genetically engineered trading - the technical analyst · genetically engineered trading gann...

4002

hc

ram

Genetically

Engineered

Trading

Gannuntangled

Beware of the cycle theory

Decision time for German bunds

.technicalanalyst.co.uk www

Advertisements - STA and Paritech

March 2004 THE TECHNICAL ANALYST 1

WELCOME

Welcome to the second issue of The Technical Analyst . Ourlaunch issue met wi th a great deal o f enthus iasm and weapp rec ia te t he he lp fu l commen ts and sugges t i ons wereceived. P lease keep sending them in !

As a response to some of these comments, we have expand-ed the Subject Mat ters sect ion to inc lude recent academicresea rch tha t i s r e l evan t t o t echn i ca l ana l ys i s . Suchresearch often doesn ' t reach the technica l analyst communi-ty, w i th pract i t ioners le f t to re ly on a few t r ied-and- testedmethods. However, there ’s a vast amount o f analys is , noton ly in economics and f inance but a lso in o ther d isc ip l inessuch as computer sc ience, mathemat ics and psychology,which is pushing the subject forward. Genet ic programmingand t ime-ser ies model l ing are just two examples and bothare featured in th is issue. I t 's impor tant that these f ind ingsare presented to a wider audience so that they can be con-tested, developed or incorporated in to everyday use.

I f you have any comments that you would l ike to see pub-l ished in a fu ture issue of The Technica l Analyst , p leaseemai l them to me at ed i tor@technica lanalyst .co.uk. I wouldvery much l ike to see our publ icat ion evolve in to an openforum for the exchange of ideas and a p lace in which techni -ca l analys is can progress in a r igorous and cr i t ica l way.

Matthew ClementsEditor

Editor: Matthew Clements (MSTA)[email protected] Editor: Jim BissMarketing: Vanessa GreenSales: Christopher LeighDesign: Paul Simpson

The Technical Analyst is published byClements Biss Economic Publications Ltd,10-12 King Edward's Road, London E9 7SF Tel: +44 (0)20 8533 3025Web: www.technicalanalyst.co.ukEmail: [email protected]

SUBSCRIPTIONS

Subscription rates UK: £275 per annumRest of world: £325 per annumFor information, please contact: [email protected]

ADVERTISING

For information, please contact:[email protected]

PRODUCTION

Art, design and typesetting by all-Perception Ltd.Printed by The Friary Press

ISSN(1742-8718)

© 2004 Clements Biss EconomicPublications Limited. All rights reserved.Neither this publication nor any part of itmay be reproduced, stored in a retrievalsystem, or transmitted in any form or byany means, electronic, mechanical, photo-copying, recording or otherwise, withoutthe prior permission of Clements BissEconomic Publications Limited. While thepublisher believes that all information con-tained in this publication was correct at thetime of going to press, they cannot acceptliability for any errors or omissions thatmay appear or loss suffered directly orindirectly by any reader as a result of anyadvertisement, editorial, photographs orother material published in The TechnicalAnalyst. No statement in this publication isto be considered as a recommendation orsolicitation to buy or sell securities or toprovide investment, tax or legal advice.Readers should be aware that this publica-tion is not intended to replace the need toobtain professional advice in relation toany topic discussed.

2 THE TECHNICAL ANALYST March 2004

CONTENTS

Product News

The Technical Analyst Talks To... Robin Griffiths, chief technical strategist, HSBC Corporate & Investment Banking Division

Book ReviewThe Investor’s Guide To Technical Analysisby Curt Renz

Commitments of Traders Report

Training & Events Diary

04

28

42

44

48


MARCH 2004

07Market Views07 Decision time for German bunds

08 Should gold bulls be running for cover?

10 Outlook for the US technology market

12 Plateau of stability for the US dollar

14Techniques14 The fall and rise of the Advance-Decline Line

16 Using neural networks for the FX markets

18 The Type 1 trade

21 Gann untangled

24 Intraday trading: Revisiting the 1-box reversal

26 Predicting option volatility with point-and-figure charts

Subject Matters30 Genetically engineered trading

34 Beware of the cycle theory

36 Central bank interventions, chartists & the FX markets

38 South-East Asian stock markets follow a non-random walk

40 Research roundup

30

Product News


MTPREDICTOR JOINS UP WITH NEW DATA PROVIDERS

e-yield launches Pairs System

eSignal launches information service for mobiles

BOLLINGER BANDS RELEASE NEW SEMINAR ON DVD

e-yield has launched Pairs System, a new low-risk trading strategy that uses pairs trading to reduce risk by simultane-ously buying and selling two stocks with similar characteris-tics.

Thierry Laduguie of e-yield told The Technical Analyst, “Because of the excesses of the last decade it will be practi-cally impossible to make

money simply by buying and holding stocks for the long term. For this reason we have developed a low-risk trading strategy to take advantage of minor to intermediate trends in stock prices.”

The Pairs System will initially be available to learn online at www.eyield.co.uk and a series of seminars are also being planned.

MTPredictor has released a major upgrade to its End-of-Day trading program and announced partnerships with two data companies, Primate Software of the US and Q-Data of the UK.

According to MTPredictor, End-of-Day 4.0 takes the trader through the complete trading process – identifying set-ups, evaluating the risk/reward outlook, determining trade size

and managing exit-stop strate-gies on-screen.

Partnering with Primate Soft-ware offers customers a datafeed for US futures and stocks, which is directly integrated into End-of-Day 4.0 for $24.95 a month. MTPredic-tor also told The Technical Analyst that the launch of the new Real-Time 4.0 software, with an integrated eSignal datafeed, is imminent.

eSignal has announced the availability of QuoTrek, a new wireless quote and market information service for mobile phones. Chuck Thompson, president of eSignal told The Technical Analyst, “With the new QuoTrek, service, subscribers can access their portfolios and comprehensive market information virtually anywhere and at anytime.”

QuoTrek offers real-time infor-mation on international stocks, indices, futures, options and forex data. Information can be displayed in line, bar or candle-stick charts. The service is available immediately for $49.95 per month for new clients and $25.00 per month for existing eSignal customers.

Bollinger Bands have launched a two-volume DVD. The DVDs are an extension of John Bollinger’s book “Bollinger on Bollinger Bands” and was produced from a two-day semi-nar held in Los Angeles.

The DVDs, which contain more than nine hours of teaching, cover advanced trading topics that are based on Bollinger Bands as well as the basics of technical analysis. Each volume of the DVD is accompa-nied by a book which contains relevant charts and formulas.

John Bollinger

Product News

BLOOMBERG ADDS PRONET’S REAL-TIME FX RESEARCH

FutureSource unveils new software

FutureSource has announced the release of FutureSource Workstation 1.1, their new real-time charting and technical analysis software.

According to FutureSource, Workstation 1.1 is designed for the serious investor and professional trader. The advanced quote component includes support for expres-

sions, Greeks, copy-and-paste DDE and over 50 customizable headings. The advanced chart-ing component includes support for overlays, under-lays, chartable expressions, studies on studies and annota-tions. Tick, intra-day, daily, weekly, monthly and continua-tion charts are included along with over 32 technical indica-tors.

Pronet Analytics will soon be available on the Bloomberg Professional platform. Shane Smith of Pronet says, "Our collaboration with Bloomberg comes at a time when foreign exchange, which is the core market on which we provide

intelligence, has also become a strategic focus for Bloomberg.”

Bloomberg recently partnered with EBS Dealing Resources International to provide an enhanced spot interbank deal-ing facility.

Testers required for SnapDragon RT

SnapDragon Systems will soon

be releasing a new version of

its real-time charting and tech-

nical analysis software, Snap-

Dragon RT, for beta testing.

SnapDragon RT is designed to

work with a variety of datafeeds

and also directly with trading

platforms, thereby eliminating

the need for a datafeed.

The aim is to increase the

product’s flexibility and ease of

use.

If you are interested in becom-

ing a beta tester please email

Adam Hartley at:

[email protected]

Product News

P A R I T E C H E N H A N C E E O D D A T A S E R V I C E

Pfscan adds Fast Track Pfscan 2 has been updated to

V2.93.13 and now includes the

Fast Track datafeed. Fast-

Track is a United States

Paritech has announced a

new look London Stock

Exchange data service cov-

ering end-of-day data for the

LSE, Alternative Investment

Market and major world indi-

ces. The data, which is triple

checked to ensure quality, is

now sourced via Reuters

and can be downloaded in

MetaStock or text formats.

Paritech have also added

US and Australian equity

datafeeds, which are down-

loadable using their Data

Director 2 software.

Updata Technical Analyst now Bloomberg compatible

Updata has launched a new

version of its Technical Analyst

software to run as an add-on to

the Bloomberg Terminal. This

gives Bloomberg users a

windows analysis system

which operates with a host of

technical analysis tools and

indicators, such as INDEXIA

proprietary indicators, not

otherwise found on the Bloom-

berg system.

charting/investment program

that specialises in dividend

adjustment for stock prices and

mutual funds.

European bond market futures have been flirting withhead-and-shoulders formations for some time. But how

likely is it that a head-and-shoulders formation will completeand in turn signal a significant reversal?

Figure 1 is a weekly candle chart for bund futures from May2002 to the present showing three peaks, akin to a head-and-shoulders formation. In November 2003 the neckline wasbroken but there wasn't a big downside push on the back of

the completed pattern. Instead, bunds bounced backand have held above this line since, but without re-testing the old highs. So what we had was a failed pat-tern.

The start of 2004 has seen the market steadily movehigher up to the 61.8% Fibonacci retracement (of themove from the all-time-high to the November 2003low). But the volume on the latest up-move has beenvery low. Whether the head-and-shoulders completes(with the recent highs being the new right shoulder)will depend on selling from here to test the green lineon the chart - the new neckline. It should be stressedthat a head-and-shoulders doesn't exist until this newneckline is broken, i.e. unless we move below the110.90 region. At that point a measured move toaround 103.00 would be expected.

A look at the 5-year bobl could explain why theNovember 2003 break of the neckline didn't producethe desired effect. The bund future is the long end ofthe European futures curve, based on 10-year bonds.A move down the curve to the 5-year bobl future givesa weekly candle chart as seen in Figure 2.

Again, there is the classic left shoulder, followed by ahead, and then a sell-off until early September when ahammer candlestick pattern was posted. This was fol-lowed by a bounce that gave a potential head-and-shoulders formation and, in turn, a neckline. However,the November weakness didn't reach the neckline(unlike the 10-year) and as of late February the neck-line is still in place. So is a new right shoulder currentlyforming for the 5-year future? Or are we going on tomake new highs by trading through 113.67?Something that will give us an early warning on direc-tion is the uptrend line that has defined the recentstrength. If this holds we can expect that move to113.67. If it doesn't we can look for a test of the neck-line at 108.40, with a break here suggesting a move toaround 103. The recent volume has been very light, (atypical right shoulder characteristic), which favours thebear argument.

