrowed from decision the- data snooping. Even if the data is ran-
ory, Stochastic Dominance, to look at dom, one of the days is bound to
the major stock indices in the US, UK exhibit the greatest or least returns and
and Japan from the 1970 to 2004. The it so happens to be Monday. Or, in the
technique looks at the probability of authors words, the possibility that "in
outcomes and, in this case, tests the practice statisticians are really searching
assertion that there is a greater proba- over so many obviously absurd anom-
bility of a more satisfactory outcome alies".
on each of the days from Tuesday to
Friday, than on Monday. Or, in other 2) Market microstructure, specifically
words, that "Monday is (stochastically) issues about settlement, dividends and
dominated by all other weekdays." taxes. For example, some suggest that
In this context, if Monday returns are returns should actually be greater over
second order stochastically dominated the weekend to take account of the
(a form that takes account of risk aver- extra calendar days. But this is at odds
sion) by the other weekday returns, with the data. Others suggest that the
then no risk averse individual who is settlement period is of more impor-
also a maximiser of expected utility tance, which would mean that Fridays'
(profits in this case) would prefer returns should be higher than Mondays.
Monday returns to the other weekday
returns. 3) Different rates of micro and macro
As can be seen in Table 1, they find information, i.e. the release of bad
strong evidence of a Monday effect for news tends to be delayed until the
most of the stock indices they study, weekend. Steeley (2001) argues that the
particularly the broadly based indices, Monday effect in the UK is related to
although the effect seems to have less- the systematic pattern of market wide
ened for some of the large cap series news that concentrates between
like the DJIA and S&P500 since 1987. Tuesdays and Thursdays.
Moreover, the effect seems highly relat-
ed to the previous Friday's return, get- 4) Differential trading patterns of vari-
ting stronger when the previous Friday ous market participants. Individuals are
return is negative and getting weaker net sellers on Mondays and individ- →

January/February 2007 THE TECHNICAL ANALYST 27




Table 1. FSD = first order dominance (hypothesis applies to all non-satiable individuals), SSD = second order dominance (hypothesis
applies to all non-satiable and risk averse individuals), TSD = third order dominance (hypothesis applies to individuals with addition-
al restrictions on their utility functions). + (-) Friday = positive (negative) returns on previous Friday.

uals behave differently on Mondays librium returns. This also tallies with Of relevance here is the strong evi-
versus other days of the week. Or else, research by Pettengill (1993 and 2003) dence of a Monday effect on days
it could be due to short selling activity - which found that none of the most when the previous Friday return is neg-
short sellers close their position on cited explanations (in 2, 3, & 4 above) ative. Market efficiency assumes that
Friday as it is difficult to monitor over are able to fully explain the effect. overreaction to information is as fre-
weekends. They sell their stocks on Clearly further investigation is sought quent as underreaction, yet the Friday /
Monday leading to a fall in prices. and the authors believe it may come Monday effect seems to refute this idea.
With regard to data snooping, the from the field of behavioural finance, This suggests the need for an alterna-
authors argue that because the same the most fertile source of theories to tive model that specifies biases in infor-
result has been more or less found for explain why the financial markets may mation processing that cause the same
all the many different indexes they not operate according to the efficient investors to under-react to some types
study (large and small cap, domestic markets hypothesis. of events and over-react to others, thus
and international), it points to an explaining the range of observed
underlying process at work. They also Implications for asset pricing results better than simple market effi-
test several variants of the hypothesis models? ciency alone.
(such as "There exists at least one day Having found such anomalous market
that is dominated by all others") to look behaviour, the authors ask if it should Based on the paper "Are there
for consistency and logic. it be interpreted as market inefficiency Monday effects in Stock Returns: A
The hypothesis tested is - according or simply that they are working with a Stochastic Dominance Approach",
to the authors - stronger than the usual bad model of market equilibrium. They 2006, by Young-Hyun Cho
one and despite some grounds for cau- argue that evidence of stochastic dom- (Department of Business
tion regarding the statistical signifi- inance of Monday returns could be Administration, Korea University),
cance (as their methods used require combined with behavioural theories Oliver Linton (Department of
large sample sizes), their findings sug- from the psychology literature to create Economics, London School of
gest that regardless of investor's atti- new asset-pricing theories that combine Economics) and Yoon Jae Whang
tudes of risk, degree of risk aversion, economic equilibrium concepts with (Department of Economics, Seoul
or seasonal variations in risk premia, psychological concepts to create an National University), with permis-
Monday returns are too low to be equi- improved asset-pricing model. sion.

28 THE TECHNICAL ANALYST January/February 2007



Jack Schwager is Investment Director at the

Fortune Group, an alternative asset manage-
ment firm and a member of the Close
Brothers Group, a London-based investment
bank. He is the lead portfolio manager for
Fortune's Market Wizards Fund, a 100%
transparent portfolio of managed accounts in
alternative strategies. Jack is best known as
the author of the best-selling Market Wizards
books. His first book, A Complete Guide to
the Futures Markets, published in 1984, is
now considered a classic and was later
expanded into the three-volume series,
Schwager on Futures.

TA: What is your role at Fortune? Are you still involved in futures type of chart that corresponds to a fixed date that
trading? changes as the amount of time up to the expiration date falls.
For example, if I were to buy yen and hold it for 3-months
JS: I'm officially an investment director running a portfolio and then look at a currency chart 3 months forward and a
on the fund of funds side of the business. My main job is the future chart, the futures chart would exactly reflect what was
selection of managers and then putting them together into a happening with my equity. The currency chart wouldn't as it's
balanced portfolio. This means that I'm not actively trading always a constant amount of time forward and is not reflect-
anymore. This is by choice as I would rather be doing what ing the change in the spread between what I've bought as
I'm now doing at Fortune. time evaporates.

