Contrary Opinion Looking at Sentiment Indicators

jan/feb

2 0

0 5

The publication for trading and investment professionals www.technicalanalyst.co.uk

Contrary Opinion Looking at sentiment indicators

Outlook for USD/NOK Capturing market extremes Software feature Several indices Applying Bollinger Bands The Bank of New York’s come together to the RSI portfolio flows monitor THE 11th EUROMONEY BOND INVESTORS CONGRESS

QUEEN ELIZABETH II CONFERENCE CENTRE, LONDON. 22–23 FEBRUARY 2005 Reserve your place at the world’s largest fixed income investor event

"The Euromoney Bond Investors Congress is unequalled in the breadth and quality of participants and is nowadays firmly established in the calendar. A continuing strength of the Congress is topical areas of discussion from which the audience can gain genuinely fresh insights and knowledge." Christopher Woods, Chief Investment O fficer for Hedge Fund Strategies, State Street Global Advisors

Now in its 11th year, the Euromoney Bond Investors Congress is the world’s premier fixed income event. It brings together more than 1500 market participants - over 650 institutional investors - to discuss and debate the issues of the day and the trends for the forthcoming year. Panels of experts from all sides of the fixed income universe argue and discuss the market’s most pressing questions. Leading economists, policy makers and visionaries will give keynote speeches. Conference speakers will include Avinash Persaud, Dr Adam Posen and Stephen Roach. The plenary sessions are complemented by a series of technical, focused workshops hosted by the conference sponsors. Topics at the Congress will include: Calling the bond market turn: why everyone got it wrong in 2004 and why they won't get it wrong in 2005 Credit spreads: is the only way up? Hedge funds: How they expect to make money in fixed income in 2005 Structured products Adding value: how forex can beef up your fixed income portfolio Emerging markets: is the party finally over?

Sponsors to date: ABN AMRO - BankTuranAlem - Barclays Glo bal Investors - Bond Exchange of South Africa - BondVision EuroMTS Indices - Dresdner Kleinwo rt Wasserstein - Fitch Ratings - MDM Financial Group - Morgan Stanley - Nykredit Realkredit A/S - Reuters - Trust Investment Bank - Westpac Institutional Bank Exhibitors to date: Cantor Market Data - CFA Institute - Chicago Board of Trade - eSpeed Inc - Eurex - GovPX inc. - International Securities M arket Association - Moody’s Investor s Service - Standard & Poor’s - SWX Swiss Exchange - Thomson Tradeweb To register for the Congress please visit www.bondcongress.com tel +44 (0) 845 130 7754 or fax +44 (0) 845 130 7753 WELCOME

Repeat patterns in historical stock prices inevitably lead to hopes that a potentially profitable market cycle exists. We look at the theory that years ending in ‘5’ signify a positive return to US stocks. Can forecasting Dow direction this year really be that easy? In the world of FX, we ask if the Norwegian Krone will signal an important turn in dollar direction. We also take a look at contrary opinion, a well established area of technical analysis, and offer an overview of the key sentiment indicators that gauge how the majority are positioned in the markets.

The Technical Analyst wishes you a happy and prosperous 2005.

Matthew Clements, Editor

CONTENTS 1 > FEATURES JAN/FEB

USD/NOK Is the tide turning? >06 Anders Soderberg, chief technical analyst at SEB Merchant Banking, reveals his reasons for predicting a USD correction

Building a better RSI using Bollinger Bands >12 Alex Douglas explains how Bollinger Bands can be used to calibrate RSI with prevailing market conditions

Sentiment indicators Sentiment indicators are essential tools for >20 the contrarian trader. The Technical Analyst presents an overview of the best

© 2005 Clements Biss Economic Publications Limited. All rights reserved. Neither this publication nor any part of it may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of Clements Biss Economic Publications Limited. While the publisher believes that all information contained in this publication was correct at the time of going to press, they cannot accept liability for any errors or omissions that may appear or loss suffered directly or indirectly by any reader as a result of any advertisement, editorial, photographs or other material published in The Technical Analyst. No statement in this publication is to be considered as a recommendation or solicitation to buy or sell securities or to provide investment, tax or legal advice. Readers > > should be aware that this publication is not intended to replace the need to obtain professional advice in relation to any topic discussed.

January/February 2005 THE TECHNICAL ANALYST 1 GET QUALIFIED IN TECHNICAL ANALYSIS

The Society of Technical Analysts (STA) represents and accredits professional and private Technical Analysts operating in the UK

Originally established in the 1960s, the STA provides its members: • Education Monthly lectures and regular teaching courses in technical analysis • Research The STA Journal publishes research papers on TA techniques and approaches • Meetings Provide members the opportunity to discuss technical approaches and markets • Representation The STA lobbies on behalf of analysts with data vendors, exchanges and regulators. The STA represents the UK at the International Federation of Technical Analysts (IFTA) • Accreditation The STA Diploma Exam is internationally recognised as a professional level qualification in Technical Analysis

The next Diploma examination day is 22 April 2005 at the London School of Economics

For more information on how to join and what is involved in passing the STA Diploma exam, visit our website at: www.sta-uk.org or call us on +44 7000 710207

I/We would like to subscribe to The Technical Analyst magazine for: £140 1 year subscription - UK £165 1 year subscription - Rest of World Please tick as appropriate Name: Title: Institution: Address:

City/State: Country: Postcode: To subscribe to The Technical Analyst magazine post a copy of this form to: Phone:

The Technical Analyst Email: Unit 201 I attach a cheque payable to ‘The Technical Analys Panther House 38 Mount Pleasant Please invoice me/my company London WC1X 0AN Alternatively, email [email protected], providing the same information as requested above. Or email [email protected], +44 (0)20 7833 1441 providing the information requested in this form www.technicalanalyst.co.uk SUBSCRIBE 32 42

CONTENTS 2 > REGULARS

INDUSTRY NEWS 04 Editor: Matthew Clements Managing Editor: Jim Biss Marketing: Vanessa Green Advertising Sales: Chris Leigh MARKET VIEWS Design: Paul Simpson USD/NOK: Is the tide turning? 06 Brent crude in 2005 08 The Technical Analyst is published by Clements Biss Economic Publications Ltd UK gilts: An uncertain year ahead 10 Unit 201, Panther House, 38 Mount Pleasant, London WC1X 0AN TECHNIQUES Tel: +44 (0)20 7833 1441 Building a better RSI using Bollinger Bands 12 Web: www.technicalanalyst.co.uk Do years ending in '5' really signal a bull run for stocks? 18 Email: [email protected] Sentiment indicators 20 Five Fibonacci studies 25

SUBSCRIPTIONS THE TECHNICAL ANALYST TALKS TO… 30 Adam Sorab, chairman, Society of Technical Analysts Subscription rates (6 issues) UK: £140 per annum Rest of world: £165 per annum For information, please contact: SUBJECT MATTERS [email protected] Gold and the Kondratiev Cycle 32 ADVERTISING Comparing moving average envelopes & Bollinger Bands 36

For information, please contact: [email protected] SOFTWARE 39 PRODUCTION The Bank of New York's Interactive Portfolio Flow Monitor

Art, design and typesetting by all-Perception Ltd. BOOK REVIEW & LETTERS 41 Printed by The Friary Press The (Mis)behaviour of Markets ISSN(1742-8718) by Benoit Mandelbrot & Richard Hudson

COMMITMENTS OF TRADERS REPORT 44 LONG-TERM TECHNICALS 46 TRAINING AND EVENTS DIARY 48

January/February 2005 THE TECHNICAL ANALYST 3 Industry News

Reuters to buy Telerate DEMARK SIGNALS LONG-TERM

Reuters's is to acquire fellow Savvis Communications SELL FOR DOW information provider Corporation. Telerate has Moneyline Telerate for $175 around 30,000 users worldwide In a New Year million. The agreement is and full integration into research note, made up of $100 million in Reuters is expected to take 18 Tom DeMark, cash and $75 million in months. creator of the Reuters-owned shares in the DeMark indicators, said that his annual Sequential ESIGNAL LINKS UP WITH US BROKER Indicator is signalling a long- term 'countdown' sell signal for the Dow. This is the first Charting provider, eSignal, has London Stock Exchange (LSE) such signal since 1972 after linked up with broker Spear, Level 2 data, Euronext which the Dow fell nearly 50% Leeds & Kellogg's (SLK) and Equities Level 2 data, OFEX over the following two years. their REDIPlus trading plat- data and Bern Stock Exchange DeMark's indicators are form. eSignal will provide data. In addition, eSignal has available on Bloomberg and charting, formula studies and completed an agreement with CQG. back testing features to AFX News Limited - a REDIPlus. SLK deals in US European provider of interna- equities, options, electronic tional economic news - that SINGAPORE futures and European equities. will expand eSignal's set of BONDS TRADE Meanwhile, As part of its current AFX offerings to plan to broaden its business in include services such as ON BLOOMBERG Europe, eSignal’s latest release German language news. (version 7.8) now includes The Monetary Authority of Singapore has chosen Bloomberg to provide the inter-dealer trading platform for the Singapore government bond. Starting in 2005, all pri- mary dealers in Singapore will use the Bloomberg platform to transact bonds electronical- ly and will be able to view complete trading information for the market which has a daily volume of around SGD 2.6 billion.

4 THE TECHNICAL ANALYST January/February 2005 Industry News

Telerate updates charting features Telerate has announced enhancements to the technical analysis functions available on its series of market data services. To its A8 1.2 system has been introduced new studies including the accumulation/distribution oscillator, a directional oscillator, divergence, RSI and Keltner channels. Version 2.0 has new instrument displays which now include the Gann Swing. 'Advanced Active8 Charting' also features new Random Walk indices, a double smoothed moving average and Fibonacci projection lines. Meanwhile, Telerate's new fast and slow stochastics, and include open interest data WebStation 3.0 now includes the directional movement indi- when its 3.1 version becomes Ichimoku Kinko-Hyo charts, cator (DMI). Telerate plan to available.

EXCHANGE NEWS: EUREX volume reaches record high

Eurex, the exchange for euro-denominated derivatives, traded a record 1.07 billion contracts in 2004 (up from 1.014bn in 2003) making it the world's largest derivatives market. In December volume increased 21% on December 2003 to 80.1 million contracts while open interest rose 35% to 61 million contracts. The Euro-Bund future enjoyed the greatest trading volume of all Eurex products. The most traded equity deriv- ative contract was the DJ Euro STOXX 50. CBOT trading volume hits record in 2004

The Chicago Board of Trade has also reported record trading volume for 2004. A total of almost 600 million contracts were traded on the exchange last year, a 32% increase on 2003 and an all-time high for the CBOT.

January/February 2005 THE TECHNICAL ANALYST 5 Market Views

USD/NOK IS THE TIDE TURNING? by Anders Soderberg

lthough nothing suggests that the four-year bear dollar correction. On average the market has retraced 7.1% trend is coming to an end soon, there are several under those conditions. In fact, the market does at present Asigns which indicate a longer pause or greater reac- reflect such divergence (Figure 1). tion may be imminent. USD/NOK The dollar Concerning USD/NOK, we begin by examining its weekly Both sentiment as well as technical factors signal an chart (Figure 2). Firstly we notice a significant probability of increased risk of a US dollar correction during the early part a forthcoming correction once the pair has traded outside of this year. As the old saying goes "When it's made it to the and then returned into its 13-week +/-5 percentage point front page, there's not much left to go" and the greenback price channel. Also, the multi-RSI shows a special pattern has certainly been front-page news during the last quarter often seen ahead of corrections where different RSIs flatten 2004. Significantly, reportable positions of non-commercial out below 30 in order to create a base prior to any reaction. accounts on the Chicago Mercantile Exchange appear con- A buy signal will be triggered once the shortest RSI (blue) sistent with this observation. They have decreased their long turns around, penetrating longer RSIs, and once the pair currency holdings, especially in CAD and EUR contracts trades outside its price band. Such a signal has already been against the dollar. triggered. In fact non-commercial accounts are currently carrying a Turning to Elliott waves for further guidance, we detect a short EUR position for the first time since reporting a small complete or almost completed wave pattern down from a short position in November 2001. Furthermore, addition of high point in October 2000 (Figure 3). However, even after the six largest currency contracts (CAD, CHF, GBP, JPY, taking this into account, it still remains unclear if the final EUR and AUD) to provide a broader understanding of dol- wave five has terminated yet. Ideally the final low should be lar positioning also suggests we have passed the point at found around 6.0000, although a temporary over-run cannot which sentiment was most bearish. In recent years, a diver- be ruled out at this stage. gence between a falling dollar index future and a decreasing Once a low is confirmed the market should begin to cor- short dollar position by non-commercial accounts has pro- rect its entire decline from a high of 9.6520 in 2000. Such a vided a particularly accurate early signal of a forthcoming correction should last many months and ultimately target the

Figure 1. Dollar index future and net non-commercial dollar posi- Figure 2. (USD/NOK weekly and multiple RSI) tioning on the CME (inverted scale).