Clive Lambert is director of FuturesTechs. Hewrites daily analysis on the key European and USfinancial futures markets.

DECISION TIME FOR GERMAN BUNDS: WILL WE GET A HEAD-AND-SHOULDERS? by Clive Lambert


Market Views

Figure 1.

Figure 2.

The gold market has been on a bull run over the last fewyears buoyed by a weak US dollar and low US interest

rates. Medium-term trend following tools have been useful intrying to take advantage of the uptrend from the $253.85 lowof 2001. One of the simplest techniques is to use movingaverages to smooth out market noise, making it easier totrade with the trend. Figure 1 shows a weekly bar chart ofspot gold with 13 and 50-week moving averages. Once theuptrend began in 2001, the 13-week moving average nevercrossed below the 50-week moving average thereby provid-ing a strong trend signal.

Chart patterns were dominated by the double bottom (Figure2) which had a base near $250 and a midpoint near $340,targeting $430. This was met on January 6th, leaving goldbulls waiting for the next uptrend. The monthly futures chart(Figure 3) shows how important the $418/430 area is. If thisis recovered on a sustained basis then chartists will be look-ing for confirmation of a larger basing pattern. Near-term tar-gets are at the $450 and $480 levels, while longer-termextensions would point towards the $600 area. The 13 and50-week moving averages remain bullish and argue in favourof a continued uptrend, but the profit taking at $430 was quiteimpressive and a near-term chart risk seems to be buildingfor a pullback to the 50-week moving average.

The pullback story is a cautious one given the strong uptrendand a US dollar that remains vulnerable to further weakness.

However, Figure 4 shows that the weekly RSI has been turn-ing lower for a few weeks now and that the peaks in this indi-cator have been at lower and lower levels. Not a clear-cutsell signal in its own right but this sort of action is associatedwith a mature trend, which gold now qualifies for. Figure 5shows the bullish resistance line from the $340.50 and $389highs, which was broken on the last uptrend in December2003. Price action has returned to this trendline and if thisline is penetrated on a sustained basis, another negative sig-nal will have been generated, arguing for a shift back to thelower support line near $327 on a range trade move.

Fibonacci retracements can also be used to gauge pullbackpotential as Figure 6 shows. If a correction to the entiremove up from the 2001 low occurs, then the 23.6% retrace-ment at $389 and the 38.2% retracement at $363 are thenear-term targets to watch for. Resistance can be expectedat the highs of the last few weeks just below $418 (the 1996high).

The ability to regain this level would put gold bulls back in thedriving seat. However, as long as $418 holds intact abovehere, a deep pullback in the price of gold towards the $360area, and potentially lower, can be expected.

Gerry Celaya is chief strategist at Redtower Research.

Market Views


SHOULD GOLD BULLS BE RUNNING FOR COVER? by Gerry Celaya

Figure 1.

Figure 2.

Market Views


Figure 3. Figure 4.

Figure 5. Figure 6.

"As long as $418 holds intact above here, a deep pullback... can be expected."

Since the US high-tech indices peaked in mid-January, fol-lowing their surge in December, they have gone through

an extended period of poor action while remaining close torecent peaks.

We use the QQQ Nasdaq-100 tracking index as our primarybellweather for the NASDAQ. As can be seen from the 9-month chart (Figure 1), we are still in the powerful uptrendthat began in March 2003. The end of February upwardmove has re-established the QQQ's above this key medium-term trendline. But having seen four closes recently belowthis line, this trend has now been seriously tested.

Looking at the 3-month timeframe (Figure 2), the market driftof the last month, after the sharp mid-January sell-off, isessentially still with us.The QQQ remains locked between thetwo green lines which mark the top and bottom of the currentboxed range, as investors remain undecided about whetherthis year will bring accelerated IT spending, higher orderrates and faster revenue growth.

But there are positives in this shorter-term timeframe: theholding of a key support at the lower level of the range wherewe now have eight or nine touches at the 36.33 level. Also,there was a high volume day on January 24th that held atthis key level and closed (fractionally) up on the day.

On the negative side, the QQQ has so far made no inroadsat all into the three major down candles of January 28th and29th, and February 19th. It's looking more likely we will seea test of these three important zones. This will determinewhether we go on to see a test of prior highs and a potentialbreakout to new high ground. The level just under $37.50,which is marked by the half-way mark up the most recentlong black candle, is likely to be a tough test. Just above thislevel at $37.50 lies the midpoint of the second long blackcandle of January 29th. These levels in the middle of majortall candles, referred to as the stomach in candlestick chart-ing, often offer excellent support/resistance levels, particularlyin the absence of convincing data elsewhere.

Market Views

OUTLOOK FOR THE US TECHNOLOGY MARKET by Dane Halling


.cnI ,langiSe 4002© - 1

00.03

00.13

00.23

00.33

00.43

00.53

00.63

00.73

00.83

00.93

00.04

1327192620221592225181427101372023169222518252811148212417033261raMbeF4002ceDvoNtcOpeSguAluJ

enil dnert gnisir htnom 9

8730.43

00:42-00:0,cimanyD )D,skcotS .S.U - QQQ(

0

M001

emuloV

M97

.langiSe 3002 © thgirypoC

00.0 :O 00.0 :H 00.0 :L 00.0 :C 00.0 :teN

emuloV )C,9,62,21(DCAM )C,01(COR )C,02(xednI lennahC ytidommoCemuloV

Figure 1.

Why the laboured performance from the QQQ's? Look no fur-ther than the semiconductor indices for some of the answer.The semiconductor fund we track is the SMH (Figure 3).Poor up-volume versus down- volume, lots of near-term over-head supply and multiple support/resistance lines, (startingwith $42), do not paint a picture of an index about to movesmoothly higher after a much needed period of consolidation. But the SMH, as with the QQQ and indeed the NASDAQComposite, has not broken down technically in a major way,and that is really key to understanding the market psycholo-

gy. The 200-day moving average is the gold standard formeasuring the medium-term health of any market. Taking thisbroader perspective, the QQQ and the SMH have bothcaught a particularly nasty strain of flu, but hospitalisation?Forget it, for now at least.

Dane Halling is the founder of Synvestor Ltd, which publishesUS & UK equity market research & technical analysis.

Market Views


Figure 2.

Figure 3.


00.53

05.53

00.63

05.63

00.73

05.73

00.83

05.83

00.93

05.93

132719262022159222518142raMbeF4002ceD

00:42-00:0,cimanyD )D,skcotS .S.U - QQQ(

0

M001

emuloV

M97


00.0 :O 00.0 :H 00.0 :L 00.0 :C 00.0 :teN



00.73

00.83

00.93

00.04

00.14

00.24

00.34

00.44

00.54

00.64

00.74

1327192620221592225181raMbeF4002ceD

8709.73

00:42-00:0,cimanyD )D,skcotS .S.U - HMS(

0

M01

M02emuloV

M31


:O 00.0 :H 00.0 :L 00.0 :C 00.0 :teN 00.0


The trade-weighted US dollar index is currently in a primedowntrend although we are currently seeing a period of

relative stability that is likely to last for another four months.Although it seems clear that being short the dollar is the rightstrategy for the long-term, it is not the best trade for now. Theexisting open short positions are so large that natural marketforces will produce a rally phase.

There is a well-known chart law being invoked here whichstates that the shape of the counter trend correction will alter-nate. Note that in mid-2002 the dollar went into a sidewaysrange for six months. Then in mid-2003, it rallied strongly forthree months and died again. Now from the early Januarytop, a six-month sideways range is the most probable out-come. The implication is that the upside for the dollar is notlarge, but it will not dump again until late June.

For now, the trade-weighted index will probably move in therange 93-97 and euro/dollar should stay in the 119-129range. Dollar/yen is now in exactly the same range as it wasin 2000, 103-110.

From late June onwards, the dollar may well enter the nextperiod of weakness, targeting 80 and taking six months to getthere. In the long run, China will probably re-peg its currencyto a basket including the dollar, euro and yen. Until that hap-pens, any renewed weakness in the dollar will impact on themajors. This implies a sterling target at 2.05, yen at 85 andthe euro at 145. The existing dollar bears have got the rightstory, but the timing is wrong for now.

Robin Griffiths is chief technical strategist within the cor-porate and investment banking division at HSBC Bank inLondon.

Market Views

PLATEAU OF STABILITY FOR THE US DOLLARby Robin Griffiths


MTPredictor TM

The software solution for complete trading excellence

Designed exclusively to find, assess and manage only the

very best trades in stocks, currencies and commodities

This is the type of trade MTPredictor can automatically uncover for you….

A Profit of approximately 7x the initial risk required to take the trade, ignoring slippage and commissions, in the UK stock GKN (October 2003)

TAKE CONTROL OF YOUR TRADING WITH THE IMMINENT LAUNCH OF THE NEW MTPREDICTOR 4.0 SERIES!

MTPredictor Ltd www.mtpredictor.com [email protected] Tel +44 (0) 208 9776191

End-of-Day and Real-time programs with automatic routines for:

· Ideal trades: Find exceptional set-ups with outstanding Risk/Reward prospects· Ideal trade size: Control your position size· Ideal trade management: Display the exit stop strategy on-screen· Ideal trading psychology: Consistent, logical trading, time after time· Systematic Elliott Wave software: Avoid the pitfalls of standard Elliott analysis· Advanced strategies: Expert trade opportunities and management plans

Techniques


Apopular tool for examining the internalcondition of the stock market has

gone awry. The Advance-Decline Line, thestandard measurement of market breadthfor more than 100 years, doesn't seem towork any more. In recent years, at criticalturning points in the equity market, theAdvance-Decline Line has signaled a posi-tive bias just before stocks turned decisive-ly negative.

In simple terms, the Advance-Decline Lineshows whether more stocks are going upthan down. More importantly, it can showwhether buying enthusiasm is spread acrossa broad number of stocks (positive), orwhether buying is narrowly focused (nega-tive). It provides investors with a simple

way to adjust the level of diversification intheir portfolios. That is, as the Advance-Decline Line expands, investors can expandthe number of stocks in their portfolios. Asthe Advance-Decline Line contracts,investors should become more selective byexcluding weak holdings.