TA: Many of the books you have written in the past are TA: Is there any difference in the use of TA across different
about the futures markets. Are there any special factors to asset classes
consider in using technical analysis and charts when trading
futures? JS: There can be a difference between currencies and equi-
ties, especially indexes. The former tend to have an intrinsic
JS: Any market that trades on what I would call constant for- trending quality over long periods of time. Equity indices
ward type charts, be it spot or 3-month forwards, those type tend to be more mean-reverting along a secular uptrend. This
of charts for a longer term trader are mis-leading. This is implies that using similar technical approaches in the two
because you don't see a chart of what your position is. If I markets may yield very different results. For example, if you
buy a 3-month FX forward today, a month from now I'm still are trading equity indices using trend following methods you
holding the position, but I'm no longer looking at my posi- are more likely to get disappointing results. On the other
tion on the chart as it doesn't show if the spread is causing a hand, in the currency markets, over time, a trend following
depreciation or appreciation of value. If you want something approach will generally make money.
that reflects your position you need a chart that will parallel
the changes in the equity of the position. That would be a TA: What about volatility and trend following. Are there →

January/February 2007 THE TECHNICAL ANALYST 31


any repercussions for this strategy in the current low volatil-

ity environment? “…IF YOU ARE TRADING
JS: Not necessarily. You can have a trending market and low EQUITY INDICES USING
volatility together; just look at equity indexes in 2006. Indeed,
in the equity markets in particular, low volatility is often asso- TREND FOLLOWING METHODS
ciated with up-trending markets. The direction of the mar-
kets is what's important; the worst case scenario for trend fol-
lowers is having high volatility in a sideways market. We don't GET DISAPPOINTING
have this at the moment.
TA: When you were trading, how did you use TA?
JS: For purposes of managing money when I was a CTA, my
partner (Louis Lukac) and I used a fully systematic approach.
The systems were based on both trend and counter-trend GENERALLY MAKE MONEY.”
technical concepts. For trading my own account, which was a
small amount of money, I combined chart analysis with pro-
tective stops. As one example, if a market had a steep down-
swing before going into a narrow consolidation, then I would
usually look for a continuation of the move. In that case, I pared to occasionally ride through a 10% drawdown you
might go short with a stop just outside of the consolidation. shouldn't be in that investment, while for some other low-
volatility arbitrage strategies, even a loss much smaller than
TA: About money management and stop losses, how should 10% would suggest something is wrong and you should
a trader decide on the best approach? redeem.

JS: That really depends on the type of markets you are trad- TA: What is your view of more systematic techniques such
ing and your individual trading style. It's a mistake to think of as DeMark?
money management rules in any general sense. If you're trad-
ing futures, for example, you have margin and you can lose a JS: Let me answer this question more broadly as pertaining
lot of money relative to your account size. In this case I to all indicators, not the DeMark indicators in particular. My
would argue that it's better, or even essential, to have an view is that you shouldn't assume that any indicator works
explicit or pre-determined stop loss. without rigorously backtesting over a broad price history and
An alternative approach would be needed if you are an across many markets. Very often traders assume that using a
equity value investor and your strategy is to look for stocks given indicator will lead to profitability because they have
that are trading very low relative to the perceived value of the seen a few charts where it has worked very well. My experi-
company. In this case, risk management through stop-losses ence is that there is a great tendency to use what I call "the
would be in conflict with the approach itself. For example, if well chosen example" to illustrate indicators. In articles or
the market falls and your stock declines along with it and the books on indicators you will probably see 100 examples of
fundamentals pertaining to that stock have not changed, then an indicator working for every one you see where it fails. But
a stop would get you out at a point where the argument for I can assure you the reality is very different. I can't answer the
being in the position is even greater than it was before the question regarding the DeMark indicators because I haven't
market fell. Risk management, however, would still have to tested them.
be achieved by other means (e.g., diversification, etc.)
TA: How can you test an indicator, such as the DeMark
TA: What about percentage stop losses? sequential, if the entry rules are specific but the exit rules are
JS: Again it depends on the markets being traded. It may be
very appropriate for a futures trader because of the leverage JS: You can test it by combining the specific signal entry with
implied by margin, but inappropriate in other applications. a wide range of exit strategies. You could use either price
For example, I have often heard fund of fund managers movements, or time, or both to define the range of exits.
boast that when any manager they invest in is down 10% (or Using time you could test the results of getting in on the sig-
some other fixed amount), they automatically redeem. This nal and then getting out a broad range of days after the sig-
one-size-fits-all stop rule always seemed absurd to me nal. You could also combine this with a back-up maximum
because there are some strategies where if you are not pre- stop-loss rule. There are a lot of variations, but if the indica-

32 THE TECHNICAL ANALYST January/February 2007


tor has value, you should see a pattern of meaningful net erly. It's far more important to have good money manage-
profits, on average, across the range of exit strategies tested, ment than an optimal entry approach.
or at least a broad contiguous range of parameters within
that range. TA: Did you talk to any traders who were trading very short
term but who didn't use technical analysis? If so, what was
TA: What was your motivation for the Market Wizards book? there approach?

