Professional Documents
Culture Documents
The CFA helped me get my job, the CMT helped me keep my job.
Craig Johnson, CMT, CFA Piper Jaffray
Every step of my career: working as an analyst, publishing my book,
becoming an instructor, can all be traced back to the MTA and the CMT Program.
Hima Reddy, CMT New York Institute of Finance
646-652-3300
go.mta.org/cmt
TABLE OF CONTENTS
LETTER FROM EDITOR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
AN INTERMARkET AppROACh TO BETA ROTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Charles V. Bilello, CMT and Michael A. Gayed, CFA
ThEREpEATINgSTORyOFONBALANCEvOLuME. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
George A. Schade, CMT
EDITORIAL BOARD
Professor, Finance
St. Marys University
President
Hood River Research Inc. & TSSB Software.com
President
Kase and Company, Inc.
Portfolio Manager
Partner
Northington Dahlberg Research
Publisher
Market Technicians Association, Inc.
61 Broadway, Suite 514
New York, New York 10006
646-652-3300
www.mta.org
Journal of Technical Analysis is published by the Market Technicians Association, Inc., (MTA) 61 Broadway, Suite
514, New York, NY 10006. Its purpose is to promote the investigation and analysis of the price and volume
practicitioner) and libraries in the United States, Canada, and several other countries in Europe and Asia. Journal of
Technical Analysis is copyrighted by the Market Technicians Association and registered with the Library of Congress.
All rights are reserved.
5IF.BSLFU5FDIOJDJBOT"TTPDJBUJPO&EVDBUJPOBM
'PVOEBUJPO .5"&'
XBTFTUBCMJTIFE*OBTB
OPOQSPUPSHBOJ[BUJPOUPDSFBUFBOEGVOE
FEVDBUJPOBMQSPHSBNTJOUIFFMEPG5FDIOJDBM
"OBMZTJT0WFSUIFZFBST
UIF.5"&'IBTCFFOBCMFUP
DBSFGVMMZEFWFMPQBDPNQSFIFOTJWF5FDIOJDBM"OBMZTJT
DVSSJDVMVNUIBUJTCFJOHUBVHIUJODPMMFHFTBOE
VOJWFSTJUJFTBSPVOEUIFDPVOUSZ
VOJ
"4"6/*7&34*5:
(BJOBDDFTTUPBDPNQMFUFMFDUVSF5FDIOJDBM"OBMZTJTDPVSTFJODMVEJOHBMM
DIBSUT
FYIJCJUT
FYBNT
BOEBDPNQSFIFOTJWFHMPTTBSZPGUFSNTVTFEJO
5FDIOJDBM"OBMZTJT4VSSPVOEZPVSTFMGXJUIBTVQQPSUUFBNPGOBODJBM
QSPGFTTJPOBMTBOETFUZPVSTFMGBQBSUGSPNPUIFSOBODFQSPHSBNT
"4"456%&/5
(BJOBOJNQPSUBOUDPNQFUJUJWFFEHFCFGPSFHSBEVBUJOHDPMMFHFCZFYQPTJOH
ZPVSTFMGUPTPNFPGUIFJOEVTUSZTUPQOBODJBMUFDIOJDJBOTBOEHBJO
PQQPSUVOJUJFTUPFBSOTDIPMBSTIJQTGPSUIFDPWFUFE$IBSUFSFE.BSLFU
5FDIOJDJBO $.5
&YBNTUISPVHIUIF.5"
MFBSONPSFBCPVUIPXZPVDBOHFUJOWPMWFEBUNUBFGPSH
bioGraPhy
Charles V. Bilello, CMT
Charlie Bilello, winner of the 2014 Charles H. Dow Award and 3rd Place Winner of the 2014 Wagner Award, is the Director of Research
themes and portfolio positioning to clients. Prior to joining Pension Partners, he was the Managing Member of Momentum Global Advisors,
As winner of the 2014 Charles H. Dow Award and 3rd Place Winner of the 2014 Wagner Award, Michael helps to structure portfolios
to best take advantage of various strategies designed to maximize the amount of time and capital spent in potentially outperforming
investments. Prior to this Chief Investment Strategist role, Michael served as a Portfolio Manager for a large international investment group,
to structure client portfolios. In 2007, he launched his own long/short hedge fund, using various trading strategies focused on taking
advantage of stock market anomalies.
inTroducTion
Buy and hold is often touted as the ultimate investment strategy when it comes
to stock market investing. The reasoning for this relates to the belief in the
Efficient Market Hypothesis, which states that because all known information
is factored into price, there is largely no edge to active and dynamic trading.
Indeed, numerous studies have documented the inability of investment
managers bench marked to a market average to consistently outperform
passive strategies through stock selection.1 However, academic studies have
also noted persistent anomalies and phenomena in the marketplace which are
consistent and exploitable, putting the Efficient Market Hypothesis in doubt.2
Many of these studies focus on momentum and seasonality, and tend to be of
intense interest for technical traders.
1
2
strategy today using Exchange Traded Funds (ETFs) as the vehicle of choice.