Perhaps the most widely accepted value ofthe Advance-Decline Line is in "diver-gences" - when the Advance-Decline Line(market breadth) improves during a periodof market weakness (bullish signal), orwhen it contracts during a period of marketadvance (bearish signal). The Advance-Decline Line has often turned down aboutfour to six months before many major mar-ket declines in the past. Because of its ele-

gant simplicity and the valuable insights ithas provided at market turning points, it hasbeen highly prized by analysts throughoutthe decades.

But in recent years something seems tohave gone wrong. For example, in July andAugust 2001 (Figure 1) when the DowJones Industrial Average (DJIA) was mov-ing in an indecisive pattern, the Advance-Decline Line began to rise vigorously, lead-ing investors to conclude that the internalstrength of the market was improving.

Many investors bought aggressively inAugust 2001, based on the strength of theAdvance-Decline Line. But the market wasactually weakening. The Advance-DeclineLine was giving off a false signal. TheDJIA plunged 20.7%, exacerbated by thetragedies of September 11th.

In early September 2002 (Figure 2), theIndustrial Average was recovering fromprevious losses. The Advance-Decline Linelead the recovery, rising to a new rally high,creating the impression that buying enthusi-asm was broadening. In fact, the marketwas weakening and the DJIA dropped15.3% to a market low in October 2002.

Again, in mid-January 2003 (Figure 2), theIndustrial Average appeared to be breakingout to new rally highs. The Advance-Decline Line added to the illusion by risingto its highest level in six months, a seeming-ly bullish indication that investors werepouring money into a broadening array ofstocks. Many investors rushed in to buyonly to find that the Advance-Decline Linewas once again providing misleading infor-mation. Over the next two months, theDJIA plunged 14.9%.

How could the time-honoured Advance-Decline Line give off such obviously falsesignals? The answer is simple but not easi-ly seen. Over the past decade, the New

THE FALL AND RISE OF THE ADVANCE-DECLINE LINE by Paul F Desmond

Figure 1.

Techniques


York Stock Exchange (NYSE) has allowedtrading in a growing number of issues thatare not, or do not trade like, domestic com-mon stocks. For example, of the 3,500issues listed on the exchange, approximate-ly five hundred are closed-end bond funds.Another six hundred issues are preferredstocks that trade more like bonds than com-mon stocks. In addition, there are morethan four hundred foreign stocks andAmerican Depositary Receipts (ADRs) thatmay or may not reflect the trends of thedomestic stock market. Lastly, roughly threehundred real estate limited partnerships arelisted on the NYSE that, like mutual funds,are not operating companies. The bottomline is that almost half (48%) of the issuescurrently listed on the NYSE are not reallycommon stocks - at least not what investorsgenerally think of as stocks. These non-operating companies have been the princi-

pal cause of the false signals given off bythe Advance-Decline Line in recent years.In each of the three cases cited above, com-mon stocks were in a generally sidewaystrend, while the bond market was stronglyrising. Thus, the common stock compo-nents of the Advance-Decline Line offsetone another, while the bond related compo-nents were rising strongly, giving theAdvance-Decline Line an overall positivebias. In other words, during each of thoseperiods, the Advance-Decline Line was, inessence, measuring the strength of thebond market, not the stock market. It's nowonder that the signals were misleading.

This is not a recently discovered phenome-non. In 1990, Lowry's Reports becameconcerned about the changing character ofthe Advance-Decline statistics. As a result, anew indicator, Lowry's Operating-

Companies-Only (OCO) Advance-DeclineLine, was compiled, which excludes all pre-ferred issues, real estate partnerships, for-eign issues and ADRs, and closed-end stockand bond funds. The remaining issues aresimply domestic common stocks listed onthe NYSE.

Throughout its twelve year history, Lowry'sOCO Advance-Decline Line has provided afar more accurate measurement of theinternal strength of the stock market, par-ticularly at critical turning points, such asthose cited above. As shown in Figures 1and 2, in each of the three cases Lowry'sOCO Advance-Decline Line was correctlyreflecting market weakness at exactly thesame time that the standard Advance-Decline Line was showing misleadingstrength.

New distortions to the accuracy of NYSEtrading data may be coming. The NYSE isworking aggressively to add ExchangeTraded Funds (ETFs) to its listed issues.The trading of ETFs (which essentially willdouble-count issues already listed) will playhavoc with not only the Advance-Declinestatistics, but more importantly with theupside and downside volume statistics thatinvestors have relied on over the years.Interestingly, Lowry's Reports originatedthe compilation of the Upside andDownside Volume statistics in 1938.Lowry's has also been compiling Operating-Companies-Only Upside Volume andDownside Volume statistics since 1990. Inan ever changing world, traders, investorsand analysts must constantly re-examine theaccuracy of the data that plays such a vitalrole in their trading and investment deci-sions.

Paul Desmond is president of Lowry'sReports, Inc. He is a past president ofthe Market Technicians Associationand was winner of the Charles H. DowAward in 2002.

Figure 2.

Techniques


What is the best combination of tech-nical indicators and oscillators

required to predict short-term movementsin the foreign exchange markets?

Initial research at Nostradamus found thatby using neural networks, we could elimi-nate all human bias when selecting an indi-cator. To do this, we took 110 well-knownmarket indicators and oscillators (e.g. RSIand moving averages) and applied them tofive years of historical data. The neural net-work then chose 40-50 indicators, weightedinto the best combination, to make predic-tions of high and lows for the next two andeight hour periods. As a result, accurate pre-dictions were achieved. For USD/DEM,the average error was around 15 pips in anygiven week (60 periods of two hours).Similar accuracy was attained with theinception of the euro (error rates are nowunder 7 pips on average).

Further research created a proprietary indi-cator, now called COMPASS, which meas-ures the direction and strength of momen-tum in market movement. Taking data fromReuters, the indicator is calculated on a onesecond basis, based on the current bid priceof the underlying currency pair.

The COMPASS dial frame shows twoextremes of upward or downward trendingmomentum, edged with reversal/signalareas to indicate the point where momen-tum is being lost (Figure 1). This is especial-ly useful for options traders looking tohedge their underlying spot position as itgives turning points at which they shoulddecrease their hedging cover after a trend isover.

The neural network updates its choice ofindicators and weighting at the end of eachperiod having learnt from the movementand volatility of the previous two or eighthours' data. This eliminates any human biasin judging the market and leaves the traderfree to make his own decision on his exact

time of entry. A glance at the historical plotof COMPASS over the previous 8 hours(Figure 2) gives an appreciation of the nextpossible signal to come and confirms ordenies his current or anticipated prediction.

Neural networks take a more advancedcombination of indicators than any individ-ual could hope to monitor and weigh interms of effectiveness. Combining this

powerful momentum and directional toolwith predicted future levels gives the tradera real-time appreciation of when a levelmight break due to the force of momentumin the trend.

Tim Finch is managing director ofNostradamus Systems Ltd.www.nostradamus.co.uk

USING NEURAL NETWORKS FOR THE FX MARKETS by Tim Finch

Figure1.

What are neural networks?

A neural network is a system of programs and data structures that approx-imates the operation of the human brain. It usually involves a large num-ber of processors operating in parallel, each with its own area of knowl-edge and access to data in its memory. Typically, a neural network is ini-tially trained or fed large amounts of data and rules about data relation-ships (for example, a grandfather is older than a person's father). A pro-gram can then tell the network how to behave in response to an externalstimulus.*

In technical analysis, neural networks refer to computer programs that areable to identify price patterns, which are then interpreted to generate trad-ing signals. For the end-user, they are a form of automatic trading systemin which the computer, rather than the trader or analyst, identifies themost relevant patterns.

* Source: TechTarget

Techniques


Figure 2.

“The neural network updates its choice of indicators and weighting at the end of each

period having learnt from the movement and volatility of the previous two or eight hours' data.”


As explained in last month's issue ofThe Technical Analyst, one lesson

learnt over years of trading is that the sim-ple approach to analysis is often all that isneeded to uncover excellent trading oppor-tunities. Arguably, one of the clearest tradeset-ups for a professional trader should bethe simple ABC correction.

The last article outlined not only how thisclean ABC correction can yield profitabletrade opportunities, but also how simpleand easy-to-recognise it could be. Takingthis a stage further, when the basic ABCcorrection occurs at a certain juncture in amarket, it can produce one of the mostprofitable trades available. This happenswhen the simple ABC correction developsas part of the first correction to the firstmove off an important high or low. InElliott Wave terms, this is a Wave 2 or Bcorrection. When this correction is com-plete, it can lead into a Wave 3 type move,which is usually the strongest and longestswing in a typical 5-wave sequence. Thisswing has the largest profit potential in anyElliott Wave pattern.

Figure 1 presents a simple ABC correctionon the UK equity ICI. The most importantpoint is that this ABC correction unfoldedas part of the first correction to the initialswing off an important low.

ICI made a major low in March 2003, whichwas followed by an initial rally off the low.The simple ABC pattern then appearedduring the correction to this initial rally(Figure 2). This is referred to as a Type 1trade set-up in MTPredictor.

Why is this set-up so important? Very oftenthis initial correction is the springboard foran extremely strong move. Moving forwardin time we can see how, once this particularcorrective ABC was complete, ICI proceed-ed to rally by over 75%, from a price of 124to approximately 220 (Figure 3).

Clearly, identifying these set-ups can resultin enormously profitable trades. This is whythe Type 1 trade is the principal trade set-upsought out of the three ABC correctivevariants. More importantly, this set-up is

straightforward to identify, with its mathe-matical simplicity - a simple ABC correc-tion that unfolds as part of the first correc-tion to the first move off an important highor low.

THE TYPE 1 TRADE by Steve Griffiths

Figure 1.

Figure 2.

Techniques

Techniques


Figure 3.

In the example shown in Figure 4, beingable to identify the end of the correctionenabled the trader to enter a very profitabletrade precisely at the start of the strongmove. This has the major advantage ofreducing the initial risk on the trade, partic-ularly compared with the potential profitfrom these Type 1 set-ups.

Expanding on this theme, Figure 4 showshow the ABC correction developed exactlyat the Typical Wave C Wave Price Targetsupport zone. (This support/resistancezone was covered in last month's article).Anticipating that the market was reversingat this zone enabled the trader to go longfrom 124 (ignoring slippage and commis-sion) with an initial protective stop loss at118 - an initial risk of only 6 points.

Now compare this with the potential prof-it as ICI rallied to approximately 220, aprofit of 96 points. Given the initial risk ofonly 6 points needed to take the trade, thisis a spectacular risk/reward ratio. Overtime, this can result in a track record wherethe losses are kept small relative to theprofits - a prerequisite for low risk/highreturn trading.

This particular trade set-up can unfold inany market and on any timeframe (fromweekly charts to five minute charts), sowhether you trade UK equities, US equities,futures or currencies, the identification ofthis set-up arguably should be a major partof your trading plan.