JS: After writing an analytical tome I wanted my next book JS: Well, almost by definition, if you're a short term trader,
to have the potential of reaching a much broader audience. I then you are using technical analysis. Nearly all short term
thought that interviewing great traders with the objective of traders I think use technicals to some extent. Maybe floor
distilling the principals of trading success would be a fun traders have some additional knowledge that they can trade
project and of interest to a much wider readership. on as they may know what the orders are and where the stops
are but even they would look at charts at some point. The real



TA: What is the obvious common characteristic shared by exception to your point is statistical arbitrage traders who use
the traders you talked to? mathematical rules rather than technicals or fundamentals to
JS: There were many common denominators even though
the traders were very different, ranging from purely technical TA: What's your view on the markets at the moment? There
to purely fundamental and from short term to long term. seems to be a consensus that the dollar will continue to fall
One example of a common factor in trading success was that and that US stocks are set for another good year in 2007. Is
their trading technique always matched their personality. This this a good moment for contrary opinion to come into play?
is not as obvious as it may seem as many traders end up using
approaches that are in conflict with their personality. For JS: I am not going to opine on whether sentiment on the dol-
example, some people design trading systems when they lar or stock market has already reached excessive levels, but I
don't really have the patience to trade that way and end up certainly agree that alarm bells should start ringing when the
sabotaging their own systems. Others may have been great market view is overwhelming in one direction. Of course,
traders on the floor and then move to trading from an office, contrary opinion works on the old argument that if everyone
a style that doesn't suit them. If you asked the traders them- thinks a market will fall then there's no one left to sell. I know
selves what makes them different, the most common answer one stock index trader who uses sentiment exclusively as a
would be summarized in a single word: discipline. The mar- trading rule and in the four years I have been monitoring him
ket is a stern master and it is uncanny how even infrequent has done extremely well in his market calls.
lapses in discipline can prove extremely costly. Great traders
apply discipline without any exceptions. Good money man- TA: Do you have any new books planned at the moment?
agement is also essential and is often underappreciated as a
critical aspect of trading. There is perhaps too much empha- JS: Not now. Writing books is hard work and I've already
sis placed on entry to the detriment of managing risk prop- done my fair share!

January/February 2007 THE TECHNICAL ANALYST 33




oday's markets pose key chal- saturated by other market participants. That brings us on to our second key
lenges to those seeking trading The components of a good strategy component: Identification. We could
profits. In contrast, markets of may be quite straight forward; however spend hours manually scouring charts
20 years ago held a plethora of trading what makes it successful is the moni- or fundamental reports looking for
opportunities, often obtainable from toring and disciplined execution of that instances where our strategy criteria fit.
trading just one product or even one strategy. For example, you may have a But that is unrealistic, so we need to
stock. So why have things changed so strategy that waits for three technical utilise technology to help us. If we have
much? Quite simply we have more mar- indicators to reach certain criteria a predefined strategy and an automated
ket participants embracing ever chang- before entering or exiting a trade. This means for searching looking for a crite-
ing technology resulting in a more may seem simple, and good strategies ria fit, then we are half way there. Then
mature market where opportunities are do not have to be complicated, but in we need to trade our signals. Some
not so apparent and need to be weeded traders are super disciplined but most
out and traded with the ultimate preci- are not: Automatic execution sounds
sion and efficiency. like we are just generating a computer
We all know that to have a strategy is to do all the work. Well, to a certain
crucial for successful trading; however extent we are, but the markets are for-
the means by which we execute that ever changing and we need to have
strategy is of much greater importance. some flexibility in what we do to
It is fair to say that you can succeed accommodate those changes. When a
with an average strategy and excellent strategy is coded into software which
discipline but not with an excellent will automatically generate signals for
strategy and average discipline. So we you, the strategy is 'boxed'
need to have a plan, but more impor-
tantly, an effective means of executing Black and white boxes
that plan. Discipline, or lack of disci- A 'Black Box' is a term used to describe
pline, is probably the number one rea- a 'closed' or 'sealed' trading system. By
son traders fail to make money and this we mean, an algorithmic strategy
after all, we are human and prone to which generates signals in any chosen
emotion, which is more often than not market but which cannot be changed or
the catalyst for poor and irrational trad- adjusted. It is then possible to have the
ing decisions. The problem is that disci- trading signals executed automatically.
pline is not always easily learnt. This is perfect if the black box strategy
However there is an overnight solution is a good one and some of the good
to adhering to a trading strategy - auto- ones may stay profitable for maybe 5-
mated trading. Screen shot 1. 10 years. But many start out as winners
and soon loose their inherent effective-
Automated trading ness and cease to be profitable. There
Automated trading is best described as are many websites which are literally a
the identification of strategy criteria, today's competitive market we need to library predefined strategies, submitted
signal generation and execution of that search for products or contracts where by their creators for others to trade
defined strategy, automated. our defined strategy criteria are in with. Collective2 ( is
Firstly, define a strategy. If we want to place. No use watching a future or the largest market place for black box
make money, we need to identify an stock for days on end waiting for your trading systems. They claim to hold and
'edge' which is not being exploited and trade criteria to hopefully appear. monitor in excess of 2500 algorithmic

34 THE TECHNICAL ANALYST January/February 2007


Screen shot 2.

trading systems which can be viewed, line) versus the performance of the becoming increasingly popular, espe-
reviewed and subscribed to. Some of underlying index (dotted line), in this cially amongst trading institutions, is
these systems are characterised by their case the S&P 500. the 'white box' system approach.
quirky names; Big Cat, Gold Survivor In addition to a graphical representa- A white box system is subtly different
and Rocket Science. Collective2 posts tion of the system's performance, a to the black box in that it is your per-
the performance of all the trading sys- detailed breakdown of the perform- sonal algorithms that make up the strat-
tems allowing suitors to shop for an ance analysis can also be viewed (Table egy. The advantage of this approach is
appropriate system. 1). the ability to easily tweak or adapt the
Figure 1 shows extracts of a graphi- The trading systems available from system to changing environments or
cal example of the performance analy- Collective2 are black box systems, writ- market circumstances, whereas a black
sis of a system listed on the Collective2 ten and coded by third parties so the box is closed and does not allow any
website. The graph shows the compar- algorithms normally cannot be changed adaptation to its factory fitted capabili-
ative performance of the system (red or adjusted. An alternative, which is ties.
Algorithmic trading using the white
box approach is rife in the hyper- com-
petitive trading world as financial insti-
tutions now feel the mounting need for
technology that aids their unique trad-
ing style. With a white box it is possible
to rapidly compose or evolve algo-
rithms to monitor, analyse and respond
to market events in a specific way.
Although the pre-defined black box
algorithmic strategies are a great place
to learn the ins and outs of automated
trading systems, the user defined white
box approach had greater flexibility and
therefore greater longevity.