Our findings are consistent with other studies which reference sector momentum
and the gradual diffusion of information across and within markets, a major
component of intermarket analysis. However, to our knowledge, no study has
yet to show quantifiably how to outperform the stock market through Utilities
rotation over time, nor has explored the signaling power of low beta leadership as
a leading indicator of heightened volatility.
liTeraTure
The idea that one can generate excess returns through defensive beta rotation is
not new. The concept is appealing in that it is intuitive to position into lower beta,
non-cyclical sectors during corrections, recessions, and bear markets, and rotate
to higher beta and more cyclically-sensitive sectors in favorable economic and
market environments.
However, some studies have put into question this approachs feasibility. Davis and
Philips (2007) argued that implementing a defensive investment strategy based
on the leading signals of bear markets and recessions (e.g., forward price/earnings
ratios, momentum indicators, and the shape of the U.S. Treasury yield curve)
would not have resulted in better results than following a buy-and-hold strategy.
However, the strategy assumptions made in this study are entirely different than
our suggested approach. Davis and Phillips used macro cyclical indicators (such
as the yield curve), valuation (such as forward P/E), and an arbitrary definition
of momentum (a 5% or 10% drop in the market over the trailing 12-months) as
their risk triggers. These assumptions are quite different than our approach, which
purely focuses on the relative price momentum of the Utilities sector and over a
much shorter time frame.
Momentum is a well-documented characteristic of markets, through both
individual stock movement over longer time periods and in the persistence of
sector strength in short-term time periods. Moskowitz and Grinblatt (1999)
note that unlike individual stock momentum, industry momentum is strongest
in the short-term (at the one-month horizon), and then, like individual stock
momentum, tends to dissipate after 12 months, eventually reversing at long
horizons. The specific time-frame of momentum drift at the one-month horizon
may be due to large-cap stocks leading small-cap stocks within a sector, and
because weekly portfolio returns are strongly positively auto correlated as
documented by Lo and MacKinlay (1990). As information by market leaders
gradually diffuses down to smaller competitors, investors act with a lag in trading
such companies, causing the aggregate to continue in its prior direction.
3
The Market Technicians Association (MTA), founded in 1973, is a not-for-profit professional regulatory organization servicing over 4,500 market analysis professionals in over 85
countries around the globe. See www.mta.org.
4
See Wilkinson (1997).
5
See Print (2002).
6
See Gould (1974).
The sTraTeGy
Edson Gould largely focused on the Dow Utilities relative to the Dow Industrials
because that was the most readily available dataset for him to make his forecasts.
However, because Dow indices are price-weighted, it stands to reason that a
more comprehensive data set should be used to not only include more stocks,
but also to more appropriately weight companies based on capitalization.
In addition, the Dow Utilities and Dow Industrial averages are not total return
indices. Dividend yield information on Dow averages is limited, but clearly has
a significant impact on investor wealth. Clarke and Statman (2000) estimated
that if a total return calculation were done using dividend approximations, the
Dow Jones Industrial Average in 1998 would have been 652,230 versus 9,181
for capital appreciation alone. Since dividends have such a large impact over
time through compounding, and because the Utilities sector tends to have a
higher dividend yield than the market average itself, a true strategy must
incorporate total return8 data.
Using data provided by Fama-French resolves both of these issues by being
market weighted and total return. Using the Fama-French price data going
back to July 1926, we developed a simple trading strategy:
When a price ratio (or the relative strength) of the Utilities sector to the
broad market is positive over the prior 4-week period, position into Utilities
for the following week. When a price ratio (or the relative strength) of the
Utilities sector to the broad market is negative over the prior 4-week period,
position into the broad market for the following week.
The basis for using the 4-week rate-of-change interval is the research
illustrating monthly momentum among industry groups9 In order to achieve
a more tactical strategy that is better able to adapt to intra-month volatility,
we converted the monthly time frame into a weekly signal. We have named
the approach the Beta Rotation Strategy (BRS) as it attempts to rotate into
Utilities when the investing environment is more favorable towards lower-beta
Source: http://www.citizen.org/cmep/energy_enviro_nuclear/electricity/deregulation/puhca/
Source: http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html
9
See Moskowitz and Grinblatt (1999).
7
8
2014 . Issue 68
equities and into the market when the investing environment is more favorable
towards higher-beta equities. Such a rotation translates the classic intermarket
relationship of Utilities relative strength as a leading indicator of market cycles
into an actual trading strategy.
Using the weekly signal from July 1926 through July 2013, the BRS shows
significant outperformance versus a buy-and-hold portfolio of both the
market and the Utilities sector. As illustrated in Chart 1 below, a $10,000 initial
investment in the strategy in July 1926 grows to $877 million in July 2013
versus $34 million for the market and $17 million for the Utilities sector.10
Second, in Table 2, we break down the performance of the BRS by decade. Here
too we observe outperformance in all decades, with some normal variability.