Steve Griffiths is managing director ofMTPredictor Ltd and developer of theMTPredictor software. Figure 4.

Efficient Distribution

Efficient distribution of market data and analytics

services to the major players in the financial

markets industry helps them to successfully achieve their

goals. In particular, TraderMade has provided

technical analysis services to major banks, corporations,

fund managers and individuals for nearly 20

years.

If your job requires you to distribute your technical opinions to your clients, you need an efficient distribution channel. Imagine this… you have just completed your EUR-USD technical analysis report and you need to send it to your clients… with TraderMade WebLink, it is already done! A dynamic link between your TraderMade charts and your web site ensures the moment you save your analysis, your clients will see it. Contact us today to get ahead of the game.

Tradermade International Limited 8 Archers Court 48 Masons Hill Bromley Kent, BR2 9JR

www.tradermade.com Europe : +44 020 8313 0992 Fax : +44 020 8313 3179 e-mail : [email protected]

Techniques


The use of percentage movements hasbeen far and away the most important of

Gann's techniques, enabling identification ofsignificant future price levels. Gann observedthat the level of future tops and bottoms hasa direct mathematical relationship to previoustops and bottoms. Gann theory should there-fore enable the analyst to project and forecastprice levels.

Like Fibonacci ratios, Gann percentages arebased on natural laws. In Gann's case this wasthe observed mathematical relationshipbetween musical harmonics. When observingthe movement of shares and other invest-ments, Gann noted that percentage move-ments from significant past levels woulddetermine the point where rises and falls werelikely to terminate. In other words, when amarket is in harmony, it reverses a previoustrend by an amount deduced through the use

of simple mathematical ratios.

Gann's original percentage levels were derivedby dividing price moves into eighths andthirds. These have since been added to bysmaller, but related, percentages, such as six-teenths and sixths. Table 1 sets out the impor-tant percentage levels.

The next high or low of a price move shouldbe at a level related by a Gann percentage to aprevious low or high, with the primary per-centage movement of 33% being the mostnotable. Once the primary percentages havebeen identified, secondary percentages can beused to confirm the level. The 50% move-ment is the most significant of these.

If the price moves through a secondary per-centage from a market top or bottom, thenthe corresponding primary percentage will

give you the new price target of the currenttrend. For instance, a move from a low or highthrough 25% will almost certainly be curtailedat 33.33%, at least temporarily, and perhapsfor many years. This explains the differencebetween the two sets of numbers and the rea-son for the dominance of the primary set.

During 2003, 25 stock market buy signalswere triggered by the use of percentage move-ments. Taking a medium-term view there wereno losses, with the least successful trade pro-viding a profit of 9% for Portugal and thebest return being 178% for Venezuela.

The following three examples illustrate theuse of these percentages in identifying per-centage moves: (see pages 22-23)

Fred Stafford is chairman of GannManagement.

GANN UNTANGLED by Fred Stafford

Primary

percentages

4.1658.33

16.6633.3366.66100

Secondary

percentages

3.1256.2512.5255075

Table 1.

"The next high or lowof a price move

should be at a levelrelated by a Gann

percentage to a previous

low or high."

Techniques


Figure 1.

1.

2.

3.

This area was the 50% retracement ofthe November 2000 to May 2002 pricerise

A fall of 33.33% from the May 2002high

A fall of 25% from the January 2003high.

The market turned at 1030.

UK Gold Mines (Figure 1)

On 3 March 2003 we anticipated a low in the1000/1030 region for UK Gold Mines. The rea-sons for this were:

Figure 2.

1.

2.

3.

4.

A rise of 50% from the July low

A rise of 25% from the October low

A rise of 16.66% from the November 6thlow

A rise of 8.33% from the November 24th

The market then fell 16.66% from the January 5th2004 high and then retraced 50% of the fall. 1600is the next level of interest, this being 8.33% fromthe previous high. This is close to the last 16.66%fall at 1580 and the 50% retracement of the July toDecember 2003 rise at 1556.

UK Gold Mines (Figure 2)

UK Gold Mines recently been trading strictly with-in Gann's rules. The recent rise was terminated by:

Techniques

Figure 3.

1.

2.

3.

This area was the 50% retracement of theNovember 2000 to May 2002 price rise

A fall of 33.33% from the May 2002 high

A fall of 25% from the January 2003 high.

The Dow Jones (Figure 3)

The flat performance of the Dow Jones IndustrialAverage in 2004 can be explained by the small mag-nitude of the Gann retracement levels:

Techniques


Iam often asked by experienced traderswhether point-and-figure charts can be

used for intraday real-time analysis. Myanswer every time is yes.

Point-and-figure charts were devised bytraders and tape readers to record the tick-by-tick prices from the floor or the tickermachine. They were the first kind of real-time chart that technical analysts ever usedand, although they are currently out-of-fashion, they continue to offer a powerfulbasis for short-term trading.

To draw intraday point-and-figure charts,you need a real-time feed. If, however, tickdata is not available, then other intradaytime frames such as every 15 minutes, whilenot ideal, may still yield meaningful results.Even if there is insufficient data for daytrading, relatively low frequency intradaydata can still provide crucial informationfor improving longer-term trading deci-sions.

The original point-and-figure charts were 1-box reversal charts. These are rarely seentoday and when they are, they are oftenconstructed incorrectly. The reason forusing 1-box charts is that they have aunique condition whereby X and O canoccupy the same column. This cannotoccur with any other point-and-figure chart.

An X and O in the same column is a pow-erful sign. For example, the presence of asingle O in an uptrend column of Xs indi-cates strength in the trend. It means thesellers have managed to force the priceagainst the trend by the value of one O, butthe buyers have seen it as an opportunity tobuy again and have pushed it higher toretake the X. I call this the 'one step back'.It is a pause for breath, a sign of a resump-tion in the trend.

The circled areas in Figure 1 show this con-dition. The price is forced down by one box

creating a O, but immediately the bulls takecharge again and push it back up by onebox resulting in an X above the O, so wehave X and O in the same column. Noticethat this happened three times during theup trend. But notice also what happens atthe top of the trend in the rectangular sec-tion. As before, the sellers push the pricedown by one O and the bulls push it backup by one X, but this time they can't get asecond X above. They make four attemptsat doing this, with the sellers forcing itdown and the buyers forcing it up eachtime. Finally the sellers push it down by oneO again but this time the buyers don't pushit back up by one X again. This showsweakness on the buying side. When the sell-ers push it down by another O, a point-and-figure sell signal is given and a downtrendstarts.

Notice that the reverse happens in thedowntrend. We have the 'one step back' inreverse. The three square sections on thechart show it clearly. During the downtrend,the buyers push the price up by one X and

the sellers immediately push it back downby one O, which then continues the down-trend. As with the uptrend, when thisoccurs during a downtrend, it is a trendenhancer. It is like a test by the bulls to seewhether the bears are serious.

Some will argue that the 1-box reversalchart is too sensitive and can only be usedfor day trading. This is not the case.Changing the box size (the value of each X& O) reduces the sensitivity. Figure 2 is a 5x 1 chart (5 pence box size, 1-box reversal)and shows the one step back during the Xcolumn uptrends in the same way.

One of the problems in using tick data isthe shear amount of data that has to bestored. A few months of a vigorously trad-ed instrument runs into tens, maybe hun-dreds of thousands of ticks, which means itis only practical to store a few monthsworth of data. This however does not pre-vent you from using 1-box reversal charts.They are not restricted to short-termtraders. Although true point-and-figure

INTRADAY TRADING: REVISITING THE 1-BOX REVERSALby Jeremy du Plessis

Figure 1.

Techniques


charts do require tick data, they can be usedwith any time frame, even daily, to goodeffect.

One of the most powerful advantages ofpoint-and-figure charts is the simple way inwhich it portrays sideways congestion with-in a range. The 1-box reversal chart showsthis congestion better than any other chartbecause every change in direction is record-ed. This allows us to have a visual under-standing of the contest between the buyersand sellers, which helps us to predict theextent of any move out of these sidewayspatterns. This is done using what is calledthe point-and-figure count, which is basedon the width of the pattern and projectingit up or down from the point of breakout.The logic being that the more vigorous thecontest, the further the price will move onthe breakout.

Using a 1 x 1 tick point-and-figure chart ofReuters again, you can see in Figure 3 howaccurate these counts have been in predict-ing Reuters' rise in the last 2 months.

Count A established in early January gives atarget of 313 which was achieved twoweeks later. Count B established in earlyFebruary gives a target of 402 which wasachieved two weeks later. Count C estab-lished in mid-February gives a target of 403which matches count B and was achievedon 17 February. Finally count D establishedon 13 February gives a target of 450 whichwas achieved on 18 February.

If you are a user of point-and-figure charts,then spending time with the 1-box reversalshould enhance your trading.

Jeremy du Plessis is head of technicalanalysis at Updata plc and is responsi-ble for the design of the UpdataTechnical Analyst software. He is aChartered Market Technician and aFellow of the Society of TechnicalAnalysts.

Figure 2.

Figure 3.

Techniques


The point-and-figure charting method isexcellent for analysing and predicting

price volatility, an essential part of optionstrading. However, because volatility ismean reverting (i.e. volatility tends to revertto its long-term average), point-and-figurecharts need to be adapted before they canbe used for this purpose.

Scaling the volatility chart

First, an appropriate box size must be cho-sen. Based on the authors’optimisations, alogarithmic scale of 5% is recommended asa growth factor from one box to the next.That is, a box is always 5% bigger than thebox immediately below.

Given both volatility and logarithmic boxsizes are measured in percent, the measure

of volatility is referred to as volatility per-cent or V% to avoid confusion. In Figure1, the long-term mean is 15.9 V%, with thehighest level around 40 V% and the lowestaround 8 V%. Applying this to the boxscale, if a box is set at 10 V%, the nexthighest box is 5% higher, which is 10.50V% and the box below is set at 9.50 V%.

To generate the 5% logarithmic scale forvolatility point-and-figure charting, start ata level of volatility that is lower than the all-time low for volatility. 5 V% is often anappropriate starting point. So the first boxis 5 V%. Then calculate the second boxlevel by multiplying 5 V% by 1.05 to give5.25 V%, the third box number by multiply-ing 5.25 V% by 1.05 to give 5.51 V% and soon until the entire range from the all-timelow to the all-time high has been spanned.

Interpretation

In equity point-and-figure charts the signalsdominate, whereas in volatility point-and-figure the trendlines govern. This meansthat volatility charts are primarily analysedaccording to the channels given by thetrendlines and, secondly, according to thebullish or bearish quality of the chart,which simply refers to the last signal. Thus,if the last signal was buy, then the chart isbullish, if it was sell, then the chart is bear-ish.