Automatic execution
There are websites available which
allow you to create and code your own
algorithmic trading system, for example
Trade Station (,
Figure 1. eSignal ( and →

January/February 2007 THE TECHNICAL ANALYST 35




whether or not to trade. If you believe

in your strategy then you should trade it
whatever the market feels like at the
time of signal generation. To have the
link between signal generation and
trade execution taken care of by an
automated process means that disci-
pline is maintained even during the
most volatile market phases.

• The competitive arena which we

speculate is forever embracing tech-
nological advances which in turn
spoil some of the fun for the traders
relying on manual signal generation
and execution.
Screen shot 3.
• The transparency of most markets
means that the edge is harder to find.
Ninja Trader ( allow the automatic execution of any Automated algorithmic systems can
which can then connect directly to the trading system via scan multiple markets looking for a
interface of a broker for automatic exe-'s gateway to the criteria fit.
cution. markets. is one broking This automation from system cre- • Trading psychology and discipline
portal which can interface with any ation through to execution really is the are difficult to master. Automatic
Collective2 black box system or system pièce de résistance in the trading world. execution removes the emotion from
individually coded on Trade Station or Financial institutions and fund man- trading making it easier to follow
e Signal. This broking interface will agers have adopted automated trading your strategy precisely.
as the core engine or backbone to their As trading technology continually
trading objectives. Individual traders advances, the race is on to find algorith-
and investors are close behind as sys- mic supremacy. As a result some strate-
tems building software, linking direct to gies are becoming forever more com-
brokers for execution, becomes main- plex, demanding a greater need for
stream. To have the link between signal automation. Now that the execution of
generation and trade execution taken trading strategies can be automated, the
care of by some automated process risks of ill discipline and human irra-
means that discipline is maintained tionality have all but been removed.
even during the most volatile market
phases. Simon Brown is a consultant at
Futures Betting.
The advantages of automated
trading For more information on automatic
The automated execution has the head- execution via,
line advantage of removing your emo- call +44(0)20 7183 0431 or visit
Table 1. tions from the final decision of

36 THE TECHNICAL ANALYST January/February 2007



arton Biggs is a Wall Street veteran who spent 30 years at Morgan Stanley
Investment Management before setting up Traxis Partners, a Connecticut
based hedge fund in 2003. He has written a real page turner that should be
fascinating to anyone involved in the markets, from private traders to hedge fund
managers. "Hedge Hogging" is similar in some ways to Edwin Lefevre's
"Reminiscences of a Stock Operator", by common consensus the classic book on
trading the markets published 80 years ago - it's a very personal reflection on many
years spent on Wall Street and the traders he has known.
Despite the title, this is not by any means a book exclusively about the hedge fund
industry. It is more a collection of reminiscences, anecdotes and opinions on the
world of trading and investing. He does talk at length about friends and colleagues
who have gone through the process of setting up and running a hedge fund. Best
of all, he describes vividly the stress and strain of running a fund and the associat-
ed responsibilities of managers who run both their own and their clients money.
What is clear is that there are many brilliant traders out there running funds, very
often in almost secret circumstances, and Biggs is in awe of these individuals who
manage to perform, or out-perform, year after year. Single mindedness appears to
Hedge Hogging be the common characteristic of most of them.
By Barton Biggs
Whilst Biggs is not a devotee of technical analysis, he concedes that even funda-
Wiley and Sons mentally based traders like himself need to look at charts because support and resist-
ISBN: 0471771910 ance levels and chart patterns have an impact on price movements. He devotes sev-
320 pages eral pages to Fibonacci numbers with an amusing story of a meeting between him-
£16.99 self and a technical research firm looking to sell him their Fibonacci-based market
calls service. Although the researcher had called two recent market bottoms, Biggs
Hedge Hogging can be
seems unsure what to make of it, concluding in the end that Fibonacci is probably
purchased from the Technical too long term for his trading style anyhow. However, he advises sceptics to keep an
Analysis bookshop. To order open mind. He goes on to emphasise, quoting research from Ned Davis, that the
please call 01730 233870 and market is always wrong at price extremes and turning points. Contrary opinion then
quote "The Technical Analyst has a part to play. He describes some of his favoured 'indicators' for markets that
Magazine". may be overbought including the VIX, Economist magazine covers, and an interest-
ing anecdote concerning what he calls the "mother-in-law factor". Biggs also gives
warning about what should be avoided, or treated with caution, in trading and
investing. These include short selling and trying to call market tops and bottoms. He
also discusses market cycles by examining the difference between cyclical and secu-
lar bull and bear markets and the behaviour of both historic and more recent stock
Although Biggs has a tendency to take personal wealth and huge Wall Street
bonuses for granted when talking about the markets and its players, he also has a
healthy dose of scepticism about the whole hedge fund industry and its ability to
really deliver for its clients when the large fees charged are taken into account. This
is an extremely well written, funny and fascinating book that has something to teach
the reader whatever his or her trading style may be. It also has the advantage of hav-
ing been written from the inside looking out.