6 THE TECHNICAL ANALYST January/February 2005 Market Views

“NON-COMMERCIAL ACCOUNTS ARE CURRENTLY CARRYING A SHORT EUR POSITION FOR THE FIRST TIME SINCE NOVEMBER 2001.”

Figure 3. (USD/NOK weekly wave count) Figure 4. (USD/NOK and daily scorecard indicator)

7.42/7.85 area (7.42=38.2%, 7.49=wave 4 high, 7.85=50% stochastics and rate of change. Each indicator is given a reaction). However, since the market will encounter very +/- score based upon them being overbought /oversold by strong resistance around 6.70, the former triangle floor, we our definition. The scores are then summed up to a total expect wave (a) of the correction to end there. Wave (b) will score, which can gyrate between +/- 7). retrace lower (to around 6.40) before wave (c) brings the As we have already shown, several indices already suggest pair upwards towards the correction target area to complete a forthcoming USD/NOK correction. We believe the cor- the total correction. Subsequently, the long-term trend will rection has already begun and that the pair will reach the resume with new lows likely. In addition, our daily 6.70 area during late Q1 or early Q2. This view reflects our USD/NOK scorecard*, which acts as an indicator of indi- expectation of a forthcoming general dollar correction, cators, reveals a clear and bullish divergence between price especially as we remain bearish concerning EUR/NOK. and score (Figure 4). The scorecard's refusal to move lower despite fresh price lows is a clear warning sign that the market is overdone. A scorecard reading of -6.00/-7.00 is normally followed by a reaction higher during the following Anders Soderberg is chief technical analyst at SEB month. (*Our scorecard includes indicators such as RSI, Merchant Banking in Stockholm.

January/February 2005 THE TECHNICAL ANALYST 7 Market Views

BRENT CRUDE IN 2005 by Jim Garland

sing the same indicators that per- formed well last year could provide a Uclearer picture of what lies ahead for Brent crude. Figure 1 shows prices since April 2004 along with a group of moving averages: the 13-, 21- and 34-day, all members of the Fibonacci sequence. Over this period moving average crossovers have done a good job in confirming strong trends in the market. Figure 2 shows that over the past year a 10- day RSI was an effective indicator of market tops, well in advance of the eventual high watermark. For example, the RSI exceeded 90 (significantly overbought) at the beginning of October last year, anticipating the eventual high of 51.56 on October 26th. In this case, the overbought signal did indeed foresee a retracement of prices. The RSI also proved an excellent indicator over intermediate market bottoms when the value fell below the 30 level. This proved true in early February 2004, early April, late August, mid-November and once again at the beginning of December. Presently, the indicator is in the middle of the range and is giv- Figure 1. ing no clear directional information. Figure 3 shows that the stochastic indicator, like the RSI, has also been a harbinger of market tops and bottoms. Recent market movement puts the stochastic indicator in the middle of its range but it must be noted that a higher low was put in just above the oversold level of 20, indicating the potential of the market to reverse its current down- ward trend. Figure 4 shows the most recent months of trading in IPE Brent, drawn with a 10-day moving average and standard Bollinger Bands. Whenever the price of the crude moved through its 10-day moving average, it then made a subsequent move to the opposite band where it found support or resistance. Most recently, Brent crossed below its 10-day average on December 22nd and even- Figure 2.

8 THE TECHNICAL ANALYST January/February 2005 Market Views

tually found support on the lower band just above $39 on January 4th. Prices are now making their way back towards another assault on the 10-day moving average which, if crossed, would indicate a move toward the upper band currently around $43.67. In terms of Fibonacci, the Brent crude contract retraced 62% of the total annual range when it found support at its lows in December around $37.30 (Figure 5). The market then advanced and found resistance back at the next level around $43. It is now in a sideways range inbetween the retracements levels. This gives us more reason to believe we are poised for a significant breakout. We are currently in a downtrend that has Figure 3. followed an unprecedented advance to all time high levels. Although the RSI is offering no clear buy or sell signals, the lagging indicators suggest a reversal or pending breakout. Furthermore, the higher lows being made by the stochastics, in conjunction with the con- vergence of the fast and slow moving averages, could imply the market is poised for a strong move to the upside. Looking at the fundamentals: Given the international oil market's relationship to polit- ical events, it would take a brave speculator to ignore developments in the Middle East, demand from fast growing China, rebuilding of tsunami-hit Southeast Asia and decisions by OPEC. As a result, any technical analyst studying his charts must keep a close eye on these fundamental factors and other world news. Figure 4.

Jim Garland is senior vice-president at GlobalView software Figure 5.

January/February 2005 THE TECHNICAL ANALYST 9 Market Views

UK GILTS AN UNCERTAIN YEAR AHEAD by Francis Bray

K gilts enjoyed a successful second half of 2004 be a regular 3-3-5 flat correction, but either way, the 104.86 following the strong and well-supported recovery June 2004 reaction low is likely to remain a significant turn- Uoff the June low at 104.86. The concerted break ing point. back above the March lower high at 110.46 in December From an Elliott Wave point of view on the monthly con- has spilled over into 2005, and is set to continue towards tinuation chart, the March 1998 high at 126.53 can be 112.45 (38.2% retracement of the 124.77 June 2003 peak to labelled the peak of Wave 3, with the triangle Wave 4 104.86 decline) and the key 114.55/115.00 area in the first underway (Elliott Wave theory states that a triangle is most half of 2005. likely to be found in this position as a Wave 4). The 8-year advance off the April 1990 bottom at 66.69 peaked at Long-term outlook 126.53. Despite the current triangle low point at 104.29 Concentrating on the wider continuation chart, for 10-year (October 1999) having already been set (marked "A"), the gilt futures (Figure 1), price action has been tracing out a terminal low has yet to be made ("E"), and as mentioned near 7-year equilateral/ascending triangle since peaking at earlier, it is expected to appear sometime beyond March 126.53 in March 1998 (actual traded peak of a former 2006. The "B" wave off 104.29 showed promising signs of benchmark), and is consolidating the impulsive 8-year testing the 126.53 Wave 3 peak, yet it never delivered as the advance off the April 1990 major reaction low at 66.69. 3 ½ year advance did not break down into 5 waves. The This lateral consolidation has surpassed the 5-year "C" wave decline to 104.86 was comparatively sharp and Fibonacci time target (.618 of the 8-year advance) and will incisive - common with many corrective "C" waves, and extend to at least match the 8-year advance which matures took a mere 11 months. The "D" wave is currently under- in March 2006. However, if there are two more legs of the way off 104.86 and should last the length of 2005 and triangle to be completed then a base beyond March 2006 is beyond. "D" waves share certain characteristics with first anticipated, and therefore action is expected to remain impulsive waves. One of these properties is evident here as between 104.86 and 126.53 for 2005. A possible alternative the wave is becoming gradually more constructive after an for labelling this near 7-year consolidation off 126.53 would uncertain beginning.

Figure 1.

10 THE TECHNICAL ANALYST January/February 2005 Market Views

Figure 2.

Medium-term outlook and ultimately force a break through there to open up the Moving in from the monthly chart to the weekly (Figure 2), upper end of the triangle around 124.77/126.53. the focus is on this recovery off the June 2004 low at 104.86. The weekly RSI and slow stochastics are all bullish Other factors - moving steadily above previous peaks inside overbought The gilt market will have to face some uncertainties this territory whilst the higher %K line on the stochastics is year with the General Election probably in Q2 (although increasing its lead over the rising %D. This is a sign of mar- many investors will already have pencilled in a foregone ket strength and strong underlying support, backed up by conclusion), the widening fiscal deficit (PSNCR) and the the momentum indicator's ability to set a higher low above prospect of slower growth forecasts for 2005. In addition, the zero line and setting fresh 18-month highs. The continuation of a split between Prime Minister Blair and prospect of further gains above 112.45 is hence looking Chancellor Brown could also weigh near-term. The favourable at this stage, opening room for a Fibonacci price Chancellor has been the architect of the "prudent" UK fis- cluster just below the psychological 115.00 level. 50% of cal stance since assuming power in 1997 with the golden the 124.77/104.86 decline ("C") lies at 114.79, and taking rule to balance current government spending over the the values of the converging triangle support and resistance course of the economic cycle. While this is looking some- lines at 105.00 and 124.10 respectively, 50% of this differ- what frayed, it has nonetheless been one of the driving ential comes in at 114.55. This highlights how important forces behind the UK fiscal stance over the last 8 years and the 115.00 level will be, and should be enough to prompt a any change may have a detrimental effect on the attraction noticeable downside correction. It will also establish of long gilts. A breakdown below the 104.50/104.86 lows whether a return to the key highs in the 124.77/125.53 area would negate the equilateral/ascending triangle theory and is possible in the longer-term. For now, a steady progres- mean the "C" wave is still ongoing. This would leave a sion towards this 114.55/115.00 area is expected in the first deeper 3-3-5 corrective flat pattern as the likely scenario half of this year before the risk of a significant setback. from the March 1998 126.53 top. Any significant decline will provide an opportunity for Next on the agenda for a potential turning point would bears to threaten a retest of the range lows at be the corrective bear channel support line, currently near 104.29/104.86. However, it will also give overbought tech- 102.00 but steadily falling towards the base of the previous nical indicators (both monthly and weekly) an opportunity Wave 4 low of one lesser degree, at 98.44 (September to unwind and for longs to take profits after a multi-month 1994). The timing for reaching this point is the first half of advance. Layers of support at 108.85 (minor pivot area), 2006. 107.54 and 106.42 (rectangle congestion) will look to defend the 104.86 low, although as in a Wave 2 correction of an impulsive Wave 1, there is a risk of a near complete retracement. Formation of a higher low above 104.86 Francis Bray is fixed income technical analyst at would prompt renewed bull pressure on the 115.00 area, Informa Global Markets in London.

January/February 2005 THE TECHNICAL ANALYST 11 Techniques

BUILDING A BETTER RSI

12 THE TECHNICAL ANALYST January/February 2005 Techniques

USING BOLLINGER BANDS by Alex Douglas

Improving upon a concept such as the RSI (Relative Strength Index) is no easy task. This indicator has clearly stood the test of time, becoming one of the most ubiquitous of all tools available to the technical analyst. This article presents an overview of the indicator and looks at how its performance may be enhanced with Bollinger Bands.

January/February 2005 THE TECHNICAL ANALYST 13 Techniques

momentum oscillator such as the RSI is constructed to give Aus an indication of the velocity of the market being studied, the idea being that momentum will often begin to turn before the actual price of the market turns and certainly never after a reversal in the price. Consider throwing a ball up into the air. At some point its upward trajectory will begin to slow, even though the ball is still moving higher. During this phase the momentum of the ball will already be in decline despite the ball having not yet met its zenith. The same is true for price trends. The velocity of the trend is likely to turn before the actual price trend turns and certainly no later than the change in prices.

Construction of the RSI To smooth out the impact on momentum of erratic price movements, Wilder suggested looking at the past 14 trading periods to find the average of gains Figure 1. Carlton Communications: The firm bull trend sees the RSI holding toward the made in 'up' periods and the average upper end of the range with plots failing to probe into oversold territory, below 30. lost during 'down' periods. Regardless of the number of up or down periods posted, the cumulative gains and losses are averaged over 14 periods to give a measure of Relative Strength (RS).