Importantly, we see the greatest outperformance during periods of market
turmoil (1926-1929, 1970-1979, and 2000-2009). We discuss this finding in
further detail in the volatility signal section below.
Third, we evaluate the rolling 3-year outperformance of the BRS to test the consistency
of the outperformance over a shorter time frame that is more in line with how many
institutional investors judge investment performance. Chart 2 illustrates consistent
outperformance during the overwhelming majority of 3-year time periods. Overall,
the BRS outperforms the market in 82% of rolling 3-year periods.
First, in Table 1, we break down the performance into various time periods
around significant legislative events for Utilities. This is important as one
could argue that the behavior of Utilities has changed over time, making the
signal more or less powerful. What we find is the outperformance of the BRS is
observable in all time periods. While the performance of the BRS did improve
following the Public Utility Holding Act of 1935, the remaining time periods
showed consistent outperformance of over 4% per year.11
10
say the least. Additionally, a strategy based on weekly positioning would have
been prohibitively expensive due to slippage and commission costs, particularly
in the pre-May Day 1975 era.12 However, just because it may not have been
possible to follow through on such a strategy in the past does not mean one
should disregard the relative strength of Utilities and its predictive power.
The lead-lag behavior of the Utilities sector can be a critical warning sign of
higher average volatility to come in the market, and can be an early tell of
whether the odds of an extreme tail event are rising.
In proving this thesis, we first examine the volatility of the market when the
BRS is in Utilities (Utilities are leading) and compare that to the volatility of the
market when the BRS is in the market (Utilities are lagging). If Utilities relative
strength is predictive of higher volatility, then we should see higher volatility for
the market when Utilities are leading and lower volatility for the market when
Utilities are lagging.
Lastly, we tested a long/short version of the BRS. When the BRS calls for a
rotation into Utilities, the long/short strategy goes long the Utilities sector and
short the market. Conversely, when the BRS calls for a rotation into the market,
the strategy goes long the market and short the Utilities sector. Table 4 shows
that the long/short strategy produced consistently positive annualized returns
over time.
12
To test whether Utilities leadership was predictive of higher volatility over shorter
time periods, we then measured the percentage of time the BRS was in Utilities
during periods of market stress.
The assumptions used in this section are no slippage or commission (more on this later).
2014 . Issue 68
11
First, in Table 6, we looked at the worst weekly declines for the market since 1926.
We found that during those weeks, the BRS was in Utilities for a significantly
higher percentage of time than overall, suggesting that Utilities strength is a
leading indicator of market volatility. In the worst 2% of weeks (declines > 5.5%)
in history, the BRS was in Utilities 58.9% of the time versus 49.7% overall.
Next, in Table 7, we looked at both the highest levels of VIX values and greatest
spikes in history. While the data set for the VIX only dates back to 1990, we find
similar results to Table 6. The BRS strategy was positioned in Utilities at a much
higher rate during periods of market stress than overall.
This is important as it may provide an additional clue as to why sell in May has
persisted over the years. If, as argued by Jacobson and Visaltanachoti (2006),
Utilities show the least differentiation among sectors between summer and
winter months, then it stands to reason that deviations in their relative strength
would have more predictive power than other sectors. Given that Utilities are
the most bond-like sector of the market, changes in interest rates are likely a
driving force behind these deviations.
13
12
We indeed see this as interest rates have tended to fall during the May through
October period and rise during the November through April period (see Table 9).
This confirms the linkage between seasonal strength in bonds (falling yields)
and more time spent in Utilities during the worst six months, consistent with
the aforementioned impact of rates on sector movement.
Finally, we need to address how the BRS compares to a simpler rotation into
Utilities in May and into the market in November. As Table 10 illustrates, the
BRS shows 3.0% outperformance vs. this strategy overall and outperformance
during both summer (1.3%) and winter (1.5%). These results indicate that the
power of Utilities to detect periods of market stress in all time periods, including
the seasonally strong winter months, outweighs a strategy of simply avoiding
stress during the summer months.
versus 94% for XLU and 96% for VTI. If we make an assumption of 0.1% per trade
for commission and slippage, the ETF Strategy still shows outperformance vs. XLU
and VTI. More importantly, on a risk-adjusted basis it also is superior, with lower
annualized volatility than both XLU and VTI over time.
conclusion
We find that the signaling power of the Utilities sector is a market anomaly that has
persisted over time. The Utilities sector has less economic sensitivity and is more
dependent upon the cost of capital than other sectors. Therefore, fluctuations
in its relative price movement can have broad implications on macroeconomic
factors. Yet, contrary to the Efficient Market Hypothesis, the information that
lead-lag dynamics may have about the near-term future may not be fully priced
in immediately by broad market averages. This lagged reaction is precisely what
makes their strength and weakness exploitable.
The implications from both a strategy and signaling standpoint are meaningful.
We find that by using a Beta Rotation Strategy based on the principles of
intermarket analysis and momentum, one could have consistently outperformed
a static buy and hold strategy over many market cycles.