Once these adaptations have been made(see box), point-and-figure charts can pro-vide clear trendlines from which to forecastvolatility. The chart of the Nasdaq volatil-ity index in Figure 2 illustrates this clarity. Itis based on a 3-box reversal and a logarith-mic 5% scale. The trendlines defining thebearish channel are evident - (please refer tolast month's issue of The Technical Analystfor details on how to draw trendlines). Twoinstances where the bearish resistance linewas touched are indicated in green. Thesecond instance is quite recent, November2003, but as in the first instance, December2002, the trendline was not broken by thebullish X-column.

Therefore, the trendline is terminated and anew one has to be drawn, shifted towardsthe right. (Note: a bearish trendline has tobe broken by an X-column, a bullish trend-line by an O-column). A break of the resist-

PREDICTING OPTION VOLATILITY WITHPOINT-AND-FIGURE CHARTS by Heinrich Weber and Kermit Zeig

Figure 1. The volatility of the S&P500 index. The red line marks the long-term average,15.9%. Volatility is clearly attracted to this red line, a feature that differentiates volatility fromprice.

Adapting point-and-figure for volatility

1. Use a 5% log scale 3-box reversal chart2. Trendlines govern3. Buy signals on high levels and sell sig-nals on low levels are systematicallyignored4. On long poles, i.e., long columns of Xsor Os, positions are neutralised on a rever-sal (three boxes in the opposite direction)

Techniques


ance line would normally provide a signal tobuy volatility (perhaps through a long strad-dle position).

Figure 3, a line chart for the same period ofthe Nasdaq volatility index, shows the diffi-culty of working with trendlines on thistype of chart. Three possible trendlineshave been drawn, all of which are valid.

Point-and-figure, once adapted, is an excel-lent tool for the analysis of volatility time-series. Trendlines are easier to draw inpoint-and-figure and do not suffer from theambiguity often associated with other charttypes. As well as informing your tradingdecisions, such as deciding whether to selloptions naked or take out covered calls,they can form the basis of any analysis thatrequires a view on future volatility.

Heinrich Weber is a partner and riskmanager of DeTraCo, an active Eurexmember firm. Kermit Zieg is a FullProfessor of Finance and Investmentsat Florida Institute of Technology'sNational Capital Region campus inAlexandria, Virginia.

Figure 2. The Nasdaq volatility index

Figure 3. The Nasdaq volatility index

“...in volatility point-and-figure,

the trendlines govern.”

TTA: HSBC publishes a great deal of market researchthat includes a lot of technical analysis. How many ana-lysts do you have in your team in London?

RG: There are three analysts in the corporate andinvestment banking division. Alan Johnson covers Asiaand Charles Morris is head of technical analysis withininvestment management. Obviously, we also have ana-lysts in our other banking divisions such as foreignexchange and capital markets.

TTA: What software and charting packages do you use?

RG: I use Bloomberg for all my charting analysisalthough we also have Thompson Financial Datastreamto construct historical charts and plot some moving aver-ages. Once one becomes fully familiar and adept atusing one system the tendency is to stick with what youknow.

TTA: What markets do you look at?

RG: We look at all the major markets from foreignexchange to energy although the equity markets proba-bly occupy most of our time. Our research is publishedin a monthly global strategy report and we also producea fortnightly FTSE chartbook which includes moredetailed analysis of individual stocks.

TTA: Do you prefer to use any particular type of chart?

RG: I use bar charts almost exclusively for my analysis.Despite the excellent software available for other charttypes such as candlesticks and point-and-figure, timerestrictions mean I tend to use only one type.

TTA: How do you assess the performance of your analy-sis?

RG: If our analysis is below par then we don't get theorders. It's as simple as that. My clients expect morethan just an insight into technical analysis. If my fore-casts impact on their portfolios and I'm wrong, then they

Interview


THE TECHNICAL ANALYST TALKS TO….

Robin Griffiths

Robin Griffiths has more than 25 years experience

in technical analysis. He is chief technical strate-

gist within the corporate and investment banking

division of HSBC Bank in London and previously

held technical analysis roles with HSBC Securities

in New York, W I Carr and Grieveson Grant. He is

also a past chairman of both the International

Federation of Technical Analysts and the Society of

Technical Analysts.

Interview


"I pay close attention to percentage deviations from the moving averages.

These provide a more quantitative method of using MAs."

lose money. Here technical analysis differs from funda-mental analysis in that I'm expected to anticipate futuremoves in the market, not merely explain what's happen-ing. If you are consistently wrong then it will be noticed.

TTA: Are there any particular indicators that you relyon?

RG: I use 25-day and 200-day moving averages forshort and long-term analysis. However, I also pay closeattention to percentage deviations from the movingaverages. These provide a more quantitative method ofusing moving averages rather than relying upon a pure-ly graphical interpretation.

TTA: What fundamentals, if any, do you look at?

RG: I attach a great deal of value to the analysis ofdemographics for my longer-term views. This is a some-what overlooked area of research that has significantimplications for economic fundamentals and market lev-els over a longer-term horizon.

TTA: What’s the importance of demographics to themarkets?

RG: Demographics and shifts in a country's age/popu-lation profile impact on future savings ratios which, inturn, has implications for stock market levels.

TTA: What are your favourite patterns and techniquesthat you use for forecasting?

RG: I'm a big fan of Elliott Waves and Gann analysis.My Elliott Wave analysis has proved successful in antic-ipating movements in the Dow and euro/dollar over thepast two years. Gann retracements of 50% and 100%have also proved reliable in estimating market correc-tions in certain stock and currency markets. You'll haveto become a client to find out more about my forecastsfor the year ahead!

Subject Matters


Don't worry about finding the best tech-nical trading strategy. Let your comput-

er do it for you and it will formulate, backtestand modify a strategy over and over again,such that its development is akin to the evo-lution of a species through millions of yearsof natural selection. This is the messagefrom the relatively new branch of computerscience, genetic programming, the principalsof which are rooted in evolutionary theory.

John Koza, a consulting professor atStanford University, explains the concept,"One of the central challenges of computerscience is to get a computer to do whatneeds to be done, without telling it how todo it. Genetic programming addresses thischallenge by providing a method for auto-matically creating a working computer pro-gram from a high-level statement of theproblem. Genetic programming achieves thisgoal of automatic programming by genetical-ly breeding a population of computer pro-grams using the principles of Darwinian nat-ural selection and biologically inspired opera-tions."

Our study

Genetic programming has recently beenapplied to the financial markets and, specifi-cally, technical analysis. In 2003, the authors

of this article showed that an evolved tradingrule could outperform a buy and hold strate-gy on the S&P500. To do this, we used sev-eral monthly technical data sets for theS&P500 from 1960 to 1990 as the materialfrom which the trading rules could evolve(see Box 1). We then let the computer maketen evolutionary runs, with each run goingthrough 100 generational steps.

The 10 evolutionary runs were carried outunder three different environmental condi-tions, giving a total of 30 runs:

Experiment 1: 10 runs using the total port-folio value as the fitness function (i.e. themeasure of performance).

Experiment 2: 10 runs using the greatestnumber of 12 month periods with well-per-forming returns as the fitness function.

Experiment 3: 10 runs with the aim ofproducing separate trading rules for when tobuy and when to sell. This process, wherebythe two rules are considered separate speciesand are therefore developed independently, isknown as Coevolution. The fitness functionused was the number of 12-month periodswith well performing returns as inExperiment 2 above.

Mating

Figure 1 illustrates how two trading rulesmight have been mated to produce two newtrading rules in one generational step. Thegenomes (subtrees) to be exchanged aremarked by a dashed box.

The top left-hand trading rule in greenreads: Buy the S&P500 if a) the 3-monthmoving average is less than the 6-monthmoving average OR b) if the 12-month mov-

GENETICALLY ENGINEERED TRADINGby Mukund Seshadri and Lee Becker

opening, closing, high, low of currentmonth

2, 3, 6 and 10-month moving averages

3 and 12-month rates of change

price resistance markers - two previous 3-month moving average minima and twoprevious 3-month moving average maxi-ma

trendline indicators - a lower resistanceline based on the slope of the two previ-ous minima and an upper resistance linebased on the slope of the two previousmaxima

Box 1.

The technical datasets used

ing average is less than the 3-month movingaverage AND the 6-month moving average isgreater than the month-end closing price. Ifyou are already long, then sell if both a) andb) become false.

The top right-hand rule in blue reads: Buythe S&P500 if a) the lower trendline is lessthan the month-end closing price OR b) ifthe first local maximum of the 3-monthmoving average is greater than the secondpreceding local minimum of the 3-monthmoving average. If you are already long, thensell if both a) and b) become false.

The new trading rules created by crossovermating are then tested for fit against theS&P500 data from 1960 to 1990, and theprocess of mating and testing then continuesuntil the evolutionary run is complete.

Forecasting performance

The best performing rule from each run washarvested and then used to make forecastsfor the out-of-sample period, 1991 to 2002,allowing a comparison of their forecastingperformance (based on an initial investmentin the S&P500 of $1,000 at the beginning ofthe period- see Table 1).

It is worth noting that the coevolved trad-ing rules (which give separate rules for whento buy and when to sell) gave the best aver-age out-of-sample performance. The bestsingle trading rule, however, with a return of$4,001 in the out-of-sample period, wasderived from the "consistency of perform-ance" fitness function in experiment 2 (seeBox 2).

Our use of genetic programming to derivetechnical trading rules is by no means thefirst of its kind and its application does notas yet present a miracle answer to technicaltrading. Previous attempts to use genetic pro-gramming for acquiring technical tradingrules had not been able to establish that GP-evolved technical trading rules could outper-form a buy-and-hold strategy if transaction

Subject Matters


OR

< AND

MA3 MA6 < >

MA12 MA3 P MA6

OR

< >

LT P

OR

< >

MA12 MA3 MX1(MA3) MN2(MA3)

MX1(MA3) MN2(MA3)

CROSSOVER MATING

OR

< AND

MA3 MA6 < >

LT P P MA6

Figure 1. Genetic Trees - Crossover Mating

Ave performance in Experiment 1



Buy and hold

Out-of-sample(1991-2002)

In-sample(1960-1990)

$5,457

$11,269

$7,990

$14,521

$2,638

$2,812

$3,014

$3,142

Table 1.

Subject Matters


costs were taken into consideration. Ourstudies have since described an approach thatincludes transaction costs which can outper-form a buy-and-hold strategy, at least if divi-dends are excluded from stock returns (seeBox 3).