Average of gains in up periods RS = Average of losses in down periods

Once this figure is determined the index can be created with the following formula: ⎡ 100 ⎤ RSI =100− ⎢ ⎥ ⎣1+ RS ⎦

This creates an oscillator with a range of 0 to 100 that allows for meaningful comparisons of the momentum of a particular market at different price levels as well as allowing comparisons between different securities.

Interpretation Wilder noted that moves in the RSI either above 70 or below 30 often suggest the imminent appearance of a top Figure 2. Carlton Communications: Manually shifting the oversold/overbought bands from or bottom respectively. Although he 30/70 to 40/80 during a bull trend provides a slightly better result.

14 THE TECHNICAL ANALYST January/February 20052004 Techniques

does not specifically refer to the area above 70 as 'overbought' or the area below 30 as 'oversold', Wilder did mention these terms in relation to oscillators in general at the beginning of the section on the Relative Strength Index in his classic book "New Concepts In Technical Trading Systems". While some would disagree, it has become common market practice to refer to the area above 70 as the overbought zone and the area below 30 as the oversold zone. A probe outside of the 30 to 70 band followed by a modest correction and then a subsequent return toward (but not exceeding) the previous trough/peak outside of the 30 to 70 band sets up a strong indication of market reversal - the failure swing (Figure 1). This signal is triggered on a break of the previous low/high established by the correction. The concept of divergence, common to many technical indicators, is particu- Figure 3. Carlton Communications. Bollinger Bands can be placed on the RSI to create auto- larly useful when interpreting RSI. As matically adjusted levels of oversold/overbought based on market activity Wilder notes, "Although divergence does not occur at every turning point, it does at most significant turning points. When divergence begins to show up after a good directional move, this is a very strong indication that a turning point is near. Divergence is the single most indicative characteristic of the Relative Strength Index." An excursion into overbought or oversold territory which is accompanied by divergence is seen as a particularly noteworthy signal. Analysts can also employ classic pattern analysis on the RSI as well as taking note of areas of support and resistance using the same methods as applied to price charts. In fact, due to the propensity of the RSI to turn before prices change direction, it is not uncommon to find signals appearing on RSI before they appear on the price chart. However, for the purposes of this article we should investigate the concept of overbought and oversold a little more thoroughly.

Overbought and oversold Figure 4. Carlton Communications: In order to make it easier to see when the RSI has moved outside of the Bollinger Bands into oversold or overbought territory we need to flatten out the In a market that is broadly moving side- bands and re-plot RSI in such a way as to represent the position of the RSI in relation to the ways, the parameters of 30 and 70 are Bollinger Bands. This is achieved by plotting %b(RSI). generally adequate to give an indi- →

January/February 2005 THE TECHNICAL ANALYST 15 Techniques

cation of the overbought or oversold status of the market. However, in a firmly trending market it is not uncommon for an oscillator such as the RSI to gravitate toward one end or the other as dictated by the direction of the price trend. A market rallying firmly may rarely see an RSI plot fall below 30. Similarly, a bearish market may rarely see an RSI plot above 70. As the best signals are often given by the RSI falling into oversold territory during a bull trend or an exploration into overbought territory during a bear trend, it has become common practice to shift the bands delin- eating oversold and overbought territory, based upon the prevailing market direction. During a bull trend the bands are often adjusted so that the oversold area increases to be any level below 40, while the overbought area contracts to Problem oscillator gravitating toward one end or be any level above 80 (Figure 2). In the On first glance this manual adjustment the other depending on the direction of case of a bear trend it is common to of the bands seems like a reasonably the price trend. But the bsic problem lower the band to 20-60. good solution to the problem of the remains - Exactly how do we know when is the right time to be manually adjusting these bands?

Solution An elegant solution to this problem that automatically adjusts the overbought and oversold levels was discussed by John Bollinger in his epony- mous book, "Bollinger on Bollinger Bands". Quite simply, this approach involves plotting Bollinger Bands on the RSI with the upper and lower bands (Figure 3). The basic theory underlying Bollinger Bands when applied to price action is also valid when the bands are used with indicators. Thy provide a measure of the volatility of the instrument being studied through the use of the statisti- cal tool of standard deviations. These standard deviations are plotted both above and below a simple moving average to create a dynamic band which contracts when volatility is low and expands when volatility increases. Due to the requirement of the standard Figure 5. Carlton Communications: Placing the regular RSI and the normalized RSI, or deviation calculation to square varia- %b(RSI) on the same chart we can see that %b(RSI) gives us signals that would otherwise have tions from the mean, Bollinger Bands been missed. are very quick to react to sudden sharp

16 THE TECHNICAL ANALYST January/February 2005 Techniques

moves in the instrument (price or indi- RSI lies in relation to the Bollinger cannot claim that %b(RSI) also gave a cator) being studied. Bands with the lower band represented divergence signal as the price low of When applying Bollinger Bands to an by a horizontal line at 0 and the upper 11-Sep-03 was marginally lower than indicator, Bollinger suggests changing band represented by a horizontal line at the price low of 01-Oct-03. the standard parameters for both the 1. An excursion below 0 or above 1 Nevertheless, divergence is commonly time period and number of standard represents the RSI moving outside of seen on the %b(RSI) while the regular deviations. In general terms it is sug- the Bollinger Bands. Given the similar- RSI is giving no hint of divergence or gested that longer time periods should ities to the stochastics indicator (which an imminent turn. In this instance be used when applying Bollinger Bands uses the terms %d and %k) this indica- %b(RSI) was clearly indicating a bias to to indicators, with 50 periods suggested tor has been given the name %b(RSI). the upside. for a standard 14-period RSI. The Interpretation methods for this new number of standard deviations is only indicator are the same as for the regular Summary slightly increased from the default of 2 RSI. During a bull trend the most pow- Placing Bollinger Bands on RSI to gen- to 2.1. erful signals generated by RSI are likely erate dynamic oversold/overbought Instead of using 'fixed' levels to indi- to be failure swings and signs of diver- regions helps to overcome the tenden- cate the overbought and oversold gence following probes into oversold cy of an oscillator such as RSI to grav- zones, the upper Bollinger Band territory. As seen in Figure 5, the firm itate towards one end or the other dur- becomes the marker of overbought ter- rally in Carlton Communications during protracted trending moves. ritory and the lower Bollinger Band ing 2003 had the RSI gravitating toward Examining the position of the RSI in marks the oversold zone. The dynamic higher levels, failing to return below 30 relation to the Bollinger Bands with nature of Bollinger Bands removes the once the rally got underway in mid- %b(RSI) will often reveal noteworthy need to make manual adjustments to March. During this same time period, divergences that are simply not revealed the overbought/oversold zones and %b(RSI) did manage to probe into with RSI while other signals such as ensures that these levels are appropriate both overbought and oversold territory, tops/bottoms and failure swings often for current market activity whether the giving us valuable clues as to the likely appear slightly earlier on %b(RSI). underlying instrument is moving side- direction of the market. The most ways or is bullish or bearish. notable difference visible between RSI Obviously, plotting Bollinger Bands and %b(RSI) in Figure 5 is in the way on the RSI and then attempting to use these two indicators reacted to the lows the upper and lower bands as indicators of September/October. of overbought and oversold territory RSI fell to a low of 37.5 on 29-Sep- makes life difficult when the Bollinger 03, failing to reach the oversold region Bands are continually expanding and below 30 and thus failing to reach the contracting, thus causing the absolute level Wilder stipulated as a requirement level of overbought and oversold to be for a significant bottom or failure continually shifting. Again, Bollinger swing. If we had had the foresight to has suggested a very elegant and simple lift the oversold/overbought levels to solution to this problem. 40/80 in reaction to the bull trend, the When using the 'fixed' levels of 30 slip below 40 to 37.5 would have given and 70 we are interested in where the a bottom signal and even a minor diver- RSI is in relation to these bands. Is the gence signal. However, the %b(RSI) RSI below 30, above 70, or somewhere indicator required no manual adjust- in between the two? Once we replace ment and gave a very clear indication of the 30/70 levels with Bollinger Bands an imminent turn higher as it hit its low our interest shifts to where the RSI lies 11-Sep-03, the same day as the pullback in relation to the Bollinger Bands. In low in price. Subsequent activity in order to make this relationship easier to %b(RSI) saw the appearance of a clas- see we need to flatten out the Bollinger sic failure swing as the indicator lifted Bands so that they become horizontal from -0.18 to 0.42 before slipping back lines on our indicator, just as the 30 and to -0.08. The failure swing was con- 70 levels were (Figure 4). firmed on 06-Oct-03 as %b(RSI) lifted The following formula shows where back above 0.42, the same day that Alex Douglas is a consultant and co- prices took their first significant step founder of G7FOREX.com. He is a (RSI − Lower RSI BBand) towards resuming the underlying bull board member of the International %b(RSI) = (Upper RSI BBand − Lower RSI BBad) trend. In this particular example we Federation of Technical Analysts.

January/February 2005 THE TECHNICAL ANALYST 17 Techniques

DO YEARS ENDING IN '5' REALLY SIGNAL A BULL RUN FOR STOCKS?

Five is the magic number for US stocks according to some market commentators in the popular press. Many recent stories have drawn attention to the fact that years ending in '5' (1905, 1915, 1925, etc) have seen the Dow end higher in every decade since 1880 (Table 1). Is this part of a 10- year stock cycle or simply coincidence? No other year in the decade has equalled this performance which has seen returns in '5th' years averaging around 32%, far exceeding the average returns of other years (Figure 1).

18 THE TECHNICAL ANALYST January/February 2005 Techniques

Year 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th

Up 7768127 61094

Down 5564056238

Table 1. Up and down years 1895-2004

5 % return % 0

-5

-10 12345678910 years ending in

Figure 1. Dow average returns (%) for each year-ending since 1885

arry Williams, CEO of includes references to interest rate have been 'up' in 8 out of the 12 Commodity Timing, a consul- cycles, the weather, and a psychological decades since 1885 (12 x 63%). And the Ltancy, says the 5th year perform- propensity for humans to think in ten likelihood of any one of those year- ance may be due to the US presidential year blocks - remain unconvincing. Jeff endings being an 'up' year on all 12 election cycle which has shown a strong Hirsch of The Stock Traders Almanac occasions is 4% (63%12 x 10). This is tendency for stocks to rally in the year told The Technical Analyst, "I've yet to analogous to asking ten people to toss prior to the election year and the year find anything other than coincidence to a coin that is weighted 63% in favour of after an election where the incumbent explain the phenomenon." heads, 12 times in a row. The probabil- retains his presidency. If this is so, then The 5th year phenomenon is also less ity of one person throwing 12 heads in years ending in 5 rallied because of the noteworthy when you consider what succession is small but not unreason- presidential election cycle in eleven out has happened to the Dow since 1885. ably so. of the twelve decades. The general trend upwards (which Finally, while it remains difficult to However, the presidential election includes the effect of inflation), has dismiss the phenomenon outright, it cycle could apply equally to other odd- meant the Dow has seen more 'up' should be remembered the sample size numbered years in the decade since years than 'down' years. In the 120 upon which the observation is made is elections always take place in even years from 1885 to 2004, 76 years have small. There are only 12 decades of years. Yet years ending in 1, 3, 7 and 9 been 'up' and only 45 have been 'down'. data available for analysis, meaning show no exceptional returns when In other words, 'up' years account for there are only 12 observations for each compared to years ending in 2, 4, 6 and 63% of the total. year-ending. As such, it is probably best 8. In fact, most explanations offered This means that each of the ten year- to look elsewhere for guidance on stock for the 5th year phenomenon - which endings can, on average, be expected to market direction in 2005.