January/February 2007 THE TECHNICAL ANALYST 37

38 THE TECHNICAL ANALYST January/February 2007
by M Dempster and V Leemans

n the quest for a trading algo- departs from a similar principle by ber of recursive training epochs ne was
rithm, artificial intelligence developing a fully layered system where set to 10. The optimal values of these 3
methods have been employed risk management, automatic parameter parameters did not change significantly
tuning and dynamic optimisation are when different FX data sets were used.
to construct systems that are bet-
combined - a self-learning system We extended RRL to take account of
ter than, or at least as good as, which we call 'Adaptive Reinforcement inputs other than past returns and the
their human counterparts in tim- Learning (ARL).' current position. Experiments were
ing trade entry and exit opportuni- conducted to include signals originating
ties. The problem, however, has Layer 1: The machine-learning from 14 popular technical indicators.
been that even if a trading model algorithm Performance did not increase by adding
was shown to produce an accept- At the core of the system is a machine- indicators to current data, except when
learning algorithm called Recurrent a low number of past returns (M) was
able risk-return profile on histori-
Reinforcement Learning (RRL). This is fed into the system. The filtering of the
cal data there was no guarantee an algorithm that was originally pre- price data timeseries offered by the
that the system would keep on sented in 1997 by Moody and Wu and technical indicators thus did not con-
working in the future. It would applied with some success to FX trad- tribute to increased performance on
cease working precisely at the ing by Moody and Saffell in 1999. current data. This indicates that the
moment it became unable to adapt Machine learning is a subject belonging RRL algorithm is able to efficiently
to the changing market conditions. to artificial intelligence concerned with exploit the information in past returns
the development of algorithms and timeseries.
In our research, we have attempted to techniques that allow computers to During the training phase, the trans-
obtain a usable, fully automated and "learn". 'Reinforcement learning' is a action cost factor (d) was not fixed to
intelligent trading system. To accom- sub-area of machine learning, and con- the actual bid-ask spread, but was
plish this, a risk management layer and cerned with how an agent ought to take instead left as a tuning parameter.
a dynamic optimisation layer are added actions (or adjust the parameters in this Setting a higher cost factor, necessitates
to a known machine-learning algorithm case) in an environment so as to maxi- a higher expected raw profit before
(see Figure 1). The middle layer man- mize some notion of long-term reward engaging in a trade. Consequently the
ages risk in an intelligent manner so (e.g. Sharpe ratios). parameter d will influence the perform-
that it protects past gains and avoids As demonstrated by Gold (2003), ance and risk profile of the resulting
losses by restraining or even shutting RRL lends itself perfectly to a rolling- trading strategy.
down trading activity in times of high window approach. Here, the model is Large jumps in FX returns, as e.g.
uncertainty. The top layer dynamically trained for a number of epochs (ne) on when central banks intervene, can
optimises the global trading perform- a training set of length Ltrain and then introduce instability into the underlying
ance in terms of a trader's risk prefer- applied to a test set of length Ltest. RRL algorithm. To prevent the weights
ences by automatically tuning the sys- Next, the training window is advanced from taking unreasonably large values,
tem's parameters. by Ltest points and the whole procedure all weights were rescaled by a factor f <
While the machine-learning algo- is repeated. The out-of-sample per- 1 as soon as one of the weights hit a
rithm, which is at the core of the sys- formance of the model is the sum of certain threshold value. This also pre-
tem, is designed to learn from its past the performances on the individual non vents the weights from drifting slowly
trading experiences, the optimisation overlapping test sets. We found that the upwards without bounds. This phe-
overlay is an attempt to adapt the evo- optimal values of Ltrain and Ltest on nomenon is undesirable because the
lutionary behaviour of the system and our FX (midpoint quote) datasets were user may get incorrect model evalua-
its perception of risk to the evolution around 2000 and 500 ticks respectively, tions at some point because of the
of the market itself. This research which are the values chosen. The num- numerical instabilities associated →

January/February 2007 THE TECHNICAL ANALYST 39

enter a trade is sepa- seen the situation and would not have
rated from the trade suffered a loss. If the market behaves
recommendations 'irrationally' according to the model, it
made by the basic is likely that its deviant behaviour will
structure in Layer 1. persist for some time. For this reason a
While Layer 1 outputs cool-down period is included in Layer
the preferred position 2. After the position has been closed
to hold in the model out the system needs to wait a while
and thus idealised before trading can be resumed. The
world, Layer 2 evalu- length of this optimal period of non-
ates this trade recom- activity was determined experimentally
mendation by consid- as a fixed number of basic time inter-
ering additional risk vals (here 1 minute).
factors relating to the An important feature of the risk man-
real world before tak- agement layer is that allows the evalua-
ing a decision to tion of the strength of the trading sig-
Figure 1.
trade. These risk fac- nal given by Layer 1. This is accom-
tors impose require- plished by looking at the unthresholded
ments on real world output generated. For example, a very
trading systems that strong buy signal corresponds to an
with very large weights. are hard to include in the trading model output close to +1. The performance
We also improved the performance of itself, but can more easily be treated in management systems offer the possibil-
the RRL trading model by implement- the risk management overlay structure. ity of validating a trading signal only
ing a better position updating scheme. More than half a century ago J. M. when the output exceeds a specified
The traditional update scheme some- Keynes, the famous economist noted non-zero threshold instead of just
times causes indecisive switching that: "The markets can stay irrational applying a sign-function. This thresh-
between different positions from one longer than you can stay solvent." old is a parameter (y) that is not fixed in
tick to the next. This can generate huge When a trade goes bad, a psychological advance but that can be set by optimi-
transaction costs and is highly undesir- tendency exists to keep the position sation in Layer 3 to influence the risk-
able. This behaviour is reduced by open in the hope that the market will return trade-off made by the trading
recalculating the output of the trading reverse itself and the trade will again system as a whole.
model twice. The first time is after the turn profitable. This implies a risk- It is common knowledge in the trad-
new inputs have been received and a seeking attitude towards losses as ing community that automated trading
new trading signal is generated. This opposed to a risk-aversion with regard systems often work for a certain while
trading signal is in fact based on the to profits. However, the market could until they suddenly stop being prof-
weights that were calculated in the pre- very well not reverse itself and eventu- itable. At that point, the system must
vious time period. Here a second eval- ally force the close out of the position shut down to avoid further losses and
uation of the model is added after the at a huge loss. This characteristic of should be redesigned thoroughly tested
weights are updated for the current human psychology needs to be avoided before making it operational again. The
period. Since the weight updating is by a successful automated trading sys- performance management layer makes
designed to improve the model at each tem. Layer 2 avoids this pitfall by build- sure that anomalous performance is
step, it makes sense to recalculate the ing in a trailing stop-loss for every detected at an early stage to protect the
trading decision with the most up-to- trade. A stop-loss is set and adjusted so profits already earned. In that case, the
date version of the parameter. This that it is always x basis points under or system is automatically shut down and
recalculation may result in a different above the best price ever reached dur- a warning is issued. The detection mod-
trading signal than was previously out- ing the life of the position. At this ule employs a measure known as maxi-
put by the model. This final trading sig- point x is an unknown parameter that mum draw-down. If at a certain point
nal is used for effective decision making will influence the performance. A the draw-down in cumulative profits
by the risk and performance control numerical value for x will be obtained from its maximum proves larger than
layer. via the optimisation in Layer 3. an amount z, the auto-shut-down pro-
If a position is closed out before a cedure is initialized. This number (z)
Layer 2: Risk and performance trade exit signal is given by the system, can be determined as a Value at Risk by
management the current market behaviour was clear- using a fitted draw-down distribution,
One of the strengths of the layered ly not expected by the RRL trading or it can be set more pragmatically to a
structure is that the decision to actually model. Otherwise it could have fore- fixed value. →