Outperformance is achieved by timing exposure to beta using a rolling 4-week
relative strength signal of the Utilities sector to the market. The strategy rotates
into Utilities when the investing environment is more favorable towards lowerbeta equities and into the market when the investing environment is more
favorable towards higher-beta equities. Importantly, because the Beta Rotation
Strategy spends roughly half of its time in Utilities, it is also able to benefit from the
compounding effect of higher dividends. We observe consistent outperformance
in the vast majority of periods, and that the rotation signal may offer further
insights into explaining seasonal patterns such as sell in May and go away.
2014 . Issue 68
13
We also find that the strength in Utilities often serves as a warning sign of
increased volatility and extreme market movement in the short-term, allowing
active traders to better manage risk during potential periods of heightened
market stress. This important finding can add to ones trading arsenal in seeking
higher risk-adjusted returns, and in reducing the possibility of tail events by simply
respecting intermarket relationships and price history.
furTher research
Although beyond the scope of this paper, there are a number of broader
implications that our findings may have on the investing and trading landscape,
particularly as it relates to volatility. Among these are: (1) implementing option
overlay strategies, (2) hedging, (3) timing of gross exposure or leverage, and (4)
tactical asset allocation.
references
Clarke, Roger G., and Meir Statman, 2000, The DJIA Cross 652,230 in 1998, The
Journal of Portfolio Management.
Davis, Joseph H., and Christopher B. Philips, 2007, Defensive Equity Investing:
Appealing Theory, Disappointing Reality, Vanguard Investment Counseling &
Research.
Day, Wang, and Xu, 2001, Investigating Underperformance by Mutual Fund
Portfolios, School of Management, The University of Texas at Dallas.
Gould, Edson, 1974, The Utility Barometer: An Early Stock Market Indicator,
Edson Goulds Findings & Forecasts Newsletter.
Hong, Torous, and Valkanov, 2005, Do Industries Lead Stock Markets? Journal of
Financial Economics.
Jacobson, Ben, and Nuttawat Visaltanachoti, 2006, The Halloween Effect in US
Sectors, The Financial Review.
Kamstra, Kramer, and Levi, 2003, Winter Blues: A SAD Stock Market Cycle,
Federal Reserve Bank of Atlanta Working Paper No. 2002-13a.
Lo, Andrew W and A. Craig MacKinlay, 1989, When Are Contrarian Profits Due to
Stock Market Overreaction? The Review of Financial Studies.
Moskowitz, Tobias J. and Grinblatt, Mark, 1999, Do Industries Explain
Momentum? The Journal of Finance.
Murphy, John J., 2004, Intermarket Analysis, John Wiley & Sons, Inc.
Murphy, John J, 1999, Technical Analysis of the Financial Markets: A
Comprehensive Guide to Trading Methods and Applications, New York Institute
of Finance.
Pring, Martin J., 2004, Technical Analysis Explained, McGraw-Hill.
Russell, Philip S. and Violet M. Torbey, 2002, The Efficient Market Hypothesis: A
Survey, Business Quest Journal.
Wilkinson, Chris, 1997, Technically Speaking: Tips and Strategies from 16 Top
Analysts, Traders Press, Inc.
14
BIOGRAPHY
George A. Schade, CMT
George A. Schade, Jr., CMT has researched the origins and background of technical indicators and their creators. He has written on the
stochastic, advance-decline, and on balance volume (OBV) indicators. He received the 2013 Charles H. Dow Award for his paper on
the development of OBV. He contributed an extensive biography of Ralph Nelson Elliott compiled over several years. He coordinated
two of the largest collections of historical materials given to the Market Technicians Association. His objective is to expand and clarify the
historical record of technical analysis. Since becoming a member of the MTA in 1987, he has served on numerous association committees.
THE ExPRESSION
THE CALCULATION
Daily OBV is calculated by adding the days volume to a cumulative total when
the securitys price closes up, and subtracting the days volume when the
securitys price closes down.4 All the volume is assigned to correspond to the
direction of the days close. If price does not change from the prior days close,
yesterdays total OBV is recorded. The cumulative count can begin at a baseline
index in order to avoid negative numbers.
The term on balance means the net effect or result after considering or
offsetting all relevant factors. The expression has a long history to describe the
state of trade.
THE THEORY
On May 24, 1893, The New York Times reported that loans due had been paid
and, in recent days, the Bank of England has received 466,000 of foreign gold
on balance. 1 On October 18, 1908, The New York Times reported that another
Every price change is the result of a sale or purchase of shares of stock. The
number of shares involved in a transaction constitutes volume.
It is not price action, but volume - the amount of money, the supply and the
demand - which best tells the story. Humphrey B. Neill (1931)5
OBV is a momentum indicator that relates volume to the direction of price. Its
basic assumption is that volume precedes price, or as Granville posited in 1963,
volume often has a distinct tendency to precede price.6 The assumption is
grounded on the dynamic shifts of supply and demand for a security.
Writing in 1935, Harold M. Gartley, arguably the Eras finest market technician,
suggested why we study volume: Theoretically, the reason we study volume is
because it is believed that it is a measure of supply of and demand for shares.