January/February 2005 THE TECHNICAL ANALYST 19 Techniques

20 THE TECHNICAL ANALYST January/February 2005 Techniques

SENTIMENT INDICATORS

Sentiment indicators are used for gauging market opinion and, as such, are effective tools for contrarian trading and investing. We present ten of the most widely used indicators.

ontrary opinion holds that volatility. Option prices tend to rise as Russian Debt Crises in 1998 (Figure 1). when the vast majority of mar- market volatility increases. Options are The VIX is a contrary indicator which Cket participants are bullish or priced using a derived 'implied volatili- means that high readings suggest the bearish, then a reversal in trend is ty' measure, the expectation of future market is oversold and may be near a imminent. In other words what the volatility during the life of the option. bottom while low readings indicate majority think is assumed to be wrong The original VIX was constructed overbought conditions and market so it's best to position yourself on the using the implied volatilities of eight complacency. The average level of the opposite side of the market. To quote S&P 500 (OEX) options thereby repre- VIX since 1995 has been around 20. As Warren Buffet, "Be brave when others senting the implied volatility of a hypo- such, high readings indicated a buy sig- are afraid and afraid when others are thetical at-the-money OEX option with nal and low readings a sell signal. brave." exactly 30 days to expiration. In 2003, However, the VIX is rarely used in iso- The following sentiment indicators the CBOE introduced a new VIX lation and is most useful when used in are widely used tools for following his measure. This uses options on the S&P tandem with traditional technical tech- advice. 500 index and uses a wider range of niques such as RSI and moving aver- strike prices rather than just at-the- ages. Bollinger bands and moving aver- money series. This measure is consid- age envelopes can also be applied to The VIX ered more representative since it incor- VIX charts in order to identify the In 1993, the Chicago Board Options porates information from option prices underlying trend and establish extreme Exchange (CBOE) introduced the over the whole volatility skew not just levels. Volatility Index (VIX). Since volatility from at-the-money options. often signifies financial turmoil, VIX is Historically, the VIX rises during peri- often referred to as the "investor fear ods of investor fear and falls as panic Commitment of Traders Report gauge". The VIX is based on real-time subsides. This effect can be seen in the The COT report was first published for options prices and so reflects investor's VIX behaviour isolated during the agricultural commodities in 1962 and changing expectations of future market Long Term Capital Management and exists because the Commodity Futures Trading Commission (CFTC) requires that traders report their daily positions. The COT report, released every Friday, provides a breakdown of each Tuesday's open interest for markets in which 20 or more traders hold positions equal to or above the reporting levels established by the CFTC. What is important is the change in the net trader position for each category from week to week. The net trader position is the long contracts less the short contracts The CFTC classifies traders into three groups: commercial traders (hedgers), non-commercial traders (large speculators), and small traders (small speculators). Commercial hedgers are institu- tions and individuals who operate → Figure 1. in the cash market of the underly-

January/February 2005 THE TECHNICAL ANALYST 21 Techniques

be later in indicating market reversals. Michael Krauss, global head of fixed income technical analysis at JP Morgan, says, "I have always used the large speculator positions and looked for extremes in long or short positions". Craig Ferguson, chief technical analyst at ANZ Bank agrees, "We look at the non-commercial speculative positions because these positions really are the swing factor in most markets. We believe that commercial positions are more long-term in contrast with the speculators who constantly change their position according to the underlying trend". The COT report is available at: www.cftc.gov.

Figure 2. The Sentix ing commodity. The non-commercials are assumed to be the smart money and The Sentix indicator was established in are speculative traders who are general- have access to superior market infor- Germany in 2001 by Manfred Hubner ly classified as fund managers. Small mation. The idea therefore is to posi- and Patrick Hussy. The index began speculators include all speculators with tion oneself on the same side of the with the S&P 500 and has since positions below reportable limits as market as the commercials. In practice expanded to include 9 additional stock, specified by the CFTC. however, most professionals tend to bond and currency markets. Sentix is To some extent, opinion is divided as look at the non-commercials or large available on Bloomberg, Reuters and to what category to look for in the speculators. George Slezak of website shortly on Updata. COT report. For comparison, we pres- consultancy Commitments of The Sentix is a survey of investor ent commercials' and non-commercials' Traders.com says COT data should be opinion taken every Friday and pub- positions for the S&P 500 future in treated as an overbought and oversold lished on Monday morning. Sentix polls Figures 2 and 3. Most text books sug- indicator but commercials tend to hold a total of around 1200 private investors gest paying most attention to the com- their positions longer than non-com- and institutional professionals such as mercials. This is because larger hedgers mercials as so their net positions may fund managers for both their one- month and six-month outlook for each market. These two private and institutional polls are published separately and also aggregated to produce the 'headline' Sentix index (See Figure 4). A Neutrality Index is also listed which is the poll of all investors who expect the market to move sideways. The neutral level for the headline Sentix is zero. The index is calculated by subtracting the number of bears from bulls and divid- ing by the total. An overall positive number is bullish and a negative number is bearish. If the index deviates far from zero then conditions may be considered overbought or oversold. The separate institutional index may also be viewed as the 'smart' money with the private investor index being used as a Figure 3. contrary indicator. www.sentix.de

22 THE TECHNICAL ANALYST January/February 2005 Techniques

ratios are available for CBOE traded equities, equity indices, US Treasuries, the short-term interest rate index and oil. www.cboe.com/data

The Consensus Bullish Sentiment Index The Consensus Index of Bullish Market Opinion is a weekly poll of professional brokers and advisors. The index is a contrary indicator indicating when market conditions may be overbought or oversold (Figure 6). The poll is taken every Friday. The index covers 31 markets including foreign exchange, bonds and commodities plus a generic US stocks Figure 4. index. An index of 75% or above signifies an overbought level and a reading of 25% or less indicates oversold conditions. Richard White of Consensus Put/call ratios put volume. When the market is bear- Inc. recommends the index is best used The put/call ratio usually refers to ish, the ratio is above average as puts to compliment another indicator such those published by the CBOE, the exceed calls. as Bollinger bands or moving averages. world's largest option exchange. The The put/call ratio works best as a www.consensus-inc.om put/call ratio for any market is simply contrary indicator when the ratio is far the volume of put options traded on above or below its long-term average the exchange divided by the volume of level indicating the market is oversold Advisors Sentiment or overbought (Figure 5). The volatility call options. As such, the ratio gives an This indicator was established in 1963 of daily data means a clearer picture is indication of market sentiment. When and is published every Tuesday. The usually given by applying an appropri- the market is bullish, the ratio falls report takes the opinions of 100 inde- ate moving average. Daily put/call below average as call volume exceeds pendent market advisors in the US and measures the balance of those with a bullish, bearish and neutral outlook. The typical reading, which implies market equilibrium, is 45% bulls, 35% bears and 20% neutral. This is a bull/bear difference of 10%. The report works as a contrary indicator so a bull/bear difference of around 30 and above may be an indication to sell. If the report falls below 0 this is considered bearish and so a signal to buy. Dominic Hawker, head of research at Investor's Intelligence says the report is targeted primarily at the professional market and is especially popular with US fund managers. The report is available exclusively at: www.investorsintelligence.com.

Figure 5.

January/February 2005 THE TECHNICAL ANALYST 23 Techniques

Barron's Confidence Index Developed in 1932, the Barron's Confidence is the longest established market sentiment index. The index measures the ratio of the average yield on 10 top-grade bonds to the average yield on 10 intermediate-grade bonds. The discrepancy between the two establishes a measure indicating investor confidence in the economy. If investors are opti- mistic about the economy, they are more likely to invest in speculative bonds, thereby driving speculative bond yields down, and the Confidence Index up. On the other hand, if they are pessimistic about the economy, they are more likely to Figure 6. move their money from speculative grade bonds to conservative high- grade bonds, thereby driving high- grade bond yields down and the Confidence Index down.

Vickers Insiders Index Insiders (e.g. company directors) are assumed to have superior information about their firm's future performance so may buy their company's stocks if they expect the share price to rise and sell if they expect it to rise. This weekly index is published by Vickers Stock Research Corporation, part of the Argus Group, and reports insider transactions for the past six months for companies included in the Dow.

Odd Lot Balance Index Figure 7. The Odd Lot Balance Index shows the ratio of odd lot sales to pur- chases (an odd lot is a stock transac- tion of less than 100 shares). The assumption is that the odd lots are Other indicators: executed by the market's smallest traders who are the least well The Coppock Indicator informed market players. When the Edwin Coppock developed the Coppock Indicator with one sole purpose: to Odd Lot Balance Index is high, odd identify the commencement of long-term bull stock markets and successfully lotters are selling more than they are did so in 1988 and 1994. The indicator was originally designed for use on the buying and are therefore bearish on Dow Jones but is also suitable for other market indices. It is typically calculat- the market. Therefore it is best to ed by taking a 10-month moving average of the sum of the 11- and 14-month buy when they are selling (as indi- rates of change. The Coppock indicator is available on CQG and Bloomberg. cated by a high OLBI) and sell when they are buying (as indicated by a low OLBI).

24 THE TECHNICAL ANALYST January/February 2005 Techniques

FIVE FIBONACCI STUDIES by Yuriy Zubarev

When Leonardo Fibonacci was solving a rabbit population problem in the 13th cen- tury he might not have known what kind of affect his discovery would have on our modern world of investment and trading. The significance of his finding has been constantly demonstrated and built upon ever since. But even today the majority of market analysts tend to reduce Fibonacci studies to only two practical applications - Fibonacci retracements and extensions - in spite of there being a number of other Fibonacci studies that can be used to good effect.

This article provides a summary overview of five tools which, although visually different, are all geometrical representations of Fibonacci ratios. They are time projections, channels, ellipses, fans, and spirals.

January/February 2005 THE TECHNICAL ANALYST 25 Techniques

Time projections Channels Ellipses Fibonacci time projections indicate Market behaviour is not only reflected Fibonacci ellipses identify the underly- when a price event is supposed to in chart patterns as large swings, small ing structure of price moves. We can occur. The base for drawing this shape swings or trend formations, it is also find long and short ellipses, fat and thin is two critical points: two highs, two expressed in peak-valley formations. ellipses and those that are flat or have a lows or a low and a high. Fibonacci lev- Fibonacci channels make use of peak steep angle. There are very few market els are projected into the future based and valley formations to forecast major price moves that do not follow the pat- on those points (see Figure 1). How far changes in trend direction. The reliabil- tern of a Fibonacci ellipse. In general, out the tool forecasts depends on the ity of Fibonacci channels depends on the shape of an ellipse is defined by the size of the price swing considered. which tops and bottoms are used to ratio of the major axis to the minor Longer-term projections tend to be locate the channel base line (see Figure axis. An ellipse is turned into a given when larger price swings are con- 2) - only major tops and bottoms Fibonacci ellipse where this ratio is a sidered, and vice versa. It's worth not- should be considered. The channel member of the Fibonacci series. ing, however, that it is impossible to say itself should contain one or more A Fibonacci ellipse starts at the begin- whether those levels will mark peaks or prominent side swings. The trigger line ning of a wave A in a 3-wave pattern valleys. If the price is declining or rising is located at the widest swing away and its width is chosen so that the approaching a given time projection from the base line. Once the base line beginning and the end of a wave B is level, this level will likely mark an end and trigger line are located, support and tied between the extreme points of the or a pause to that particular trend. Time resistance lines can be drawn weeks and ellipse's minor axis. As long as the price projections should always be combined months into the future. stays inside the ellipse it is likely to with other Fibonacci tools for more reach the level indicated by the end of dependable signals. the major axis. When Fibonacci ellipses can be plotted, they provide a solid overall picture of the total price pattern, no matter how many waves or sub-waves occur. (See Figure 3).

Figure 1. The period between an original price swing of VZ stock is multiplied by Fibonacci ratios of 2.618 and 11.09 to give timing forecasts for another two swings.

26 THE TECHNICAL ANALYST January/February 2005 Techniques

Figure 2. Prominent peaks and valleys create an opportunity to see major support and resistance lines using Fibonacci channels.

