jan/feb 20 07 Books Barton Biggs .technicalanalyst.co.uk .technicalanalyst.co.uk Hedge Hogging with www futuresbetting.com Automated trading with
Markets Software FTSE 100 Outlook for
HSBC Investments Inside a Trend Following Fund Inside a Trend Following
and Beta Measuring Alpha Measuring Alpha
Trading the VIX Volatility Strategies:
Interview Jack Schwager Jack Schwager HSBC Investments The publication for trading and investment professionals The publication
WELCOME
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.
CONTENTS 1 > FEATURES JAN/FEB
FTSE 100 outlook The technical view for UK stocks remains broadly positive although, as Steven Wesiak of >07 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 >10 equity fund and how he selects the best global stocks.
Interview Jack Schwager >31 Jack Schwager is a well known figure in the 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
CONTENTS 2 > REGULARS
INDUSTRY NEWS 04 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 TECHNIQUES 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 Web: www.technicalanalyst.co.uk Email: [email protected] INTERVIEW Jack Schwager 31 SUBSCRIPTIONS
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: [email protected] BOOKS Hedge Hogging with Barton Biggs 37 ADVERTISING
For information, please contact: [email protected] AUTOMATED TRADING SYSTEMS Measuring alpha and beta using Excel 39 PRODUCTION A machine-learning method for automated trading 43 Art, design and typesetting by all-Perception Ltd. Printed by The Friary Press
ISSN(1742-8718)
January/February 2007 THE TECHNICAL ANALYST 3 Industry News
Reuters offer news DOW JONES TA RESEARCH 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 strategies.
BARCLAYS CAPITAL LAUNCH Alex Rudolph 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 www.market-source.co.uk
4 THE TECHNICAL ANALYST January/February 2007
presents 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 community.
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 www.technicalanalyst.co.uk
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
www.technicalanalyst.co.uk +44 (0)20 7833 1441 [email protected] Market Views
FTSE 100 OUTLOOK: A POSITIVE CHART BUT CAUTION IS REQUIRED by Steven Wesiak
he uptrend that started back in 2003 is still robust and showing Tno 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
EUR/USD: AN ICHIMOKU PERSPECTIVE 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 Ocan 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 [email protected]
“LONGER TERM THESE CHARTS TELL US THAT THE 1.40S IS A VERY POSSIBLE AREA FOR EURODOLLAR TO MOVE TO.”
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 [email protected] or call +1 (646) 733 5000 or +44 (0) 20 7265 3782.
Visit us at www.theice.com.
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. Techniques INSIDE A TREND
10 THE TECHNICAL ANALYST January/February 2007 Techniques FOLLOWING FUND
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 Techniques
outperforming the index, even if the index is trending smoothly so to speak; “BREAKOUT INVESTING TENDS TO BE 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 Techniques
“WE DON’T USE PERCENTAGE STOP LOSSES BECAUSE THEY TEND TO TAKE YOU OUT OF SOME VOLATILE MARKETS THAT PERHAPS YOU SHOULD BE IN.”
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 selection?
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):
www.technicalanalyst.co.uk +44 (0)20 7833 1441 [email protected] Techniques
SPREADING FINANCIAL FUTURES: NEW TECHNIQUES FOR A HIGHLY LIQUID MARKETPLACE 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 of short-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 to trade 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 spread; a short lived but highly active trade based around the in the case of the Eurodollar. This means there are an enor- 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 two differential between the M8 and U8contracts (RED) - (Hourly) July to Dec 2006. Charts → by Reuters Metastock futures contracts. Whilst
January/February 2007 THE TECHNICAL ANALYST 15 Techniques
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 Techniques
Optimisation 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
January/February 2007 THE TECHNICAL ANALYST 17 Techniques
18 THE TECHNICAL ANALYST January/February 2007 Techniques
TRADING THE DAX OPEN 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 Shour 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
Techniques
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 IS ONE THAT EXPLOITS AN OPENING expected gap and its subsequent follow 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 at www.pitchforktrader.com. He can be contacted at: [email protected]
Figure 3.
January/February 2007 THE TECHNICAL ANALYST 21 Techniques
TRADING VOLATILITY: VIX STRATEGIES
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 Tmarket 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 Techniques
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 Techniques
• 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
24 THE TECHNICAL ANALYST January/February 2007 e-yield
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
e-yield To test run our Scoring System call us on 0870 873 8811 or email [email protected] or visit www.eyield.co.uk Techniques
Monday Blues The worst day of the week for stocks?
26 THE TECHNICAL ANALYST January/February 2007 Techniques
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- T 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 explained? 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 Techniques
“MARKET EFFICIENCY ASSUMES THAT OVERREACTION TO INFORMATION IS AS FREQUENT AS UNDERREACTION YET THE FRIDAY/MONDAY EFFECT SEEMS TO REFUTE THIS.”
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
Interview
THE TECHNICAL ANALYST TALKS TO...
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 Interview
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- YOU ARE MORE LIKELY TO 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. RESULTS…IN CURRENCY TA: When you were trading, how did you use TA? MARKETS A TREND JS: For purposes of managing money when I was a CTA, my FOLLOWING APPROACH WILL 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 discretionary? 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 Interview
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
“YOU SHOULDN’T ASSUME THAT ANY INDICATOR WORKS WITHOUT RIGOROUSLY BACKTESTING OVER A BROAD PRICE HISTORY AND ACROSS MANY MARKETS.”
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 trade. 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 Software
AUTOMATED TRADING WITH FUTURESBETTING.COM
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 Tprofits. 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 (www.collective2.com) 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 Software
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 (www.tradestation.co.uk), Figure 1. eSignal (www.esignal.com) and →
January/February 2007 THE TECHNICAL ANALYST 35 Software
“IF YOU BELIEVE IN YOUR STRATEGY THEN YOU SHOULD TRADE IT WHATEVER THE MARKET FEELS LIKE AT THE TIME OF SIGNAL GENERATION.”
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 (www.ninjatrader.com) 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- FuturesBetting.com's gateway to the criteria fit. cution. markets. FuturesBetting.com 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 FuturesBetting.com, line advantage of removing your emo- call +44(0)20 7183 0431 or visit Table 1. tions from the final decision of www.futuresbetting.com
36 THE TECHNICAL ANALYST January/February 2007 Books
HEDGE HOGGING
arton Biggs is a Wall Street veteran who spent 30 years at Morgan Stanley Investment Management before setting up Traxis Partners, a Connecticut Bbased 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 prices. 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 A MACHINE-LEARNING TRADING SYSTEM 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 I 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 MEASURING ALPHA AND BETA
USING EXCEL by Ron McEwan
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 Aof 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 volatility.
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) intraday.
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
Conclusion 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) [email protected]
January/February 2007 THE TECHNICAL ANALYST 47 February - May 2007 GET QUALIFIED IN TECHNICAL ANALYSIS
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 (www.sta-uk.org) 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: www.sta-uk.org 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 or email Draw or add notes, labels, Best Point and Figure Indicator logos and images charts in the world Optimised Trailing Stops
For details: type IDOC UPDATA
All Trade Marks are the property of their respective owners