(italics in original)7 A shift in the supply or demand for a stock will change the
stocks volume. In 1984, Granville wrote that:
Price will rise only after the volume equation is thrown out of balance by
quietly increased demand. Conversely, when heavier, silent selling occurs,
supply will overcome demand, and only then will price fall. In either case the
alteration in supply and demand must take place before the move in price. 8
Frank Vignola and wife Maude Vignola Woods, who in 1951 asserted that
buying and selling volume are best analyzed when measured cumulatively,
explained their Continuous Volume Curve, as they called OBV, as follows:
The CONTINUOUS VOLUME CURVE is. [e]xtremely sensitive to price
movement, and will indicate the relative balance between buying and selling
at the peaks and valleys of market trends. 9
The assumption that volume often precedes price was expressed in 1934.
The Wetsel Market Bureau, Inc. (Wetsel) was a subscription market advisory
service (Humphrey B. Neill was its Vice-President in 1931). In 1934, Wetsel
published a 26-lesson trading course. The discussion of volume and prices
relationship in Lesson 20 stated that volume tends to lead the price movement
and it is in this respect that volume may constitute a forecaster. 10
Neill (1895 - 1977) wrote in 1931 that volume will give you indications
of pending moves, often when nothing else will.11 Thirty-two years later,
Granville wrote that on-balance volume can be a particularly effective early
warning of future price movements.12
According to Wetsel,We have seen individual occasions when price would hardly
tell you anything about probable future price movement and where volume
alone would contribute possibly as much as 85% or 90% to the importance of
price and volume combined as indicators.13 Long ago, technicians recognized
that volume often precedes price is a valid working assumption.
16
1. A volume index number made by giving the sign of the price movement to
the daily volumes, and accumulating the plus and minus movements
On April 17, 1925, Clay spoke at an ASA meeting on a panel announced days
earlier:
The psycho-technical index built out of these five elements looks much like a
price chart with the false movements eliminated.
It has the very distinct merit of often moving contrary to the course of the market
itself. This index is not used independently, but rather in conjunction with the
economic indexes which formerly constituted the chief reliance of Mr. Clay. 15
Clay used OBV as one element of an index rather than as an independent
indicator. In spite of intense research, additional information about Clays index
has not been found.
Clay was a prominent economist, statistician, and investment counselor in the
years prior to 1945. From 1912/1913 to 1927, he worked at Moodys Investors
Services as an economist and statistician, rising to Vice President and Chief
Economist. In 1915, he wrote a 371-page book entitled Sound Investing.16
Moodys published a revised edition in 1920. In 1916, Clay penned an article
for Moodys Magazine - How and When to Buy and Sell - in which he proffered
detailed rules for analyzing accumulation and distribution as the really
important subject of this article is the when. (emphasis in original) 17
From August 1919, to May 1922, he was Staff Economist and columnist for
Forbes Magazine. In November 1920, Forbes published Clays column which it
regards as one of the most important articles it has ever published. 18
In 1927, Clay was retained as an expert witness on the valuation of stock by a
group of founding Ford Motor Company shareholders who were contesting a
$30 million income tax assessment on a stock sale. Clay spent February 7, 1927,
giving extremely technical testimony under forceful cross-examination.19 The
shareholders group won.
In the Thirties, Clay was an economist for the United States Shares Corporation
and Supervised Shares Corporation, investment counsel and trust companies,
and in his consultancy Clays Economic Service. In August 1931, he was elected
director and economic adviser of the General Shares Corporation. In October
1934, he became Editor of Brookmire Bulletins, published by Brookmire
Economic Service, a national business forecasting firm established in 1910.
Clay was among the first 605 applicants approved by the Securities and
Exchange Commission as registered investment advisers under the Investment
Advisers Act of 1940 in New York, New Jersey, and Connecticut.20 In this group
were renowned technicians Ralph Nelson Elliott, William D. Gann, and Harold
M. Gartley.
2014 . Issue 68
17
At the 1932 ASA meeting, Clay conceded that the rules which he believed to be
dependable, based on his years of research concerning the cyclical movements
of prices, had all between 1928 and 1931 broke down.22 He concluded that
this breakdown was caused by the fact that the rules were applicable to some
eras but not to others. 23 Clay realized that his former rules had broken down
following the Great Crash of 1929, but the fermenting vigor of technical studies
was opening new frontiers.
Volume, in the early Thirties, became a fertile field for analysis. The increasing
volume of stock trading up through 1929 made volume worthy of study.
Clay, ardent economist and business statistician, saw value in technical concepts
such as OBV. By 1932, he had changed some of his strong views expressed at
the meeting held in April 1925, where he had remarked he rejected using an
automatic barometer or combining certain barometrical returns, [to] obtain
a barometer or index number which is supposed to move ahead of the stock
market and indicate its course. 24
Clays analytical work, grounded in business statistics, was shaped by the 1929
Crash as well as by the post-Crash Emergent Age of technical analysis. The
psycho-technical index showed his interest in technical analysis and crowd
psychology. Clay was thinking technically.