Figure 3. Fibonacci ellipses with ratios of 11.09 and 4.236 identify target price and timing. →

January/February 2005 THE TECHNICAL ANALYST 27 Techniques

Figure 4. A Fibonacci fan

Figure 5. A Fibonacci spiral

28 THE TECHNICAL ANALYST January/February 2005 Techniques

Figure 6. Combining the studies (including retracements and extensions)

Fans shapes as in Figure 6. It's also good Fibonacci fans are drawn using tops or practice to confirm trading signals on “FIBONACCI IS bottoms. They are made up of three two or more time frames. trendlines which start from a common THE MOST point. The three trendlines project into More than one way … the future with slopes of 38.2%, 50% Fibonacci is not just a numbers game; IMPORTANT and 61.8% (although additional levels it is the most important mathematical can be used) and these lines can then be representation of natural phenomena MATHEMATICAL used to indicate levels of support and ever discovered. The financial mar- resistance. (See Figure 4). kets seem to recognize this fact and REPRESENTA- Fibonacci studies are now widely used Spirals by analysts and traders. Yet this analy- TION OF Fibonacci spirals provide the optimal sis has, by and large, been dominated link between price and time analysis. by the use of Fibonacci retracements NATURAL Corrections and trend changes occur at and extensions. The studies and all those prominent points where the shapes outlined in this article should PHENOMENA Fibonacci spiral is touched on its provide a more diverse toolbox with growth path. If the correct centre is which to apply Fibonacci principles chosen, Fibonacci spirals identify sup- EVER to the markets. port and resistance levels that can pin- point turning points in the market with DISCOVERED.” remarkable accuracy. (See Figure 5).

Fibonacci confluence Yuriy Zubarev is president of To significantly enhance reliability, trend reversals or continuations can be Solution Realm Software Inc. See confirmed with two or more Fibonacci www.fibonaccisolution.com

January/February 2005 THE TECHNICAL ANALYST 29 Interview

THE TECHNICAL ANALYST TALKS TO…

Adam Sorab

Adam Sorab is head of technical research and trading at CQS, a hedge fund management company. Prior to this he was a director of Absolute Return Strategies at Deutsche Asset Management from 2000 to 2004. He has served as a director on the UK's Society of Technical Analysts executive committee since 1993 and been chairman since 1996. Mr Sorab is also a director of the International Federation of Technical Analysts.

TTA: What does your current role with CQS involve? investment process. The reality is that professional investors have for a long time realised that the two approaches are AS: CQS manages about $4.7 billion in hedge fund strate- not mutually exclusive but instead can be highly compli- gies. My work involves the development of technical strate- mentary. While it is true that many hedge funds today use gies to support our trading activities and the management technical analysis exclusively to generate investment strate- of portfolios. gies I don't think this reflects a move away from fundamental analysis. The reality is that it's the fundamentals that TTA: Your first City job was with Bank of Credit & actually move markets. The advantage a technician has over Commerce from 1984-86 as a money market dealer. How an economist is that he or she may pick up on a change in has the City's perception and use of technical analysis sentiment and trend before the change in the fundamentals changed since then? has been picked up.

AS: Technical analysis has become increasingly accepted TTA: What are your favourite TA techniques that you use within the City over recent years. In reality it had always had in your trading strategies? a strong following amongst FX, fixed income and commodity professionals. The bear market that developed in the last AS: I tend to favour the KISS approach to technical analy- few years did a lot for promoting TA amongst equity mar- sis and as a result tend to focus on prices and volumes ket participants. When the STA started out, over 25 years rather than complex oscillators and second order deriva- ago, technicians in the UK were not popular and tended to tives. I tend to look for classic support/resistance levels and huddle together in the corners of pubs sharing chart trendlines to pick my spots to enter and exit markets. For insights. Today technical analysis forms an increasingly long and medium-term analysis I like candlestick charts, important part of all research published and even gets good Ichimoku Morning Clouds, point-and-figure. For short term coverage from the financial markets television media. As a analysis, particularly in the futures market, I also like using result technicians find a greater and more receptive audi- Market Profile. ence for their research and are able to obtain better resources from their employers to produce their work. TTA: What are the major developments in TA at the moment? TTA: Do you think TA is now a major part of all hedge funds' strategy? Do you detect a move away from funda- AS: TA is always evolving and the advent of cheap comput- mental analysis to any extent? ing power has accelerated this progress in recent years. I don't see very much that is really new these days in TA. AS: Hedge funds are very different to many classic invest- However, there is a lot more cross fertilisation of ideas ment houses. Not least because the portfolio managers tend across disciplines and markets. For example while many to invest their own money alongside their clients and earn Japanese techniques have existed for centuries, it's really the bulk of their fees from their share of the absolute only been in the last 10 years that these have been properly returns. As a result they are very focused on positive returns taken up by western analysts. I think IFTA deserves a lot of and are open to all and any approaches that bring them credit for encouraging this ongoing exchange of ideas closer to achieving these goals. Accordingly, technical analy- between technicians all over the world. From my point of sis is employed in one form or another by pretty much all view the real development in TA is the extent to which the hedge funds; even those who have a firmly fundamental ordinary private investor now has access to TA software

30 THE TECHNICAL ANALYST January/February 2005 Interview

and analytics. In the past TA was the City's secret weapon. er is that they change so slowly as to be almost worthless as Today, the investment bank prop desks have to compete an aid in supporting day-to-day position taking and strategy. with Joe Public when we get to big chart points. This has got to be a good thing in the long run; even if it does make TTA: As chairman of the STA what are your ambitions for the markets a little choppier whenever we get to major chart the organisation? points. AS: I hope to see the STA continue to grow in size and TTA: What do you think about backtesting TA techniques resources. Over the past 25 years the STA has developed and trading strategies? into one of the oldest and most respected TA societies in the world. I would like to see it build on this success by AS: I have systems that allow me to set up a mechanical expanding its membership and increasing the services it is trading strategy and then backtest it to establish how well it able to offer its members. A major part of the STA work would have fared had it been trading in the past. The soft- relates to its accreditation program and the STA Diploma ware even allows me to optimise the system to maximise (which also forms the basis of IFTA's international DITA2 the historic returns. This all sounds great in theory however exam). In the future, I would like to see more people use in practice it's a lot more complex than that for a number this resource to get properly qualified as technical analysts. of reasons. Firstly, it's not obvious that a static system is the best way to trade markets. Instead more adaptive/reactive TTA: How much time do you devote to your role with the approaches may be more suitable. Secondly, it's not obvious STA? that one should always be just looking to maximise returns especially if the price of doing so is high volatilities or huge AS:: I am fortunate to have a large volunteer board at the adverse excursions. Thirdly, it's unlikely that the systems STA so the work gets spread around many people. Each that work best in the past will always work as well in the month we hold formal board meetings and hold lectures. In future. As a result, systematic approaches have to be treated addition to that a lot of work gets done via email over the very carefully by investors even when the backtested results weekends. Several years ago the STA elected to employ a look fantastic. There are a number of excellent books on professional administration company to deal with the day to this subject which are worth reading. One good one I read day needs of running the society. Today STA recently was Thomas Stridsman's "Trading systems that Administration, along with all the other active board mem- work". bers, does a tremendous amount of the real work. This makes my job a lot easier. TTA: To what extent, if any, should people combine behavioural finance and quants with TA? TTA: How has the membership of the STA changed in recent years both in number and nature of members (pri- AS: To my mind behavioural finance and technical analysis vate traders/professionals)? are one and the same thing albeit approached from completely different angles. Behavioural finance attempts to AS: In the early years the STA membership was mainly pro- understand how psychology affects financial decision mak- fessional and then in the 80s it began attracting a lot more ing and financial markets. It has developed out of acade- private investors. Today the membership is about 800 in mia's recognition of the fact that classical financial econom- total. About 55% of these are professionals working in the ics is a blunt tool for supporting trading decisions in mod- financial markets. ern developed markets. Technical analysis attempts to use price action and volumes traded to interpret trends or TTA: What changes do you plan to make to the STA syl- changes in market psychology. In this way the two fields are labus? very similar. AS: The STA syllabus is constantly under review and we are Quantitative finance is a very different field and I think planning to make a number of changes to it in the near highly complementary alongside TA when used to assist in future to reflect the proposed changes to IFTA's interna- portfolio construction and risk management. tional exam accreditation programme. Precise details of these changes are yet to be announced but in essence they TTA: How important are fundamentals? involve the incorporation of several Japanese techniques and the addition of Ethics within the syllabus. We think AS: As an economist by training I can assure you that fun- these changes will greatly enhance the syllabus and ensure damentals are very important. The problem I have as a trad- the exam retains its international high standing.

January/February 2005 THE TECHNICAL ANALYST 31 Subject Matters

32 THE TECHNICAL ANALYST January/February 2005 Subject Matters GOLD AND THE KONDRATIEV CYCLE by Michael Alexander Michael Alexander uses 'reduced prices' to reveal Kondratiev waves in modern price data. He goes on to use his findings as a tool for modeling the price of gold.

50 to 60 year economic cycle expenditures. By running a deficit the was first described in detail by government can put more money into AN. D. Kondratiev. Kondratiev the hands of consumers than it takes in cycles are most readily apparent in taxes, which increases the amount of monetary data such as prices and inter- money available for spending. est rates. Figure 1 shows a plot of the Monetary policy affects the money sup- US Producer Price Index (PPI) (in red) ply and the Federal Reserve has a vari- over the period 1790 to the present. ety of means by which it can increase Prior to WWII, the Kondratiev cycle the amount of money in circulation. can be seen clearly as periodic waves in Both of these policies are stimulatory prices spaced about 50 years apart. in that they can increase the rate of Kondratiev, working in the 1920's, economic growth (at least in the short- predicted a deflationary depression term) and produce rising prices. using his cycle theory. He was right, but To account for these effects, I define subsequent attempts to apply his theo- stimulation (S) as the sum of govern- ry have been unsuccessful due to the ment debt and money supply relative to absence of declining prices after the real GDP (See box). Great Depression. Since then, inflation From equation 4 it follows that veloc- has been an ever-present fact of life as ity is equal to Price divided by S. I call the steadily rising red line in Figure 1 the P/S ratio the "reduced price". shows. Near-constant inflation over the According to standard quantity of past 70 years is the result of stimulato- money theory, velocity is supposed to ry economic policy that began with the be constant and so reduced price New Deal in the US and continues to should be approximately constant as this day. Stimulatory policy falls into well. It is not. As shown in Figure 1 two categories: fiscal and monetary. (black), reduced price shows the same Fiscal policy involves government wave-like pattern as regular prices. →