40 THE TECHNICAL ANALYST January/February 2007

Layer 3: Dynamic optimisation sation of utility. model calibration or fitting and hence
of utility For EUR/USD, we allowed the sys- greatly reduces the risk of data-snoop-
The trading system up to Layer 2 is able tem to trade only during the active ing. The system performed very well on
to trade automomously and manage hours that correspond to sufficient liq- the high-frequency historical FX
risk appropriately. However, the sys- uidity (9am to 5pm London time). dataset that was examined.
tem's risk-return profile depends on a During these hours, the average bid-ask A first possible avenue for future
number of parameters (d, n, ?, x, y and spread in the inter-dealer FX market research is to examine whether the
z) and it is the task of Layer 3 to search was lower than 2 pips. As execution trading performance of this system can
for optimal values of these parameters speeds and liquidity are usually very be further improved by feeding other
so that some utility of the resulting high on interdealer electronic trading information than price data into the
risk-return profile is maximised. platforms like EBS and Reuters3000, system. In our earlier research, we
Therefore, it is key to find a suitable no further slippage was assumed. demonstrated that order book or order
utility function that is dependent on the (Note: Our model assumes that the flow information is able to enhance the
cumulative profit, on a measure of trading system always takes one posi- performance of automated trading sys-
trading risk and on the supervising tion or another and that at every trade tems. Secondly, the risk management
trader's personal risk aversion. exactly 1 unit of a certain currency is layer could be extended to control sev-
The length of the time for this top- bought or sold). eral automated FX trading systems that
layer optimisation interval is set to a The ARL system was able to earn trade different currencies, in an attempt
multiple of the training interval length 5104 pips over 2 years, or more than to emulate the actions of a human FX
Ltrain so that it is safe to superimpose it 26% per annum. Given the absence of trader.
on the previous layers. It is sensible to any serious draw-downs, this is not a
choose a relatively lower frequency for bad performance. Preliminary results Michael Dempster is Emeritus
this optimisation as the parameters on recent FX market data covering a Professor of Management Studies
being optimised determine the risk period up to January 2004 indicate that (Finance) and Director of the
management characteristics as well as the profit curve exhibits a diminishing Centre for Financial Research at the
the adaptive behaviour of the underly- slope. This could signify that markets Judge Institute of Management,
ing learning algorithm. As these are are becoming more and more efficient University of Cambridge. Vasco
both aspects of the system that typical- and that it is becoming more difficult to Leemans is a Research Student at
ly span multiple time periods or even make profits when only price informa- the same institute.
multiple trades, more time is required tion is fed into the system.
to see the impact of these hyper- To illustrate the benefits of our struc- Article extracted from "An Automated
parameters on trading performance. At tural approach and the top-level utility FX Trading System Using Adaptive
the start, the 5 parameters (d, n, ?, x, optimisation, a series of experiments Reinforcement Learning", 2004, M A
and y) are initialised with 5 randomly was run to compare performance and H Dempster and V Leemans, Working
chose values because there is no infor- utility of performance with and with- Paper 18/2004, Judge Institute of
mation available about what values out dynamic optimisation turned on. Management, University of
result in good performance. The opti- We found that the ARL system per- Cambridge. With permission.
misation then searches for values that forms consistently better than the
result in optimal utility of performance benchmark which illustrates that the References
over some past period of time. This addition of Layer 3 contributes signifi- Gold, C. (2003). FX trading via recurrent rein-
framework not only eliminates the cantly to better performance. forcement learning. IEEE International
problem of having to choose and fix Conference on Computational Intelligence in
hyper-parameters a priori, but also Conclusions Financial Engineering 2003.
allows the trading system to dynamical- We have developed an automated trad-
Moody, J and Wu, L. (1997). Optimization of
ly adapt to changing market conditions ing system based on adaptive reinforce-
trading systems and portfolios. Decision
and even to take into account changing ment learning. The parameters that Technologies for Financial Engineering. Edited by
risk preferences of the user over time. govern the learning behaviour of the Abu-Moustafa Y, Refenes A. N., and Weigend,
machine-learning algorithm and the A.S. London: World Scientific, 1997, 23-25.
Applications to FX trading risk management layer are dynamically
The ARL trading system presented optimised to maximise a trader's utility. Moody, J. and Saffell, M. (1999). Minimising
here was fed historical FX data and its A risk-aversion parameter allows con- downside risk via stochastic dynamic programming.
out-of-sample performance on this trol of the system's trade-off between 6th International Conference Computational
data recorded and analysed. A few strategy risk and return. ARL automat- Finance 1999. Edited by Abu-Mostafa Y.S.,
LeBaron B., Lo, A., and Weigand A.S., MIT
more experiments were conducted to ically tunes the hyper-parameters and
Press, Cambridge, MA, 403-415.
test the impact of the dynamic optimi- thus eliminates the need for manual