Although little is known about Clays index, his formulation of a cumulative
volume indicator impresses when one considers the then prevailing demands
of studying volume. Harold M. Gartley (1899 - 1972) described them in 1932.
Between September 19, 1932, and December 5, 1932, Gartley wrote twelve
articles published in Barrons National Financial Weekly headlined Analyzing the
Stock Market. The next year Gartley compiled the articles in a course entitled
Stock Market Studies. The course led to Gartleys seminal book Profits in the Stock
Market: The Gartley Course of Stock Market Instruction published in 1935.
Barrons announced that from source material supplied by one of the bestequipped laboratories in Wall Street, the articles will present the most
modern work on the interpretation of security-price movements.25 In
November 1932, Barrons featured Gartleys landmark article The Significance
of Volume of Trading.
Gartley wrote that the accuracy and completeness of volume data were
problematic. Analysts faced the inability to obtain essential data without an
almost prohibitive amount of abstraction from the official sheets which list
all of the transactions on the Stock Exchange, and compilation, as well as
that in the past few years, notably from 1928 to date, the number of shares
18
Wetsels course does not contain a calculation similar to that of OBV, but it
explains how volume tends to lead price, OBVs basic assumption. Alphier was
likely referring to the explanation of this conceptual assumption rather than to
an explicit OBV calculation that is not evident in Wetsels course.
Volume is the pressure gauge for measuring the balance between supply and
demand, and for determining the quality of buying and selling in a stock or market
Average. VOLUME CAN BE ANALYZED TO BETTER ADVANTAGE when data is
arranged in a time series or on a cumulative basis. (emphasis in original)34
Vignola used three series to analyze volume which he integrated with a 10-day
moving aggregate of daily price changes in stocks or market indices called the
Price-Curve.The first series was a 10-day moving total of aggregate volume called
the Aggregate Volume Curve. Saturdays volume was doubled to account for the
2014 . Issue 68
19
2
short session. The second was a 30-day moving total of aggregate volume named
the Major Volume Curve. These are time based series, but Vignolas third series
differentiated between buying and selling volume.
The third series was the Continuous Volume Curve which is made by adding the
total daily market volume of trading to a base index figure, each day the market
advances; and by subtracting the volume on days when the market declines.35
Saturdays volume was not doubled in this series. Vignola suggested a base number
of at least 50 or 100 million. He did not use the term cumulative volume.
Chart 2 shows the Dow Jones Industrials, 10-day Price-Curve, 10-day Aggregate
Volume Curve, and Continuous Volume Curve in low volume conditions which
Vignola claimed prevail in over 80% of all intermediate and major signals. Note
how trend line breaks triggered buy signals.
20
M. V. Woods published another course in 1955, but it did not further amplify
how to use continuous volume curves.
COMMODITY CONTRACT
PRICE
VOLUME
OPEN
INTEREST
= UP
Gotthelf believed his OBV method detected accumulation (open interest and
volume rise) and distribution (the reverse) which led to overbought or oversold
markets. Overextended markets, notably those following a long period of
accumulation, could experience dramatic corrections which in turn gave buy
and sell signals.
Gotthelfs method is more akin to a tabular score than to a cumulative count,
but it shows that the term OBV was used in futures trading more than fifteen
years before Granville wrote his OBV book in 1963.
CONCLUSION
Several people working in different decades and living thousands of miles from
each other conceived OBV - a remarkable and not uncommon story of creativity.
Recognition for OBVs invention must be shared.
Alphier opined, Im confident Granville came upon OBV independently, but I
have two courses in my library, one written in 1955 [Vignola] and one in 1934
[likely Wetsel], which use the concept. I also can be confident that Granvilles
two predecessors didnt have any contact with each other.45 Until proven
otherwise, the same must be said of Clay, the Vignolas, and Gotthelf.
We cannot shadow their brilliance, but these conclusions should not surprise.
The indicators simplicity and its sensible logic explain why several people (and
these are the ones we know) developed OBV from their own studies.
DATE
VALUE
While Granville popularized OBV, Paul Clay and clearly, Frank Vignola and Maude
Vignola Woods, had earlier originated OBV, while Edward B. Gotthelf used the
term. Volume was uniquely important to the analytical work of Granvilles
predecessors. They believed volume could presage the direction of price, and
a cumulative count was a valid way to analyze buying and selling volume. All
were intelligent observers of volume in stock and futures markets who merit
recognition.
= DOWN
Over time, the net plus and minus days would be counted. If on balance net
pluses outnumbered net minuses, Gotthelf would be a buyer. If minuses
exceeded pluses, he would sell. If net pluses and minuses were about even, he
2014 . Issue 68
21
ENDNOTES
N. Y. TIMES, May 24, 1893, On the London Exchange (all NYT citations are to
the digital archives).
Gould, Jr., Edson B., 1974(est.), A VITAL ANATOMY (Anametrics, Inc., New
York, NY), 1.