The findings of our research call intoquestion the Efficient Market Hypothesisand provide strong evidence to suggest thedevelopment of genetic programming maysoon be able to help traders exploit price pat-terns in a systematic way. More powerfulcomputing and/or the consideration of evenmore technical datasets at the outset, shouldlead to significant developments in this field.

References:

Koza, J. (1992) Genetic programming - On the program-ming of computers by means of natural selection. MITPress, Cambridge, MA.Allen, F. and Karajalainen, R. (1993) Using GeneticAlgorithms to Find Technical Trading Rules, Rodney L.White Center for Financial Research, The WhartonSchool, Technical Report 20-93

Neely, C.J., Weller, P.A., and Dittmar, R. (1997) Is tech-nical analysis in the foreign exchange market profitable? Agenetic programming approach, Journal of Financial andQuantitative Analysis, 32, pp. 405-426.

Mukund Seshadri and Lee A. Becker ([email protected], [email protected])Department of Computer Science,Worcester Polytechnic Institute, UnitedStates.

Box 3.

A NEW APPROACH

We incorporated a number of significant changes from that of previous researchers such as Allen & Karjalainen and Neely.These include:

using monthly as opposed to daily data in order to reduce the number of transactions, with the result that there is only atransaction, on average, once every two years (transaction costs are assumed to be 0.5% for each buy or sell)

reducing the operator set and increasing the number of technical indicators

using a complexity-penalizing factor in the fitness function to avoid overfitting as well as to improve comprehensibility. Theresult being that the forecasting ability of the trading rules improved for the out-of-sample period (1991 to 2002)

using a fitness function which considers the number of periods a rule performs well, and not just its total return or averageexcess return

using co-evolution of a specialized buy rule and a specialized sell rule.

If you are out of the market and either A becomes true or B becomes true, you should buy.

If you are in the market and both A and Bbecome false, then you should sell.

B

A 12-month rate of change < 3-month rate of change

1st local maximum of the 3-month moving average > 2nd preceding local minimum of the 3-month moving average.

Best performing trading rule

Box 2.


Subject Matters

Is the Singapore stock market cyclical innature? If so, can these cycles be used

reliably as a basis for investment decisions?The presence of cycles in Singapore's

Strait Times Index (STI) is a hot topicamong the Singapore financial community.Many commentators believe the STI followsa 14-year cycle, with alternate 7-year bull and7-year bear runs. According to the 14-yearcycle theory, the STI is now in the third yearof a 7-year bull run, which will lead to a peakfor the STI in 2008. So will the STI peak in2008?

Analysis of cycles

The most widely used method for the analy-sis of cyclical patterns is spectrum analysis.A classic example of applying spectrumanalysis to cycle theory is the study ofsunspot activity, in which a cycle period ofaround 11 years is found. Assuming sunspotactivity reaches its peak this year, one mayexpect sunspot activity to peak 11 years later.

Equally, analysts frequently apply spectrumanalysis to measure cycles in stock markets.The analysis reveals the periodicity as well asthe magnitudes of the cycle(s). Figures 1 and2 illustrate the cyclical patterns in theSingapore and US stock markets.

As shown in Figure 1, the STI possessesmany different cycles, including a 143-weekcycle (2.7 years) and a 350-week cycle (6.7years), while Figure 2 shows that the USstock market comprises cycles of 107 weeks(2.06 years) and 195 weeks (3.75 years).

The 6.7 year cycle for the STI is of partic-ular interest. If this holds true, and assum-ing the STI was at the bottom of the cycle in2001, the next time the STI will be at thebottom of the cycle will be in 2008(2001+6.7). This prediction is the exactopposite to that derived from the 14-yearcycle theory, which states that the STI is inthe third year of a bull run and will peak in2008.

Lessons from the US stock market

But do the cycles identified by spectrumanalysis remain constant over time?Unfortunately, the Singapore index yieldsinsufficient data points to test this. Datafrom the US stock market, however, can beused as a proxy.

We divided 40 years of US stock marketdata (from 1964 to 2003) into two equal sub-periods: the first sub-period from 1964 to1983 and the second sub-period from 1984to 2003. Spectrum analysis was then applied

to the two sub-periods, the results of whichcan be found in Figures 3 and 4.

Figures 3 and 4 show that the cyclical pat-terns in the US stock market have changedsignificantly. The cycle periods are around200 weeks (3.8 years) and 55 weeks (about 1-year) in the first sub-period, and around 115weeks (2.2 years) and 70 weeks (1.3 years) forthe second sub-period. If investors had usedthe data at the end of first sub-period todetermine their investments, their returnswould not have matched their expectations.

Wing-Keung Wong (Associate Professor)and Chen Dujuan (PostgraduateStudent), Department of Economics,National University of Singapore

Note on cycles:

The prices of many commodities such as wheatand corn reflect seasonal cycles because of theiragricultural nature. In these cases, cycles are easilyexplained and understood. For financial securities,cycles are more difficult to identify and, once iden-tified, difficult to explain. Theories to explain suchcycles range widely in their scope and include ananalysis of sunspots, planetary movements andhuman psychology, which are not within the scopeof this study.

BEWARE OF THE CYCLE THEORYBy Wing-Keung Wong and Chen Dujuan

Figure 1. Spectrum for the Singapore Strait Times Indexfrom 1973 to 2004

Figure 2. Spectrum for the US Standard & Poor's Index from1964 to 2003


Subject Matters

Figure 3. Spectrum for the US Stock Market from 1964 to 1984 Figure 4. Spectrum for the US Stock Market from 1985 to 2003

1.

2.

3.

4.

5.

6.

The cycle may be periodic, but the period does not repeat precisely. For example, a significant body of evidence suggeststhat short- to medium-term cycles (less than or equal to four years) may vary in periodicity by about 10%. Long-termcycles (more than four years) typically vary by 15% to 18%.

Cycles are often not symmetrical in shape. For example, cycle bottoms may be periodic, but the cycle tops may not be andvice versa.

One cycle may be observed within other more powerful longer-term cycles, which will result in different interpretationsbased on the individual's analysis about when tops or bottoms occur.

Cycles can be distorted by market manipulations and irrational human logic.

One cycle can be out of phase while others are not, leading to greater complexity.

There may be many unobservable core cycles and the cycles that appear in the stock market may result from the combina-tion of these core cycles. Therefore, even though the core cycles may remain unchanged, the cycles that appear in the stockmarket may change dramatically. (Tai and Wong 2003)

WARNING - Using cycle theory for investment decisions …

Although there is a great deal of evidence for the presence of cyclical patterns in stock markets, some characteristics of cycles are worthpaying attention to if cycle theory is to be used as a basis for investment decisions.

Subject Matters


Central banks frequently intervene in for-eign exchange markets to reduce volatili-

ty or to correct misalignments. Such opera-tions may be successful if they drive awaydestabilizing speculators. However, specula-tors do not simply vanish but may reappearon other foreign exchange markets. Using amodel in which traders are able to switchbetween foreign exchange markets, wedemonstrate that while a central bank indeedhas several means at hand to stabilize a spe-cific market, the affect on the other marketsdepends on how the interventions are imple-mented.

Interventions are motivated by the desireto check short-run trends or to correct long-term deviations from fundamental values(Neely 2001). Although central banks seemto believe in the power of intervention oper-ations, both the theoretical and the empiricalliterature remains sceptical about its useful-ness.

One noteworthy exception is Hung (1997)who argues that central bank interventionsmay be successful in the presence of trend-extrapolating chartists. First, a central bankmay try to destroy technical trading signalsby breaking a price trend. Second, a centralbank may stimulate positive feedback tradingby inducing a price trend in order to guidethe exchange rate closer towards its funda-mental value.

While chartists display bandwagon behav-iour, fundamentalists expect the exchangerate to converge towards its fundamentalvalue. However, the fundamentalists onlyperceive the fundamental value on average.When the exchange rate is equal to its funda-mental value, half the fundamentalists viewthe exchange rate as undervalued and half asovervalued. Consequently, the net impact offundamentalists is zero. But as the distortionin the market becomes larger, the influenceof fundamentalists increases. Due to thisnonlinear weighting scheme, the model gen-erates interesting dynamics. Overall, thischartist-fundamentalist approach has provento be quite successful in replicating the styl-ized facts of financial markets.

The model

We develop a model in the spirit of thischartist-fundamentalist approach that allowsus to investigate the effectiveness of centralbank interventions. The model makes thefollowing key assumptions:

Our model examines the consequences ofinterventions by a single central bank withina system of linked foreign exchange markets.In particular, we study the effectiveness ofthe two most common intervention strategies(Neely 2001). First, the "leaning against thewind" rule (LAW), which aims at reducingpositive feedback pressure. For instance, ifthe price of a currency goes up, the centralbank takes a short position. Second, the"targeting long-run fundamentals" rule

(TARGET), which means that the centralbank always trades in the direction of thefundamental exchange rate. If the exchangerate is below (above) its fundamental value,the central bank submits buying (selling)orders.

The two strategies, LAW and TARGET,are in turn assessed under two different con-ditions, according to whether the interven-tions have been used to drive the rate closerto the fundamental value (unbiased interven-tions) or away from it (biased interventions).Thus, a total of four types of interventionare considered in the model, the results ofwhich are summarised in Table 1.

Unbiased interventions

LAW interventions which are unbiased (i.e.are not contrary to fundamental values) sta-bilize all markets. The reason is that this ruledestroys or at least weakens the trading sig-nals of chartists. Since the intervention mar-ket is less distorted, it draws in chartists fromthe other markets so that these markets alsobenefit from the intervention operations.

TARGET interventions also calm down allthe markets because they effectively work likean increase in the power of fundamentalists.If more demand is based on mean reversion,exchange rates are indeed driven closertowards fundamentals. As in the previouscase, more chartists enter the interventionmarket so that all markets profit from thispolicy.

Biased interventions

Central banks sometimes attempt to shift theexchange rate away from fundamentals inorder to boost the domestic economy.According to the model, both LAW andTARGET interventions successfully result inmoving the exchange rate away from the fun-damentals. Moreover, exchange rate fluctua-tions are completely eliminated in the inter-vention market. Since the intervention mar-ket is now very unattractive for the chartists,

CENTRAL BANK INTERVENTIONS, CHARTISTS &THE FX MARKETS by Frank Westerhoff and Cristian Wieland

Fundamental analysis is time-consumingand requires intensive research.Fundamentalists are therefore regarded asexperts who specialize in one market andthus remain in that market.

Since chartists use rather flexible extrap-olative trading rules, they may easily wan-der between markets.

Fundamentalists bet on mean reversion,whereas the philosophy of technicalanalysis is to ride on a bubble (Murphy1999).