Figure 1. PPI Commodities Index and Reduced Price since 1790

January/February 2005 THE TECHNICAL ANALYST 33 Subject Matters

concept of stimulation is applied to gold and the US went back on the stan- 1. S = (Debt + Money) / GDP gold. dard. What this means is the price for Government debt is calculated by gold given by the stimulation model summing government deficits over Modeling the price of gold (the true value of gold as "money") time. For money supply I use the M3 Gold is a unique asset in that it is both rose above the market price in 1862 and measure. (For the period prior to a commodity and a monetary instru- didn't fall back until 1879. 1959 when M3 data is unavailable, I ment. As a metal, it is subject to the From 1879 to just before World War use M2). I now write a standard same sorts of supply and demand pres- I, the US dollar was on the gold stan- quantity of money equation for sures that apply to other metallic com- dard. This is shown in Figure 2 by the inflation: modities. Unlike other commodities, close correspondence between the 2. Money x Velocity = Price x Output however, gold has often served as price of gold predicted by the stimula- I can rearrange this and substitute money and thus has use as a store of tion model and the actual POG. Of GDP for output and obtain: value (i.e. savings). The willingness of course, this is necessary since the whole 3. Money / GDP = Price / Velocity gold owners to keep large stocks of the purpose of a gold standard is to regu- If I include government debt in my metal out of economic use means that late stimulation in such a way that this definition of money I obtain gold can trade (for years) way out of correspondence occurs. During WWI 3. (Debt + Money) / GDP = Price / line with the supply and demand stric- the US (and the rest of the WWI bel- Velocity tures that apply to other commodities. ligerents) went off the gold standard Substituting equation 1 into equation Figure 2 outlines the history of gold just as the US had done for the Civil 3 above, gives: prices. The money role of gold means War. After the war, the US (and eventu- 4. S = Price / Velocity that we should be able to apply the ally the rest of the western world) went stimulation concept used to produce back onto gold. But this return to gold reduced prices to the price of gold was not at parity. That is, the actual More importantly, it continues to show (POG). Figure 2 shows a stimulation- price of gold was well below what the a wavelike pattern after the Great based model (bold black line) for the stimulation model calls for. For parity Depression, after which ordinary prices price of gold. This model simply to be reached it would have required stop doing so. Reduced price clearly assumes that the gold price is linearly politically unacceptable levels of defla- shows a Kondratiev trough in 1946 and related to stimulation. For the period tion after World War I and so the gold a peak in 1981. These dates are close to 1860-1915 the POG and the stimula- standard was re-established under what those estimated from a consideration tion model closely agree. The only dis- was called the gold-exchange system. of interest rates or stock market cycles, crepancy is in the 1862-1879 period Throughout the 1920's the economic both of which correspond to the during which the US dollar was off the authorities were able to maintain stable Kondratiev cycle. It would seem that gold standard. In order to finance the currencies well above the value consis- reduced price captures the Kondratiev Civil War, the US government had to tent with gold. When the world econo- cycle that underlies the actual price. Or print money in the form of greenbacks. my fell into depression after the 1929 put another way, the Kondratiev cycle Doing so took the US financial system stock crash, the gold exchange system for actual prices has been distorted by off gold in 1862. It was only in 1879 was no longer tenable and one country stimulation. In the next section the that the dollar came back to parity with after another went off gold. The US left gold in 1933 and a new POG was established at $35 an ounce by a deval- uation of the dollar. World War II introduced another round of stimulation. As a result, the price of gold given by the stimulation model once again rose above the POG. This time the world did not return to the gold standard. Instead, the world currencies were pegged to the US dollar, which was then pegged to gold at $35 per ounce. The strong trade position of the US coupled with the vast gold reserves of the US treasury meant that the US could keep this new dollar system going for decades despite the Figure 2. Annotated price history of gold lack of parity between the dollar and

34 THE TECHNICAL ANALYST January/February 2005 Subject Matters

The Kondratiev Cycle gold. Eventually the system collapsed in 1971 just as the gold exchange sys- Economists recognize four major cycles, or regular fluctuations, in the economy: tem had before it and gold began to trade as a commodity. After more than 1. Kitchin's short-wave cycle of average duration 3-5 years, discovered two decades of under-valuation, the in 1930; POG exploded upward after 1971. 2. Juglar's cycle of average duration 7-11 years, discovered in 1862; Now freed of its monetary characteris- 3. Kuznets' medium-wave cycle of average duration 15-25 years, discov- tics, gold began to behave as a com- ered in 1923; modity. As the world economy pro- 4. Kondratiev's long-wave cycle of average duration 45-60 years, discov- gressed through the inflationary ered in 1922. Kondratiev summer period, gold rose in price along with oil and other com- Many researchers believe the cycles are inter-dependent. In particular, they note modities reaching a peak around the that the average length of each of the four cycles is slightly longer than double time of the K-peak in 1980-81. That is, the length of the immediately preceding shorter cycle. Kondratiev's cycle may gold rose partly as a response to its pre- itself be followed by a centennial cycle, the secular trend, identified by Braudel, vious under-valuation, and partly as a Simiand, and Larousse in the 1930's. commodity following the K-cycle as The long-wave cycle in the capitalist economy was discovered by the Soviet defined in reduced prices (the thin economist N. D. Kondratieff (1892-1930) in 1922. Kondratiev's methodology black line in Figure 2). involved the analysis of 21 economic indicators including the price index, the By 1980, gold had become extremely rate of interest, wage rates, rents, volume of production, consumption, exports, overvalued relative to its value as imports, employment, as well as their standard deviations, for Britain, France, money. The very fact that gold has been Germany and the US. In studying volumes, Kondratiev used per capita data. He used historically as money suggests that calculated deviation from the trend through the method of least squares. In it has a higher value as money than any order to filter out noise caused by the shorter cycles he employed nine-year mov- other use. Thus, once the Kondratiev ing averages. downwave got underway after 1981 we Kondratiev confirmed the presence of a long wave-cycle in 15 of the 21 series. should expect gold to fall in price along Significantly, in the case of the price level and the rate of interest the evidence with other commodities, but at a faster was strong. Kondratiev's ultimate conclusion was that he obtained sufficient rate, because of its overvaluation. Note empirical basis to support the hypothesis of the existence of a long-wave eco- how gold has fallen in price along with nomic cycle in the capitalist economies he studied, with an average duration of the reduced price. A few years ago, gold 54 years. He allowed a 25 percent deviation from this average. In particular, reached parity with the dollar. That is, Kondratiev identified three historic waves: gold was no longer either over or undervalued compared to the dollar. (i) First wave: rising phase from 1780-90 to 1810-17; falling phase from Since then gold has climbed. 1810-17 to 1844-51. Right now, the model value for gold (ii) Second wave: rising phase from 1844-51 to 1870-75; falling phase is about $320 and it has been rising at from 1870-75 to 1890-96. about $13/year since the beginning of (iii) Third wave: rising phase from 1890-96 to 1914-20; falling phase 2001 due to stimulation from interest started 1914-20. rate cuts and deficit spending. Thus, at Kondratieff was exiled to Siberia by Bolshevik officials who flatly rejected his a POG at or below $320, gold looks conclusions. To the faithful there could only be one falling phase of the capitalist cheap. Indeed this was demonstrated economy, followed by the socialist revolution and the dictatorship of the prole- recently by the rally that followed when tariat. And, following that, there was to be only one rising phase. gold's valuation fell below parity in Kondratiev died in the Gulag in 1930 at the age of 38. His work was later updat- 1999. ed by other economists using his original methodology. They found that the falling phase of the third wave ended 1947-48, and that there is a (iv) Fourth wave: rising phase from 1947-48 to 1973-80; falling phase Michael Alexander is the author of started 1973-80. Stock Cycles (2000, Writer's Club Press), The Kondratiev Cycle (2002, Writer's Club Press) and Cycles in Antal E. Fekete, Professor Emeritus, Memorial University of American Politics (2004, Newfoundland iUniverse.com)

January/February 2005 THE TECHNICAL ANALYST 35 Subject Matters

A COMPARISON OF MOVING AVERAGE ENVELOPES AND BOLLINGER BANDS by Joseph Man-joe Leung and Terence Tai-leung Chong

Trading rules based on simple moving averages are the most frequently tested of all the different trading strategies. However, when the market lacks a clear trend or when prices fluctuate significantly around a trend, it is more appropriate to use channels for capturing short-run fluctuations. Two such channels that are popular with traders are moving average envelopes and Bollinger Bands. Yet few empirical studies have explored their profitability. This article attempts to fill this gap by comparing the performance of these two trading techniques on 11 stock market indices.

Constructing channels The data Moving average envelopes and Index To Bollinger Bands are channels that are Location From based on a moving average. In this Dow Jones Industrials 29/12/2000 study we use the simple moving aver- USA 1/1/1985 Toronto 300 29/12/2000 age, where the N-day moving average at Canada 1/1/1985 BCI Global 29/12/2000 any given time is the mean of prices in Italy 1/1/1985 FTSE 100 29/12/2000 the previous N days. For moving aver- United Kingdom 1/1/1985 DAX 29/12/2000 age envelopes, the channel boundaries Germany 5/9/1989 Nikkei 225 Stock Avg. 29/12/2000 are plotted at a fixed percentage above Japan 3/1/1985 CAC40 Instantaneous 29/12/2000 and below the moving average value. France 16/7/1987 KOSPI 29/12/2000 For our analysis, we use percentage val- South Korea 4/1/1985 Straits Time Index 29/12/2000 ues (k) of +/-3% and +/-5%. Singapore 4/1/1985 Hang Seng Index 29/12/2000 In contrast, the width of Bollinger Hong Kong 1/1/1985 TWSE 29/12/2000 Bands is determined by a certain stan- Taiwan 4/1/1985 dard deviation value from the moving average - i.e. the band width depends region, the trading rule does not signal signal is generated when the price pen- on the fluctuation of prices around the a buy or sell until the stock price has etrates the upper boundary from above. mean. When volatility increases in such moved away from those extreme In this article, moving average a way that the moving average remains regions. Therefore, a buy signal is gen- envelopes of 3% and 5%, and Bollinger unchanged, Bollinger Bands expand to erated when the price crosses the lower Bands of two standard deviations are capture the price fluctuations, while boundary from below. Similarly, a sell applied to 10-, 20-, 50- and 250-day moving average envelopes will not. Bollinger Bands are normally plotted N = 10 20 50 250 with two standard deviations (used here), which should capture around Average annualised returns 95% of price movements. MAE-3% 2.4% 3.2% 3.8% 1.7% The trading rules MAE-5% 3.3% 3.3% 2.5% 1.0% The region above the upper boundary BB (2 SD) 1.9% 2.9% 1.4% 2.2% of moving average envelopes and Bollinger Bands is considered an indi- Average number of transactions cator of overbought conditions, while MAE-3% 45 41 28 45 the region below the lower boundaries is considered an indicator of oversold MAE-5% 15 20 18 7 conditions. However, because it is often BB (2 SD) 53 40 20 4 hard to predict how long a stock will stay in an overbought or an oversold Table 1. MAE and BB averages across 11 stock market indices

36 THE TECHNICAL ANALYST January/February 2005 Subject Matters

moving averages. We assume that trans- Results positive rates of return. Rates of return action costs and stock dividends are Table 1 shows the average annual rates are especially good for the Dow Jones, negligible. We also assume that short of return generated by the two trading DAX, CAC and Hang Seng Indices (e.g. selling is not allowed, and transactions rules across all 11 indices. It also shows the Dow Jones averages 5.6% across cannot be accumulated, i.e., two con- the average number of transactions the 12 trading rule permutations - not secutive buying actions are not allowed. generated. shown here). However, both rules gen- Since there are about 250 trading days erate a negative rate of return in most each year, the performance is evaluated Observations cases for the Nikkei 225 and the in terms of annualized rates of return On average, both trading rules generate KOSPI. Note that the number of transactions generated by moving average envelopes falls with a lower k. This is because the larger the value of k, the more the price fluctuations will be captured by the envelopes and the fewer trading signals will be observed. Note also that the number of transactions generated by both channels falls with increasing N. In general, when N = 10 and 20, the number of transactions for Bollinger Bands and for MAE-3% are close. While for N = 50 and 250, the number of transactions for Bollinger Bands is closer to the number of transactions for MAE-5%. Overall, when used to formulate a mechanical trading rule, moving average envelopes perform better than Bollinger Bands for 10-, 20- and 50-day moving average periods, while Bollinger Bands perform better on a Figure 1. Moving average envelope on Dow Jones Index (20 day moving average, 3 percent envelope). Source: Telerate 250-day moving average. We therefore suggest using moving average envelopes for short-term investment and using Bollinger Bands as a long- term investment tool. However, since technical trading rules are usually designed for short-term investment purposes, it would be reasonable to conclude that moving average envelopes are better than Bollinger Bands in practice. Therefore, despite the fact that Bollinger Bands can capture sudden price fluctuations that moving average envelopes cannot, they do not out-perform moving average envelopes in terms of profitability.

Joseph Man-joe Leung and Terence Tai-leung Chong are professors of economics at The Chinese Figure 2. Bollinger Bands on Dow Jones Index (20 day moving average, 2 standard deviations). Source: Telerate University of Hong Kong

January/February 2005 THE TECHNICAL ANALYST 37 www.londonstockexchange.com

There’s new competition on the horizon.

Introducing euroSETS™ Dutch Trading Service The London Stock Exchange is leading the way in opening up the competitive landscape to ensure that investors get access to the largest pools of liquidity in Dutch securities at the best price. The service will be up to 40%* cheaper than Euronext and this sustainable level is aimed at attracting a wide cross-section of the market.

The most liquid securities will be included: namely, the constituents of the two major Dutch indices (currently 50 securities in total), which account for around 90 per cent of trading in the Dutch market

Competition waits for no-one and we remain committed to a competitive single market in Europe. Working in partnership with trading firms from the Netherlands and the United Kingdom we can offer you the opportunity to access efficient and reliable trading technology, a broad range of international securities and market expertise on a global scale enabling you to fulfil your business objectives.