21
22
23
2
3
Neill, Humphrey B., 1931, TAPE READING AND MARKET TACTICS: THE THREE
STEPS TO SUCCESSFUL STOCK TRADING (B. C. Forbes Pub. Co., New York, NY),
repub. 1959, 1970 (Fraser Publ. Co., Burlington, VT), 41.
Granville, Joseph E., 1963, GRANVILLES NEW KEY TO STOCK MARKET PROFITS
(Prentice-Hall, Inc., Englewood Cliffs, NJ), 19. By 1984, Granville was firmer
asserting, Volume precedes price. Granville, Joseph E., 1984, THE BOOK OF
GRANVILLE (St. Martins Press, NY), 91.
Gartley, Harold M., 1935, repub. 1981, PROFITS IN THE STOCK MARKET: THE
GARTLEY COURSE OF STOCK MARKET INSTRUCTION (Lambert-Gann Pub. Co.,
Pomeroy, WA), 299.
Id., 317.
24
25
Gartley, Harold M., Nov. 7, 1932, Analyzing the Stock Market: The
Significance of Volume of Trading, BARRONS NATL. FNCL. WKLY., 22.
26
Id.
27
28
29
Library of Congress.
30
Alphier, James E., 1988, The tragic neglect of the old masters, TECHNICAL
ANALYSIS OF STOCKS AND COMMODITIES, vol. 6:10, 396.
31
THE MAGAZINE OF WALL STREET AND BUSINESS ANALYST, Mar. 30, 1946, vol.
77: 797.
8
9
Wetsel Market Bureau, Inc., 1934 U.S.A., repub. 1998, A COURSE IN TRADING
(Donald Mack ed., Financial Times Prof. Ltd., London, UK), 164.
10
11
12
13
14
King, Willford I., 1932, Forecasting Methods Successfully Used Since 1928,
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION, vol. 27, 179: 317. The
other elements were the price movements of the twenty stocks having the
largest daily volume, adding 1 to the index on up days and subtracting 1 on
down days, resistance ratios to measure liquidation and short sales, and
relative strength.
15
Clay, Paul, 1915, SOUND INVESTING (Moodys Magazine and Book Co., New
York, NY).
16
Clay, Paul, Nov. 27, 1920, Confidently Predicts Boom in 1921, FORBES, 117.
18
19
20
22
32
33
34
35
Id.
36
37
Id., 3-B:3.
38
39
Id., 1-C:3.
40
Gotthelf, Philip, Dec. 1986, The making of a system B.C. (before computers),
FUTURES, 57.
41
Id.
42
Id.
43
Id.
44
45
bioGraPhy
Carl Aspin
Carl Aspin recently retired as President of Aspin Engineering Services in Prescott, Az. He received a degree in Engineering Physics from the
Colorado School of Mines. This was followed by post graduate engineering training at General Electric in Utica. NY. He was employed
at Motorola for more than 30 years in a wide range of R&D, engineering, and management positions. He is a graduate of the Motorola/
absTracT
Stock and market index sequences are characterized by an original statistical
algorithm that identifies in real time, prices and price levels that have a
significant and recurrent influence on the magnitude and direction of price
fluctuations in their immediate neighborhood. These prices, called balanced
ladder values, provide a basis for the identification of trends, support and
resistance levels, channels, and bands. A unique off-balance volume sentiment
chart is described. The use of balanced ladder levels as a basis for standard
technical analysis is demonstrated through rule-based graphical analysis of
stock and market index examples.
i. inTroducTion
An ongoing tension exists between bullish and bearish psychology, forming
technical support and resistance levels. Price movements are driven by ever
changing market expectations and psychology, and a balance or relaxation
of the tension between the pressures of supply and demand can be detected
by means of probabilities and statistics. Consequently, on the occurrence of
a balance value, it becomes possible to detect the presence of support and
resistance barriers. Trends are observed as a drift in balance values between
bands and channels defined by support and resistance barriers.
means and a variance that is the sum of the variances of the two distributions.
Let V(T) denote the variance after T intervals. Then V(T) = V(1)*T. The increase
in the variance is directly proportional to the time. As the standard deviation is
equal to the square root of the variance it follows that = 1 T1/2. Therefore the
distribution of the sum of two distributions, each with mean zero and standard
deviation 1, is N{0, 1 21/2} and in general, the sum of T distributions, each
with mean zero and standard deviation 1, is N{0, 1 1/2} (Hull, 2000). Let:
Each of these values, Xt, is the difference in the natural logarithms of prices
separated by one unit interval. If T successive fluctuations are added, then a
record of the logarithms of the prices is generated from P(0) to P(T). Note that
and in general:
FIGURE 1
t = 1, , k.
24
FIGURE 2
A reason for this divergence is shown in Figure. 3. The market index fluctuations
do not satisfy the requirements for the addition of independent and identically
distributed normal distributions. We observe anomalous fluctuations that fall
well outside 3-sigma statistical limits, the red boundaries. These limits should
enclose 99.74% of all fluctuation data. A large number of these anomalous
fluctuations occurred during the recent market crises. The probability that
such fluctuations would ever occur if sampling was from a normal distribution
is vanishingly small. Anomalous fluctuations such as these produce what are
known as fat tails.