To limit the risk of being caught in abursting bubble, chartists prefer marketswhich are not too distorted. To be pre-cise, chartists trade forcibly in those mar-kets which display price trends but whichare not too misaligned.

The behaviour of fundamentalists tendsto stabilize markets whereas the activity ofchartists is typically destabilizing. If amarket attracts an increasing number ofchartists, the exchange rate is likely to bedriven away from fundamental values (andvice-versa).

Subject Matters


they wander to the other markets. Volatilitytherefore increases in these related markets.

Conclusions

We provide the first multi-foreign exchangemarket framework based on a chartist-funda-mentalist approach to evaluate central bankoperations. Simulation analysis reveals thatcentral bank interventions may succeed incalming down markets. However, the marketin which intervention takes place eitherattracts more or drives away some destabiliz-ing chartists, depending on the nature of theintervention, which may lead to the destabili-sation of other markets.

References:

Hung, J. (1997) Intervention strategies andexchange rate volatility: A noise trading perspective,Journal of International Money and Finance, 16,779-793.

Murphy, J. (1999) Technical Analysis of FinancialMarkets, New York Institute of Finance, NewYork.

Neely, C. (2001) The practice of central bank inter-vention: Looking under the hood, Federal ReserveBank of St. Louis Review, 83, 1-10.

Frank Westerhoff, Department ofEconomics, University of Osnabrueckand Cristian Wieland, RAAD Consult,Munster.

© 2004, Frank Westerhoff and Cristian Wieland

Further details of this study will be available in thepaper "Spill-over dynamics of central bank inter-ventions", Frank Westerhoff and Cristian Wieland,due for publication in the German EconomicReview at the end of 2004.

LAW

TARGET

LAW

TARGET

Volatility in Intervention

Market

Distortion in Intervention

Market

Volatility in Other Markets

Distortion in Other Markets

Unbiased

Intervention

Biased

Intervention

Table 1. Summary of simulation results

Subject Matters


Are the returns of the major South-EastAsian stock markets forecastable? If so,

can those returns be forecast by models thatrely entirely on one variable - the stock priceitself ?

To seek answers to the above questions, weresorted to time-series modelling, a method-ology which is rooted in the same principlesas technical analysis. A time-series modeldemands nothing more than the historicalrecords of the variable under investigation,whereby the movements of the variable areexplained solely in terms of its own past.

Parallels have even been drawn betweenthe recent trend in non-linear time-seriesmodelling (where the output from a model isnot proportional to the sum of its input vari-ables) and technical analysis. Clyde andOsler (1997) argued that technical analysiscould be viewed as a simple way of exploring

the non-linear behaviour of financial time-series. For example, patterns such as head-and-shoulders are clearly attempting to findsome kind of non-linearity in the series.

In our study we looked at daily stock mar-ket indices from the five major South-EastAsian countries (ASEAN-5: Indonesia,Malaysia, Philippines, Singapore andThailand) from January 1990 to October2001. From this data, we computed the per-centage daily returns (based on the pricemove from the close of one trading day tothe next). Figure 1 provides an example ofthe resulting time-series; a plot of the dailyreturns from Singapore's Strait Times Index.

The data was divided into two periods.Data from January 1990 to October 2000was used to create six time-series models(two linear and four non-linear) and a ran-dom walk model. The seven models were

then used to generate 1-day, 1-week, 1-month, 3-month, 6-month, 9-month and 1-year forecasts.

The forecasts from these models were thencompared with the actual data fromNovember 2000 to October 2001 and theirperformance was measured using the rootmean squared error (RMSE) method.

Forecasting Performance

The forecasting performances of the sevenmodels are summarized in Table 1 (note thatmodels with better performance have smalleraverage values). In addition, the averageranking of the models (based on RMSE foreach forecast horizon) is given in Table 2.

On average, linear models are superior tonon-linear models for forecast horizons of

SOUTH-EAST ASIAN STOCK MARKETS FOLLOW A NON-RANDOM WALKby Venus Khim-Sen Liew, Kian-Ping Lim and Chee-Keong Choong

-10

-5

0

5

10

15

20

500 1000 1500 2000 2500 3000

Figure 1. Daily returns of the Strait Times Index

Subject Matters


1-month and longer, whereas non-linearmodels are superior for 1-day forecasts. For aforecast horizon of 1-week, linear models areat most comparable with non-linear models.Thus, although there is evidence of non-lin-earity on stock returns (Tse, 2001), informa-tion on non-linearity seems to produce littlegain in the prediction of stock returns.

More significantly, Table 1 shows that theRMSE values of the random walk modelsare substantially greater than all the time-series models considered, for nearly all theforecasting horizons and across all five coun-tries. In other words, the random walk modelranked last in all cases, with the only excep-tion being the 1-week horizon for the StraitTimes Index (in which the random walkmanaged to rank second out of the sevenmodels under study). This suggests that thereturns of the ASEAN-5 stock markets donot follow a random walk and are fore-castable by time-series models, thus provid-ing further justification for the work of tech-nical analysts.

Venus Khim-Sen Liew, Faculty ofEconomics and Management, UniversitiPutra Malaysia.

Kian-Ping Lim, Labuan School ofInternational Business and Finance,Universiti Malaysia Sabah.

Chee-Keong Choong, Faculty ofAccountancy and Management,Universiti Tunku Abdul Rahman.

Forecast Horizon Random Walk Linear Models Non-Linear Models

RMSE

Jakarta Composite Index

1 Year 1.854 0.001 - 0.006 0.009 – 0.0299 Months 1.826 1.342 1.344 - 1.3496 Months 1.779 1.061 - 1.062 1.064 - 1.0693 Months 2.079 1.638 - 1.639 1.635 - 1.6371 Month 1.431 1.430 - 1.432 1.433 - 1.4341 Week 1.264 1.403 1.4031 Day 1.480 1.414 1.413

Kuala Lumpur Composite Index

1 Year 1.789 1.296 - 1.299 1.295 - 1.2989 Months 1.706 1.343 - 1.346 1.341 - 1.3456 Months 1.793 1.263 - 1.288 1.260 - 1.2703 Months 1.767 0.819 - 0.825 0.815 - 0.8231 Month 1.389 0.882 - 0.897 0.886 - 0.9151 Week 1.256 1.222 - 1.233 1.211 - 1.2301 Day 1.932 0.006 - 0.044 0.022 - 0.036

Philippines Composite Price


Strait Times Index


Stock Exchange of Thai

1 Year 2.135 1.677 1.677 - 1.6799 Months 2.115 1.725 1.725 - 1.7276 Months 1.981 1.559 - 1.565 1.560 - 1.5693 Months 2.361 1.799 1.795 - 1.8021 Month 2.421 2.161 - 2.184 2.167 – 2.1971 Week 2.105 1.726 - 1.767 1.718 - 1.7891 Day 2.297 0.036 - 0.089 0.013 - 0.155

Table 1. Forecasting Performance by the RMSE


ASEAN-5Average Random Walk Linear Models Non-Linear Models

RMSE

Table 2. Ranking of Forecasting Models by Forecast Horizon

Subject Matters

RESEARCH ROUNDUP

Local traders are less likely to recognise lossesthan non-local traders.

This is the finding from a behavioral study conducted by Alex Frinoand Hui Zheng of the University of Sydney and David Johnstone ofthe University of Wollongong. Consistent with other studies, theyfind evidence of a disposition effect (the propensity of traders to ridelosses yet realise gains) for both on-floor professional futures traders(locals) and a matched sample of non-local traders. After controllingfor potential differences in trader characteristics, comparisons reveal astronger disposition effect among locals than non-local traders. Theauthors hypothesise that because locals must trade profitably to sur-vive, it is improbable that they are more irrational in their loss ridingthan non-locals. To the contrary, they find evidence to show thatpaper losses for local traders are more likely to be turned into papergains by the time of the next transaction, when compared to non-local traders. This result is consistent with the hypothesis that locals,by their presence on the trading floor, have privileged albeit short-lived information on order flow that allows them to form relativelyaccurate probability predictions of the direction and strength ofshort-term market price shifts.

The propensity for local traders in futures markets to ride losses:Evidence of irrational or rational behavior? Alex Frino, David Johnstoneand Hui Zheng, Journal of Banking & Finance, Volume 28, Issue 2,February 2004, Pages 353-372

NYMEX dominates the IPE in setting the priceof crude oil,

according to Lin and Tamvakis of CASS Business School in London.The authors look at the interaction of the two main price setting ener-gy markets (London's International Petroleum Exchange (IPE) andNew York's Mercantile Exchange (NYMEX)) when both of them areopen (synchronous trading) and when only London is open (asynchro-nous trading). Specifically, they test the hypothesis that London isaffected by New York by analysing the transaction duration of the IPEBrent futures contract, both when the NYMEX WTI futures contractis being traded and when NYMEX is closed. Using tick-by-tick dataobtained from IPE, transaction durations are found to form two dis-tinctive and inverted U-shaped patterns. A statistical model is appliedto the data, which shows that the parameters of the IPE in the morn-ing and the afternoon are significantly different from each other,underlining the dominant effects of NYMEX on IPE trading. Theresults from the current analysis reinforce previous results by theauthors, which indicate that NYMEX is a leading price setter in crudeoil futures prices and has a dominant effect on the IPE-traded con-tracts.

Effects of NYMEX trading on IPE Brent Crude futures markets: a dura-tion analysis. Sharon Xiaowen Lin and Michael N. Tamvakis, EnergyPolicy, Volume 32, Issue 1, January 2004, Pages 77-82

Abstracts reprinted with permission from Elsevier.

• elbitapmoC esabataD KCOTSATEM )deriuqer noisrevnoc on(

• no nuR ISC ,erofneT ,GREBMOOLB deeF tenretnI nwo ruo ro

• yaD fo dnE elbaliava segnahcxE 03 revO

• a kciT ,seirotsiH ecirP gnoL elbaliava ataD lavretnI dn

• htnom rep 72£ sa wol sa morf trats secirP

– morf lairT eerF ruoY daolnwoD ku.oc.atadpu.www

sisylanA lacinhceT decnavdArof stsylanA lacinhceT decnavdA

ve si tsylanA lacinhceT atadpU morf tcepxe dluow uoy gnihtyreA lacinhceT decnavda ylurt a a ot noitidda nI .metsys sisylan

ni erawtfos eht sloot fo tsoh yrateirporP AIXEDNI sedulc fo yarra tsav a gninnacs rof rethgilhgiH lufrewop a ,srotacidnI

sihpos tsom eht dna airetirc otacidnI detacit noitasimitpO r dlrow uoy sevig osla tsylanA lacinhceT .dnif lliw uoy seludom

tuB .strahC erugiF & tnioP ssalc ees ,ti rof drow ruo ekat t’nod .yadot ypoC noitaulavE lluF ruoY daolnwoD ,flesruoy rof

Subject Matters

Key price levels play a strong role in determiningstock market activity.