London is Europe’s leading financial centre and is one of the world’s key investment capitals. Set within the European time zone, it is the natural apex of the global triangle of New York, London and Tokyo.

Changing the market place - www.londonstockexchange.com/dutch *As at 06/04/04

Client Services London - Client Services Amsterdam - Client Services Laura Pandit Scott Coey Maurice Tordoir Tel: +44 (0)20 7797 1328 Tel: +44 (0)20 7797 1131 Tel +31 (0) 65 420 1074 [email protected] [email protected] [email protected]

The London Stock Exchange crest and logo are trademarks of London Stock Exchange plc. Approved for the purposes of sections 21 of the Financial Services and Markets Act 2000 by Cazenove & Co. Ltd of 12 Tokenhouse Yard London EC2R 7AN Software

THE BANK OF NEW YORK'S INTERACTIVE PORTFOLIO FLOW MONITOR

low of funds information is firmly established in +100 to -100. This scale is a normalized index of cumu- the market as important data that can give traders a lative net daily flows in local currency terms. A bar chart Fnoticeable edge. Often, the information is passed (not shown) below each chart displays the net capital flow around anecdotally and is used to explain short-term price for each day that contributes to the aggregate chart above moves. Captured and presented in a verifiable form, this it. The charts, however, are somewhat crude and there is data may also be used for identifying medium- and longer- no source data available to support them. term trends. The Bank of New York's position as the world's largest The iPFM Flow Matrix custodial bank has given it unique access to daily cross- The flow matrix is an add-on to the iPFM which ranks border portfolio flows in the equity and fixed income mar- those investment classes and indicators that exhibit the kets. This has led to the development of the Interactive greatest correlations. The iPFM Flow Matrix sorts and Portfolio Flow Monitor (iPFM), the bank's portfolio flows service used primarily by fund managers and hedge funds. The iPFM uses proprietary data from the Bank of New York's global custody activity. The bank currently has $8.5 trillion worth of assets under administration which represents over 20% of the total global custody market. According to Michael Woolfolk, head of research at The Bank of New York, "This is a sufficiently large sample to fairly reflect total custodial flows for both major and emerging markets." Speaking as an in-house user of the iPFM, he continues, "In the strategy department we are primarily interested in Figure 1. Total portfolio flows and spot USD/CHF any flows likely to trigger FX transactions. Sustained changes in net flows are often a harbinger of changes in the underlying real money flows that ultimate determine the value of currencies."

The iPFM on-line The iPFM was launched in 1998 and became available on the Bank of New York website in 2001. At present, the service only includes portfolio inflows into each country although the service is expected to be expanded to incor- porate bilateral flows in the near future. The iPFM allows Figure 2. Equity flows and Dow the user to plot portfolio flows into 26 countries including the US, UK, Japan, the Eurozone, Australia, New Zealand, Canada and Scandinavia. Also included are 14 emerging markets. The flows are divided into three 'investment classes'- fixed income, equity or the total of both. Against each of these classes can be plotted up to three 'indicators': 25 FX rates (against the domestic currency), 20 benchmark government bond yields and 34 equity indices (see Figures 1-3). All data is updated daily and can be plotted for periods ranging from one month to two years. All iPFM charts represent flows in yellow on a scale of Figure 3. Fixed income flows and 10-year gilt

January/February 2005 THE TECHNICAL ANALYST 39 Software

"REAL MONEY FLOWS ULTIMATELY DETER- MINE THE VALUE OF CURRENCIES." Figure 4. MICHAEL WOOLFOLK, THE BANK OF NEW YORK presents the iPFM data identifying the major trends and as the bank readily admits, it does not have a significant correlations in currency, equity and fixed income markets share in two important areas - official flows such as cen- ranking them on an index of +/- 100. This index is a tral bank reinvestment of intervention proceeds and the measure of relative correlation and says nothing about the Japanese market. Japanese investors still prefer using actual correlation that may exist between an investment domestic custodial banks. class and indicator. For example, even if an actual correla- Basic flow-of-funds interpretation implies that positive tion is low, it may have an index of +100 if it is the high- or increasing capital inflows into a country will support est correlation that the iPFM identifies. To aid in the iden- the domestic currency. This appears obvious as invest- tification of trends, the ranked output is colour coded in ment in US fixed income or equity instruments require the shades that vary from dark blue (positive) to dark red (neg- buying of dollars. However, as a guide to currency move- ative). The matrix (Figure 4.) shows that the strongest cor- ments, the iPFM seldom provides explanations for market relation over the past 3 months was between Brazilian behaviour. For example, over the past 3 months, capital equity inflows and the domestic stock market. Clicking on inflows into the US have been positive although the dollar this square displays the chart (see Figure 5). continued to weaken against the euro and other currencies. The Bank of New York accepts that the service may Not a predictive tool have limited forecasting ability. "The iPFM," says The Bank of New York's claim that their custodial hold- Woolfolk, "is designed to be a value-added explanatory ings are representative of total capital flows is open to tool for financial market professionals and not a predictive question. The iPFM only looks at private sector cross-bor- tool." der flows into each country which means the system can- Nevertheless, The Bank of New York's position as a not, as yet, measure and display net capital flows, surely a custodial bank affords it a unique advantage in having more important determinant of market prices. Moreover, access to portfolio data when no daily cross-border capital flow data is publicly available elsewhere. As Woolfolk explains, "We choose to focus on providing the most up to date and accurate cross-border capital flow data available in the market."

More information is available from www.globalmarkets.bankofny.com Figure 5. Brazilian equity inflows and stock market (IBOV)

40 THE TECHNICAL ANALYST January/February 2005 Book Review

THE (MIS)BEHAVIOUR OF MARKETS

n the late 1980s and early 1990s, Chaos Theory was a fashionable way of looking at the world, complete with essential vocabulary I(turbulence, determinism, fractals) and posters of Mandelbrot sets. Yet today, far from being a theory with which to illuminate, it is too often used to blag - especially in the financial markets where 'black box' solutions abound. Benoit Mandelbrot's book will do much to set the record straight. This is the first book he has written on finance for lay readers and there are few others who could write about the application of fractals - a branch of mathematics that perceives hidden order in the irregular- ity of nature - to the financial markets with such authority and, thanks to his co-author Richard Hudson, clarity. Mandelbrot's CV is impressive: professor of mathematics at Yale, inventor of fractal geometry, and winner of numerous scientific medals and prizes. Richard Hudson also ranks high amongst his jour- nalistic peers, having been managing editor of the Wall Street Journal Europe for six years. The book is organised into three parts. Part one provides a convinc- ing argument against the Modern Theory of Finance. It argues that markets, like most natural phenomena, operate according to a power law. This means that instead of the "mild mannered bell curve", extreme values occur much more often than would be expected. The (mis)Behaviour of Markets Why is it, Mandelbrot asks, that "the seemingly improbable happens A Fractal View of Risk, Ruin and Reward all the time in financial markets." The 6.8% decline in the Dow Jones By Benoit B. Mandelbrot and Richard L. Industrial Average on the 31 August 1998 had, according to Hudson Mandelbrot's measure of standard theory, a one in 20 million chance Published by Profile Books Ltd of happening. 328 pages, £18.99 Such incidents, often viewed as anomalous by the markets, can lead ISBN 1 86197 765 4 to financial ruin. Yet much of financial theory continues to be based on assumptions of normal distribution. This is "… the guiding princi- The (mis)Behaviour of Markets is available ple for many of the standard tools of modern finance… taught in from The Technical Analyst bookshop at the business schools and shrink-wrapped into financial software pack- reduced price of £13.29 plus P+P. To order ages... [It] is a house built on sand." please call 01730 233870 and quote "The Technical Analyst magazine". Books are According to these principles, you are "more likely to get vaporized usually posted within one working day of by a meteorite landing on your house than you are to go bankrupt in a your order. financial market." The second part of the book is humbly entitled "The New Way", in which Mandelbrot talks us through his findings with regular reference to fractal cartoons. His fractal cartoons explain how seemingly irregular chart patterns can be built up by regularly repeating patterns at many different scales. This fractal view of the financial markets can explain why all charts look similar, regardless of timeframe. In essence, his models highlight two important characteristics of the market: Price discontinuity and long-term dependence. With price discontinuity, prices move with "wild randomness" where the "fluctua-

January/February 2005 THE TECHNICAL ANALYST 41 Book Review

tion from one value to the next is limitless and frighten- half of currency speculators play some form of trending." Under these assumptions, the odds of ruin are far following game…" more realistic - more like one in ten or one in thirty. Another property of wild randomness is that volatility This hesitancy seems to stem from a respect for the is concentrated into clusters and these clusters occur at markets and a belief that the markets' understanding, all scales. This also matches the data. For example, nearly conscious or not, may be far ahead of any economic the- half of the decline in USDJPY from 1986 to 2003 ory: "The market has known for years that the Black- occurred in just ten days (out of the 4,695). If you then Scholes formula for option pricing is "simply wrong… focus in on those ten days, then most of the decline was Improving or replacing Black-Scholes is one of the liveli- further concentrated into a few minutes. And so on. est sub-disciplines in mathematical finance." Prices do not move smoothly and continuously from one Referring to the equity premium puzzle, an academic value to the next, but "jump, skip and leap - up and debate which struggles to understand why equity premi- down." ums should be as high as they are, The second important characteris- "REAL INVESTORS Mandelbrot says: tic, long-term dependency, recognises something that technical analysts have KNOW BETTER THAN "Real investors know better than the done for a while - that prices do not economists. They instinctively realize move in a random walk as predicted THE ECONOMISTS. that the market is very, very risky, by the Efficient Market Hypothesis. THEY INSTINCTIVELY riskier than the standard models say. Events of the past are still playing So, to compensate them for taking themselves out in today's prices. One REALIZE THAT THE that risk, they naturally demand and of the authors' illustrations is how the often get a higher return." Great Depression made those MARKET IS VERY, investors who lived through it more VERY RISKY.” In the third section, Mandelbrot risk averse. "Their memories provided and Hudson list what they call the a practical form of long-term dependence in the finan- "Ten Heresies of Finance". Mandelbrot's challenging cial markets. Is it any wonder than in 1987, when most of conclusion to technical analysis is that you cannot fore- those men were gone and their wisdom forgotten, the cast the financial markets. But he does not mean it in the market encountered its first crash in nearly sixty years?" same sense that proponents of the Efficient Market But, he warns, "The long-range dependencies in prices Hypothesis or Random Walk do. To him, such thinking is create a kind of tendency in the data - not towards any equally flawed because it does not recognise the long- particular price level, but towards price changes of a par- term dependency of prices. ticular size or direction." He continues, "The changes can So can any of Mandelbrot's conclusions be used for be persistent, meaning that they reinforce each other… trading and investment decisions? Well, according to him, or they can be anti-persistent, meaning they contradict yes. You may not be able to forecast direction, but with each other. Because of this, he thinks that "spurious pat- the understanding that periods of high-volatility cluster terns" make the markets deceptive and that technical together, then you can step out of the market at those analysts are the ones being deceived. periods and reduce the chance of loss and ruin. The Yet Mandelbrot's attitude towards technical analysis question he leaves unanswered is how to do this in prac- isn't as emphatic as his other views. At times he seems tice. dismissive, describing technical analysis as "financial Regardless of whether you will like his conclusions or astrology" and part of the "fun-house mirror logic of not, The (mis)Behaviour of Markets is highly recom- markets", but elsewhere he seems less certain of how mended to anyone working in the financial markets. It is technical analysis fits into his model, agreeing that: both entertaining (no doubt because of Richard Hudson's involvement but also because Mandelbrot talks "… a by-now substantial body of economics research openly about his route to scientific discovery) and suggests that there is, indeed, money to be made in such informative. Above all, however, the book is recom- a "trend-following" strategy; how much, and whether it is mended because it is a serious presentation, albeit written worth the risk and expense, is a matter of debate. But for the layman, of Mandelbrot's formidable views. clearly, the market pros have already voted: More than