42 THE TECHNICAL ANALYST January/February 2007


lpha and Beta are two important measures of a give me the flexibility of applying the two measures to
security's performance in relation to the variability shorter time frames including daily and intraday data. Many
of other assets in the portfolio, or in our case, the traders who use Excel often use the built in User Functions.
index of which the security is a part. For example, XOM is For example if you are plotting a moving average you just
part of the 30 stocks in the "Dow Jones Industrial Index" click on the "=Average()" function from the drop down list
and it also a component of a number of other indexes of available functions and apply it to your series. For what
including the S&P 500. So in accord with our definition, the we want we need a new user function that does not exist in
Alpha and Beta for XOM will be different as measured Excel. Fortunately Excel allows us to define our own user
against the two indexes mentioned (DJI &, S&P 500). This functions. This is a definite advantage for when it is neces-
is important to keep in mind because we want to measure sary to define more complex or custom mathematical func-
the relationship between apples and apples instead of tions (like Alpha and Beta). The steps required for this are
apples and oranges, unless you have an inter-market strate- fairly simple and straightforward.
gy of trading the spread between apples and oranges.
Create a User Defined Function in Excel:
Alpha and Beta 1. Open a new (or existing workbook)
Let's start with a basic definition of the two terms. Alpha is 2. Click on the Visual Basic Editor, you can use the short-
a measure of a security's performance (relative to the cut Alt + F11 to do this.
benchmark index) in excess of what Beta would predict - or 3. When in the VBA Editor, click on Inset > Module to
how much of the returns cannot be explained by the under- insert a new VBA Module into your workbook.
lying index. In the Excel code we will be creating, it can be 4. Add the code (below) for the Alpha and Beta functions
defined as the intercept of an ordinary least squares (OLS) to the module you have created (you can put both in the
regression applied to out data series. In Excel the built in same module)
function is "=Intercept()". 5. Close the VBA Editor (or use the shortcut Alt+Q).
Beta is different in that it measures the variability of a 6. Now when you are back in the spreadsheet you will have
security's return as compared to the total return of the two new functions in the Function Dialog dropdown
benchmark index. It is generally regarded as how much of under "User Defined" (shift + F3). One for Alpha and
the return can be attributed to the index. Sometimes it is one for Beta. These functions will be applied to the Log
compared to a measure of volatility. It has at times also changes (Excel Function "=LN()") of your Market data
been mentioned as a measure of how much a security will and Security data.
move relative to the index. For example, a security with a
Beta of 1.5 is said to move 1.5 times as much as the under- Alpha code for User Defined Function
lying index. This is not exactly true (if you have ever tried Function Alpha(security, market)
to measure this). What is important to know is that a secu- 'Returns Alpha (Intercept from Regression)*
rity with a Beta of 1.5 will exhibit more volatility then the Alpha = Application.Intercept(security, market)
underlying index over the period being measured. In Excel, End Function
Beta can be measured using the built in function for the
slope of the OLS Regression, "=Slope()". Beta code for User Defined Function
The above definitions are for the sake of this article and Function Beta(security, market)
a much more extensive exploration of the terms and their 'Beta (Slope from Regression)*
applications can be found in any good text exploring the Beta = Application.Slope(security, market)
Capital Asset Pricing Model (CAPM). End Function
Alpha and Beta are more commonly applied to the
monthly returns of data relative to a portfolio of securities. * The comment lines are not required. →
What I would like to do is create a tool in Excel that will

January/February 2007 THE TECHNICAL ANALYST 43

Daily data examples:

Table 1. Excel Custom User Defined Alpha & Beta

Functions applied to an index and security

Now that we have created the new Alpha and Beta func-
tions we can proceed to apply this to our daily and intraday
data. We want to use the Alpha and Beta function as a way
to aid in identifying early possible breakouts, or the begin-
ning of short term trading opportunities. Alpha measures
returns in excess of Beta. So finding high Alpha's should
help me pick securities that are showing better performance
or are running ahead of the pack. Beta will help me identi-
fy if the securities are more (or less) volatile then the index.
Alpha will help find which stock's movement is increasing
independent of the index and Beta will show me how much
of the return can be attributed to the index. Figure 1.
Here are some simple rules for using Alpha and Beta as
tools for assisting in making more informed trading deci- The results of the Alpha calculation were ranked from 1 to
sions: 30 and sorted into an output page in Excel (Figure 1). In
this example of the Daily, scan AA ranked number #1
1. For identifying stocks making strong positive moves, (highest Alpha) and GM ranked #30 (lowest Alpha read-
look for increasing changes in Alpha with decreasing ing). The Daily scan used a lookback range of 15 days. →
changes in Beta.
2. Stocks with high Beta (generally greater then 1.5) will
exhibit greater volatility then the index. They tend to
overshoot the returns of the index.
3. A stock with a high Alpha reading and low Beta readings
may move up faster then the index and possibly identify
a breakout or short term trade.
4. If the Beta readings are increasing with the Alpha read-
ings you may get strong moves with equivalent reversals
as the increasing Beta may be indicating increased

Dow as an example
In the following examples I constructed a spreadsheet that
includes data for the 30 stocks composing the Dow Jones
Industrial Index, including data for the index itself as my
benchmark. I applied the Alpha Beta code to the daily and
60 min intraday data for the 30 Dow Jones Industrial stocks
over a period of 15 days (daily) and 100 periods (60 mins)

Figure 2.

44 THE TECHNICAL ANALYST January/February 2007

The same techniques as in the Alpha scans were applied to
the Beta calculations and the results were ranked from 1 to
30 and sorted into an output page in Excel (Figure 2). In
this example of the Daily scan, INTC ranked number #1
(highest Beta) and PFE ranked #30 (lowest Beta reading).
The Daily Beta scan also used a lookback range of 15 days.