This is the conclusion from the Helsinki School of Economics'Markku Kaustia, whose study looks at the reluctance of investors torealise losses (disposition). To do this, the author examines IPO trad-ing volume, in which case all initial investors have a common purchaseprice and the disposition effect should be easier to detect. His findingsshow that turnover is significantly lower for negative initial return IPOswhen the stock trades below the offer price, and increases significantlyon the day the price surpasses the offer price for the first time. Theincrease in volume lasts for two weeks. On a daily level, attaining newmaximum and minimum stock prices also produces a strong increase involume.

Market-wide impact of the disposition effect: evidence from IPO tradingvolume. Markku Kaustia, Journal of Financial Markets, Volume 7, Issue2, February 2004, Pages 207-235.

Finnish investors realise losses more than gains towards the end of the tax year (31 December),according to Finnish researchers Mark Grinblatt of The Anderson School at UCLA and NBER and Matti Keloharju of the Helsinki School ofEconomics. Moreover, the authors continue, Finnish investors repurchase the same stocks recently sold. The repurchase rate depends on themagnitude of loss, firm size and how late in the year the sale takes place.

Tax-loss trading and wash sales. Mark Grinblatt and Matti Keloharju, Journal of Financial Economics, Volume 71, Issue 1, January 2004, Pages 51-76

Moving average trading rules in the currencymarkets are unprofitable,

according to Dennis Olson of the American University of Sharjah,United Arab Emirates. In his paper, the author analyses 18 exchangerate series from 1971 to 2000. The moving average trading rules wereoptimised for successive five-year sample periods from 1971 to 1995and tested over the subsequent out-of-sample period 1995 to 2000.His results show that risk-adjusted moving average trading rule profitshave declined from an average of over 3% in the late 1970s and early1980s to about zero in the 1990s. The author hypothesises that earlierprofits resulted from inefficiencies that have since been eliminated.

Have trading rule profits in the currency markets declined over time?Dennis Olson, Journal of Banking & Finance, Volume 28, Issue 1,January 2004, Pages 85-105

Book review


THE INVESTOR'S GUIDE TO TECHNICAL ANALYSIS

By Curt RenzPublished by McGraw Hill 148 pages, £12.99ISBN 0-07-138998-9

Numerous books now exist providing introductions totechnical analysis and charting, many of which

claim to give the private investor an indispensable guideto making profits on the stock market. Well establishedtext books such as those by John Murphy and CorneliusLuca may be targeted at the more experienced traderand would-be professional analyst, but should not bebeyond the scope of the novice investor. Nevertheless,Curt Renz's book, while covering little new ground, pro-vides a useful guide to individuals who have neither thetime nor inclination to delve more deeply into the sub-ject.

While there is nothing new to say on the subject of intro-ductory level technical analysis, starter-level books onthe subject continue to emerge as the profile of techni-cal analysis rises. This may be due to the failure of fun-damental analysis to adequately explain and anticipatethe stock market crash of 2001. Furthermore, ambigu-ous fundamental news continues to make profit-makingopportunities on the stock markets hard to identify.

The core of Renz's book looks at bar-chart analysis inthe context of various US stock market indices.Reiterating that "the trend is your friend" and presentingan overview of basic chart patterns such as flags, pen-nants and head-and-shoulders will not inspire experi-enced investors, but then the book doesn't claim to bemore than an introduction for those new to the subject.Everything the casual investor needs to know aboutcharting is included and no space is wasted on morecomplicated patterns, indicators and oscillators.

The author claims the basic chart formations included inhis book should cover around 95% of situations the pri-vate investor is likely to encounter. The remaining 5%of situations may offer significant profit opportunities,but their exploitation demands the time and experienceonly professional traders or analysts possess. He thengoes on to discuss the use of 10, 20 and 200-day mov-ing averages and cross-over methods as tools of confir-mation. He concludes with a set of thirteen problemswhere the reader can test his or her interpretation of dif-ferent patterns and check his answers against the solu-tions provided.

Carl Renz's book is short, well laid-out and written in anupbeat North American investor style. It is also reason-ably priced for the retail market and should prove anattractive and user-friendly introduction to technicalanalysis for the non-professional.



16 December 2003 – 24 February 2004Non-commercial net long positions and spot rates

COMMITMENTS OF TRADERS REPORT

10-year US Treasury

Source: CBOT

5-year US Treasury

Source: CBOT

-100,000

-80,000

-60,000

-40,000

-20,000

0

20,000

40,000

60,000

Dec-16 Dec-22 Dec-30 Jan-06 Jan-13 Jan-20 Jan-27 Feb-03 Feb-10 Feb-17 Feb-24

3.90

3.95

4.00

4.05

4.10

4.15

4.20

4.25

4.30

4.35

10-yr Treasury Spot

0

50,000

100,000

150,000

200,000

250,000

300,000


2.80

2.85

2.90

2.95

3.00

3.05

3.10

3.15

3.20

3.25

3.30

5-yr Treasury Spot

Dow Jones Industrial Average

Source: CBOT

Swiss franc

Source: CME

-6,000

-5,000

-4,000

-3,000

-2,000

-1,000

0

1,000


9800

9900

10000

10100

10200

10300

10400

10500

10600

10700

10800

DJIA Spot

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000


1.2

1.21

1.22

1.23

1.24

1.25

1.26

1.27

Swiss franc Spot



Euro

Source: CME

0

5,000

10,000

15,000

20,000

25,000

30,000

35,000

40,000


1.2

1.21

1.22

1.23

1.24

1.25

1.26

1.27

1.28

1.29

Euro Spot

3-month eurodollar

Source: CME

0

50,000

100,000

150,000

200,000

250,000

300,000

350,000

400,000

450,000


0.98

1.00

1.02

1.04

1.06

1.08

1.10

1.12

1.14

3-month eurodollar Spot

Pound sterling

Source: CME

0

5,000

10,000

15,000

20,000

25,000


1.65

1.70

1.75

1.80

1.85

1.90

1.95

Pound sterling Spot

Yen

Source: CME

-20,000

-10,000

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000


103.5

104

104.5

105

105.5

106

106.5

107

107.5

108

108.5

Japanese yen Spot



Nasdaq

Source: CME

-14,000

-12,000

-10,000

-8,000

-6,000

-4,000

-2,000

0

2,000

4,000

6,000


1800

1850

1900

1950

2000

2050

2100

2150

2200

Nasdaq Spot

Nikkei

Source: CME

-3,500

-3,000

-2,500

-2,000

-1,500

-1,000

-500

0

500

1,000

1,500

2,000


9800

10000

10200

10400

10600

10800

11000

11200

Nikkei Spot

Gold

Source: CEI

0

20,000

40,000

60,000

80,000

100,000

120,000

140,000


385

390

395

400

405

410

415

420

425

430

Gold Spot

US dollar index

Source: NYBOT

-12000

-10000

-8000

-6000

-4000

-2000

0


110

110.5

111

111.5

112

112.5

113

113.5

114

114.5

115

US dollar index Spot


Non-commercial

Commercial

Charts and tables: Open interest (futures only)All data provided by the Commodity Futures Trading Commission (CFTC) with permission

166,521 157,614 122,121 69,476

-123,236

-155,529

-172,435

-200,139

8,146

9,790

11,544

13,258

-13,696

-17,364

-20,434

-10,109

-34,669

-36,627

-33,868

-35,363

-85,882

-75,786

-75,498

4,417

-36,721

-45,368

-48,060

-39,744

-85,788

42,544

-31,505

-69,723

2,159

3,967

5,719

6,814

-419

-262

110

1,124

-109,445

-106,391

-119,597

-104,499

8,435

8,591

12,336

7,318

Feb 3 Feb 10 Feb 17 Feb 24

10yr Treasury -49,823 -53,187 -12,128 52,831

5yr Treasury

176,465

211,342

230,387

250,599

DJIA

-1,385

-1,837

-2,689

-4,848

Swiss franc

9443

11,003

11,479

5,374

Pound Sterling

20,893

19,906

20,159

21,076

Japanese yen

64,499

58,630

55,198

-12,181

Euro

20,446

26,242

28,409

26,367

3m eurodollar

174,972

45,682

147,772

194,247

Nasdaq

2,663

-855

-3,011

-11,892

Nikkei

Nikkei

-420

-617

-1,604

-2,884

Gold

67,564

60,286

70,305

63,058

US$ index

Feb 3 Feb 10 Feb 17 Feb 24

10yr Treasury

5yr Treasury

DJIA

Swiss franc

Pound Sterling

Japanese yen

Euro

3m eurodollar

Nasdaq

Gold

US$ index

-7,259

-6,641

-9,617

-6,137



All venues are in London

For training and events diary submissions please email us at: [email protected]

6 AprilCourse:

STA revision day Organiser:

Society of Technical AnalystsContact:

[email protected]

22 AprilCourse:

Alpesh Patel Live! secrets of an active trader Organiser:

The Information ExchangeContact:

[email protected]

Course:

Introduction to technical analysisOrganiser:

Quorum TrainingContact:

[email protected]

24 March

Course:



[email protected]

19 July

Course:



[email protected]

15 November

Course:


7CityContact:

[email protected]

4 June

Course:


7CityContact:

[email protected]

23 August

Course:


7CityContact:

[email protected]

10 November

23/24 MarchCourse:

Technical analysis & chartingOrganiser:

ChartWatchContact:Market Directional Analysis Tel: 020 7723 0684

22/23 JuneCourse:

An introduction to charting andtechnical analysis Organiser:

IPEContact:

[email protected]

23 JuneCourse:Advanced technical analysis

Organiser:

The Oxford Princeton Programme Contact:[email protected]

23/24 MarchCourse:

Advanced technical analysisOrganiser:

ChartWatchContact:

Market Directional AnalysisTel: 020 7723 0684

Late OctoberCourse:

Technical analysis & chartingOrganiser:

ChartWatchContact:

Market Directional AnalysisTel: 020 7723 0684

Late OctoberCourse:

Advanced t echnical analysis Organiser:

ChartWatchContact:Market Directional Analysis

Tel: 020 7723 0684

17 AprilCourse:

One day technical analysisOrganiser:

The Technical Analysis Workshop Co.Contact:

[email protected]

TRAINING AND EVENTS DIARY

genetically engineered trading - the technical analyst · genetically engineered trading gann...

Documents