42 THE TECHNICAL ANALYST January/February 2005 Letters

LETTERS TO THE EDITOR

SIR: I saw with interest your article in The Technical returned 572% with no threshold and 694% with a 1% Analyst magazine on the January Barometer. In the past I threshold. Strategy (2) (i.e. long/short) returned 314%. have done similar research into the indicator but have only Other time periods such as a rolling 5 year or 10 year peri- discussed the results internal to the firms where I have od gave results in roughly the same order. worked. What does this tell us? That the risk/reward of the strate- First, I think it is only sensible to look at what happens in gy is very poor. In high return years, often a lot of the gain the following 11 months for the simple reason that the indi- comes in January. However, when the indicator is wrong, it cator should indicate the likely future and not include what is often very costly to be on the wrong side or sidelines. has already passed. After all, investing can not be done For example, in 2003, the opportunity loss was nearly 30%. with hindsight. The high success ratio for the indicator predicting direction As you show in your article, 31 out of 44 years is still a is more a function of the upward drift of equities overall. good strike rate for an indicator (or 29 out of 44 if the 1% It is noticeable that the barometer was at its least effective threshold is applied). However, when you link the indicator in periods when the market trended sideways or downwards to the returns the picture changes completely. Test straight- (1966-1979, only 7 out of 14 successes, and 1930-1948, forward strategies such as (1) going long for the remaining only 10 out of 19 successes). 11 months when January was up and staying out when Conclusion. Don't use the January barometer. It is worse January was down, and (2) going long for the remaining 11 than a simple buy-and-hold strategy most of the time. months when January was up and going short when January Appreciate that when the indicator is likely to be wrong, it was down. Neither strategy comes anywhere close to beat- can be very wrong since this likely indicates a change in ing a simple buy-and-hold strategy of being 100% long trend and consensus about the market. Maybe there is throughout (i.e. 100% long every February-December), over worthwhile work to be done exploring this idea. any time period, with or without a threshold. For example, from 1960 to 2003, a buy-and-hold strategy Simon Donne, Japanese equity long/short fund, returned 1439%. Following strategy (1) (i.e. long/flat) Invicta Investment Management LLP

SIR: I would like to draw your readers' attention to a subject neuronal systems. Then we came to technical analysis meth- that appears to receive little coverage in your publication - ods - attacked by many who argue they have no real scien- artificial intelligence (AI). This relatively new discipline, tific basis. After many years of investigation, we found the which has emerged from the progress of mathematics and following: computer sciences, has immediate relevance and application Our neuronal systems fell in love with classic technical for technical analysts. More specifically, there are two pow- analysis. The weights assigned to some of its methods range erful AI tools which are already being used for forecasting upon the higher ones. The genes that codify them frequent- in the financial markets: neural networks and genetic algo- ly become dominant ones. One discipline shines over all rithms. and deserves a special mention: Elliott wave analysis. It was The link between AI tools and technical analysis is aston- the most difficult method to program. It took us more than ishingly natural, making the application of new computa- a year to show our systems how to count waves and inter- tion methods to the financial markets inevitable. Any kind pret them. To program something into a computer, you of algorithms can be used as the initial seeds for an evolu- must understand its mathematical nature completely. So we tionary process. We use neural networks as seeds and the went into Elliott theory and started translating its simple result is a genetic net, something that germinates, grows, rules into mathematical equations. We ended up with a set reproduces itself and eventually dies - something very near of mathematical theory, having defined an Elliott geometry, to being alive. an Elliott topology and an Elliott probability algebra. At the core of AI systems are algorithms. They are what Connections to chaos theory, fractals and the phase space neural networks and genetic programs optimize to solve a became increasingly evident. Once programmed, our neu- given problem. One by one, year by year, we programmed ronal systems adopted Elliott wave quickly as a key forecast- the most diverse range of algorithms into our neuronal sys- ing method. tems, based on: Quantum mechanics, decision theory, statis- tical methods and probability. Many of them proved to be useful in forecasting prices and were quickly accepted by the Mario Martin, chief programmer, Neuronal Systems

January/February 2005 THE TECHNICAL ANALYST 43 Commitments of Traders Report

COMMITMENTS OF TRADERS REPORT

6 January 2004 - 4 January 2005 Futures only (open interest) non-commercial net long positions and spot rates

10-year US Treasury Source: CBOT 5-year US Treasury Source: CBOT

300000 4.50 200000 Non-commercial net long 5.00 Spot Non-commercial net long Spot 150000 4.80 250000 4.60 4.00 100000

4.40 50000 200000

4.20 3.50 0

4.00 150000

-50000 3.80 3.00

-100000 100000 3.60

-150000 3.40 2.50 50000

-200000 3.20

-250000 3.00 0 2.00 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004

Dow Jones Industrial Average Source: CBOT Swiss franc Source: CME 50000 1.35 12000 11000 Non commercial net-long Non-commercial net long Spot 10000 Spot 10800 40000 1.30 8000

10600 30000 6000

10400 1.25 4000 20000

2000 10200 10000 1.20 0 10000

-2000 0 9800 1.15 -4000 -10000 9600 -6000

-20000 1.10 -8000 9400 06/01/2004 30/03/2004 22/06/2004 28/09/2004 21/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004

Pound sterling Source: CME Yen Source: CME

45000 2.00 80000 116 Non-commercial net long Non-commercial net long Spot 40000 Spot 114 1.95 60000 35000

112 30000

1.90 40000 25000 110

20000 1.85 20000 108 15000

10000 106 1.80 0

5000 104

0 1.75 -20000 102 -5000

-10000 1.70 -40000 100 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 44 THE TECHNICAL ANALYST January/February 2005 Commitments of Traders Report

Euro Source: CME 3-month eurodollar Source: CME

10000 120 600000 3.00 Non-commercial net long Spot Non-commercial net long Spot 118 400000 5000 2.50

116 200000 0 2.00 114

-5000 112 1.50

-200000 110 -10000 1.00 -400000 108

-15000 0.50 106 -600000

-20000 104 -800000 0.00 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004

Nasdaq Source: CME Nikkei Source: CME

15000 2400 10000 12500

Non-commercial net long Non-commercial net long Spot Spot 10000 8000 2200 12000

5000 6000 2000 11500 0 4000

1800

-5000 2000 11000

1600

-10000 0 10500 1400 -15000 -2000

10000 1200 -20000 -4000

-25000 1000 -6000 9500 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004

Gold Source: CEI US dollar index Source: NYCE

160000 480 10000 120 Non-commercial net long Non-commercial net long Spot Spot 460 140000 118

5000 440 116 120000

420 114 100000 0 400 112

80000

380 110 -5000 60000 360 108

40000 340 106 -10000

20000 320 104

0 300 -15000 102 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004 06/01/2004 30/03/2004 22/06/2004 14/09/2004 07/12/2004

January/February 2005 THE TECHNICAL ANALYST 45 Long-Term Technicals

LONG-TERM TECHNICALS

Provided by Thomas Anthonj, ABN Amro, Amsterdam

EUR-USD USD-JPY

Stalling around a projected Fibonacci-target for a 5th wave top Coming close to the old low at 101.25 the market started to hesitate, (1.3650) the market could have started a setback with a minimum but this looks like nothing but profit taking so far. As long as the mar- down-potential of 1.1760 (bottom of wave 4). But showing several ket doesn't exceed 106.78/107.03, though we expect renewed weak- subcounts within the latest advance coming from the 1.1950 handle, ness and a conclusive break below 101.25. Thereafter the downside we would only be confident of having completed a bigger 5-wave would be pretty open towards the head-and-shoulders target at cycle up once the market breaks decisively below a strong support 95.75. A break above 107.03 would on the other hand open further cluster between 1.3017 and 1.2928 (38.2 %/old top). Such a break upside potential towards 108.75/109.39. A break above the latter would give room to test 1.2470/61 (trend line/old top) which looks would more or less break the bears neck and give room to re-test like the last resort for the bulls. Above 1.2928 we expect further neckline resistance at 114.83. gains towards 1.3804/1.3921 (Fib. projections) and 1.4160 (historic top) thereafter. GBP-USD Brent Crude

Stalling right under the projected target zone for a potential 5th wave The key-reversal week down at the end of October triggered a top (1.9566-88) and breaking below the last major top at 1.8771, we stronger setback than expected. But as long as the market remains could indeed be due for a much bigger setback. But unless the mar- above key-support between 35.50 and 33.94 we don't see any ket breaks below trend line support at 1.8378 and 1.8236 on weekly strong evidence that the whole up-trend has reversed. Above, we close we have no confirmation for a reversal yet and would still expect a minimum rise to 48.42 (76.4 % of the last decline). This has expect a resumption of the up-trend shortly. Having said that, the to be cleared in order to call for a straight resumption of the up-trend next targets to focus on would be at 1.9753 with massive resistance to trend channel resistance at 53.87 and the calculated target zone between 2.0115 and 2.0165 (old top/Fib. projections) which should for this 3rd wave impulse at 56.52-58.64. A failure to clear 48.42 definitely cap the market for a while. would call for another corrective leg down to the 37/36 handle.

46 THE TECHNICAL ANALYST January/February 2005 Long-Term Technicals

Dow Jones S&P 500

Stalling right at the head-and-shoulders target (10860) the market is Stalling right inbetween a Fibonacci-target cluster for a potential 5th in danger of having completed a first 5-wave cycle up. Nevertheless, wave up at 1211/26, the market is definitely in danger of having com- we are currently running a significant risk of performing a 2nd wave pleted its accumulation phase. But unless it closes the week below setback that usually retraces 61.8 % or 76.4 % of the whole advance trendline support at 1125, we have no evidence at hand to argue for from 7197 to 10868. However, it would take a weekly close below a bigger setback yet. To at least delay the permanent risk of a bigger trend line support at 9984 to give room for a decline to 8598 and setback (wave 2) the market would have to clear strong resistance potentially even 8063/62. A straight break above 10868 would on the between 1246 and 1308. A close below trendline support would be other hand give room to extend the upside to 11350 and 11750 (old really worrisome and could trigger a setback to 954/40 (old top/61.8 tops) from where we again expect a bigger setback. %) or to 878/75 (trendline/76.4 %) at the worst.

Nasdaq Nikkei

Breaking the row of higher lows in the daily chart and showing a 5- The market managed to break the row of lower tops at 11143 recent- wave structure up from the 1108 low that already came pretty close ly so we could see a test of key-resistance between 11648 (61.8 %) to the last major top at 2328, we clearly have to expect a bigger 2nd and 11789/11969 (76.4 %/old top). Only a break above the latter wave setback that could easily erase 61.8 % or 76.4 % of the gains would eliminate the threat of missing the left shoulder of a bigger since October 02. Having said that, it would still take a weekly close inverted head-and-shoulders pattern down to around 9383. A break below trendline support at 1870 to trigger a bigger decline towards below triangle support at 10637 would point in that direction. 1522 (61.8 %) or 1387/64 (old low/76.4 %) to potentially form the right shoulder of a bigger inverted head-and-shoulders reversal pattern. Only a break above 2328 would re-open the upside for substantial gains again. January/February 2005 THE TECHNICAL ANALYST 47 Training and Events Diary

TRAINING AND EVENTS DIARY

JAN - MAR FEBRUARY FEBRUARY 11-22 9 22-23 Course: Event: Course: STA diploma course STA Meeting Bond Investors Congress Organiser: Organiser: Organiser: Society of Technical Analysts Society of Technical Analysts Euromoney Conferences Contact: Contact: Info: [email protected] [email protected] www.bondcongress.com

FEBRUARY MARCH MARCH 28 17/18 23 Course: Course: Course: Introduction to technical analysis Technical analysis and charting Advanced Technical Analysis Organiser: Organiser: Organiser: Quorum Training Chartwatch 7City Contact: Contact: Contact: [email protected] [email protected] [email protected]

SUBMISSIONS FOR EVENTS & COURSES IN 2005

Please email us at: [email protected]

48 THE TECHNICAL ANALYST January/February 2005