Figure 5. 15 day Relative Performance chart for AA, DJI,

& GM

Intraday examples:

Figure 3.

The 15 days of ranked results for highest ranked Alpha

security (AA) was plotted against the ranked results for the
Beta calculations of AA to show the relationship of the
changes in the two functions (Figure 3). Here we have a ris-
ing Alpha and a declining Beta. This may be an indication
of a stock being moved by something external from what
is affecting the Markets (DJI) overall return.

Figure 6.

The results of the Alpha calculation were ranked from 1 to

30 and sorted into an output page in Excel (Figure 6). In
this example of the Intraday scan, KO ranked number #1
(highest Alpha) and INTC ranked #30 (lowest Alpha read-
ing). The Intraday scan used a lookback range of 15 peri-
Figure 4. ods of 60 mins each (approx. 2 days).
The same techniques as in the Alpha scans were applied
to the Beta calculations and the results were ranked from 1
to 30 and sorted into an output page in Excel (Figure 7). In
The 15 days of ranked results for lowest ranked Alpha
this example of the Intraday scan, MRK ranked number #1
security (GM) was plotted against the ranked results for the
(highest Beta) and MSFT ranked #30 (lowest Beta reading).
Beta calculations of GM to show the relationship of the
The Intraday Beta scan also used a lookback range of 15
changes in the two functions (Figure 4). Here we have a
periods of 60 mins each (approx. 2 days) (See Figure 7)
falling Alpha and a rising Beta. This may be an indication
of a stock falling out of favour with investors and becom-
ing more volatile in its daily returns.

46 THE TECHNICAL ANALYST January/February 2007

Figure 9.

Figure 7.

Figure 10. 15 periods of 60 mins intraday data Relative

Performance chart for KO, DJI, & INTC

Using the Custom Excel User Functions for Alpha and
Figure 8. Beta, as outlined in this article, and applied to time frames
The 15 periods of 60 mins Intraday ranked results for high- other then monthly returns, (in this case daily and 60 min
est ranked Alpha security (KO) were plotted against the intraday data), can significantly add to a trader's overall per-
ranked results for the Beta calculations of KO to show the spective on underlying market dynamics. One should be
relationship of the changes in the two functions (Figure 8). aware however that Alpha can be very "fragile", meaning
Here we have a rising Alpha and a rising Beta. This may be that it can disappear as quickly as it appears. Therefore,
an indication of a stock being moved by something exter- attempts should be made to determine the root cause of
nal from what is affecting the Markets (DJI) overall return. the excess returns. Often it is investor perceptions rather
The rising Beta may be an indication of increasing volatili- than being based on something substantial so the better
ty too. informed a trader is, the superior will be his or her results.
The 15 periods of 60 mins Intraday ranked results for the
lowest ranked Alpha security (INTC) were plotted against
the ranked results for the Beta calculations of INTC to
show the relationship of the changes in the two functions
(Figure 9). Here we have a falling Alpha and a high but
declining Beta. This may be an indication of a stock falling Ron McEwan is employed with the US Treasury
out of favor by investors and becoming less volatile in its Department. He can be reached via email at
short term returns. (See Figure 9)

January/February 2007 THE TECHNICAL ANALYST 47

February - May 2007

The Society of Technical Analysts (STA) represents and accredits

professional and private Technical Analysts operating in the UK

Society of Technical Analysts Ltd (STA)

Diploma Course
The next STA diploma course will run from 11 January to 3 April 2007 at the
London School of Economics in the Aldwych. The Course consists of 11
Thursday evenings from 6.00pm to 9.00pm starting on Thursday 11 January.

It will be followed by a one day Revision Day (including report writing), on

Tuesday 3 April 2007 from 9.30am to 5.00pm. The course is designed to prepare
students for the diploma examination.

The exam itself lasts three hours and will be held Thursday 19th April 2007.

The course may be suitable for the annual PIA Continuous Professional
Development Programme.

For further, information please visit the STA website ( or contact
Katie Abberton on 07000 710207

Originally established in the 1960s, the STA provides its members:

• Education Monthly lectures and regular teaching courses in technical analysis

• Research The STA Journal publishes research papers on TA techniques and approaches

• Meetings Provide members the opportunity to discuss technical approaches and markets

• Representation The STA lobbies on behalf of analysts with data vendors, exchanges and regulators.

The STA represents the UK at the International Federation of Technical Analysts (IFTA)

• Accreditation The STA Diploma Exam is internationally recognised as a professional level qualification

in Technical Analysis

For more information on how to join and what is involved in passing

the STA Diploma exam, visit our website at: or call

us on +44 7000 710207

The ultimate TA ‘add-on’
for Bloomberg
Scanning, real time alerting and more...

More and more Bloomberg Professional users are running

Updata Technical Analyst to unleash the power of their Bloomberg data

Familiar Bloomberg codes Automatic creation of Risk Reward Ratios and High quality printing
for fast ticker entry Index Constituent Lists price targets for any trade and page setup
Full Windows look and feel Create your own portfolio Chart your own Excel data, Report Writer for
with multiple desktops watch lists Metastock and other documents
High speed scans for Market Maps of any list formats and presentations
any technical criteria of instruments Schedule Automatic Sophisticated technical
Hundreds of TA techniques Chart Bloomberg CIX Chart Runs analysis alerts with screen
and tools covered Indices Optimise and Test any pop ups, SMS messaging
Indicator or email
Draw or add notes, labels, Best Point and Figure
logos and images charts in the world Optimised Trailing Stops

For details: type IDOC UPDATA<GO> on your Bloomberg

Or to arrange a Trial or Demonstration
Call: 020 8874 4747 Email:
All Trade Marks are the property of their respective owners