FIGURE 3
In an ideal bell-shaped distribution the probability that Z = ln[P(t+1)/P(t)]/
exceeds +/-3 approaches zero much more quickly than is the case when
anomalous fluctuations are present. In the presence of anomalous fluctuations the
areas to the left and right of -Z* and Z*, respectively, Figure 4, are larger than the
corresponding areas in a standard normal distribution, hence the term fat tails.
and thus desensitizes the controls. An analogous approach is used here with
market-based data. Although we may be able to infer the underlying causes
for anomalous market fluctuations we cannot eliminate the causative factors
in general. However, knowing that anomalous data is not part of the normal
historical reference distribution we can logically make adjustments so that the
subsequent reference distribution is not unduly perturbed. Accordingly, the
fluctuations that are outside of the 99.74% probability boundaries, Figure 3,
are iteratively scaled back in size until they fall within recalculated +/- 3 sigma
boundaries. The adjusted variogram of Figure 5 is the result. The approximation
to the Markov model is greatly improved. Hence, prior to the analysis of any
fluctuation sequences, anomalous fluctuations are rescaled as described. Note,
however, the original values in the price data sequence are not changed so
there is no change in the original price chart. It is recognized that although this
procedure brings the variogram into close agreement with the results expected
from a Markov process it does not mean that the revised set of fluctuations is
ideally normal or Gaussian.
FIGURE 5
FIGURE 4
2014 . Issue 68
25
According to Kahn (2006), support is that price level at which the aggressive
selling of the bears has waned sufficiently to be offset by the rising aggressiveness
of buying by the bulls. Conversely, resistance is the level at which the aggressive
buying of the bulls has waned sufficiently to be offset by the rising aggressiveness
of selling by the bears. The support level acts as a lower price barrier in a region
previously dominated by bearish activity while a resistance level acts as an upper
price barrier in a region previously dominated by bullish activity. Ladder values
are affected by the presence of support and resistance levels as these levels are
approached. Where the bulls and the bears offset one another we expect to find
a transient point of equilibrium or balance. From a technical analysts perspective
it is useful to be able to objectively determine when and where price barriers and
their associated balance points occur.
26
FIGURE 7
2014 . Issue 68
27
Next refer to Figure 10. After 4 bearish days the last close is now 8.75. The
descending ladder has continued to be active. During this interval the ascending
ladder has remained inactive at 10. Note that a gap has opened up between
the two maximum expected utility loci, i.e. between 9.02 and 9.42. With a
gap between the utility functions the next closing price must fall into one of
three probability zones: the upper bullish zone, the lower bearish zone, or the
intermediate zone where balance occurs. If the closing value is in the bullish
zone then the ascending ladder is active and adjusted. Similarly, if the closing
value is in the bearish zone then the descending ladder is active and adjusted.
If, however, the closing value falls into the intermediate gap then both ladder
values are active and a balance value results.
Referring again to Figure 10, if the close is less than 9.02, then prices remain
FIGURE 8
According to the example of Figure 8, given 1 = 0.03, the maximum expected
price for the next day is 10.30; the minimum expected price is 9.70. However,
suppose a loss of 5% occurs such that the next days close is at 9.50. This
represents a significant bearish event. Figure 9 shows the status at the close of
day one. The first fluctuation after the origin has produced two ladder values,
an upper ladder value at 10 and a lower ladder value at 9.50. Since the most
recently changed ladder value is the location of the ongoing Markov process the
active ladder is 9.50 while the origin at 10 is passive. At the active descending
ladder level there are new bullish and bearish reference loci, i.e., maximum
expected utility functions. Since the ladder at 10 is passive the original bullish
and bearish reference loci are unchanged. The bullish locus stemming from the
bullish ladder and the bearish locus stemming from the bearish ladder are not
needed in any subsequent discussion in this paper. Therefore they are not shown
in the graph as their continued representation causes unnecessary graphical
clutter. Hence, only two reference loci are displayed: the bearish maximum
utility function, ULx(t), originating from the ascending ladder, and the bullish
maximum utility function, UGx(t), originating from the descending ladder.
FIGURE 9
28
FIGURE 10
in the bearish zone. It is readily calculated that the probability of this event is
greater than 84%. If the close is greater than 9.42 then conditions have become
bullish. The probability of this happening at random is less than 1% and unlikely.
However, in case of this event the bullish ladder becomes active and is set equal
to the closing price. At the same time the bearish ladder becomes passive. If the
close does not fall in either the bullish or bearish zone then it falls in the gap
between 9.02 and 9.42. The probability of this happening at random is 14.9%.
In this case the two ladder values are logically equivalent, i.e., in balance, and
define a new origin. At this point the ladder generation process begins again
with the next fluctuation. It is obvious from symmetry considerations that if the
first fluctuation is positive or bullish the process will evolve in a similar fashion
to the one just described.
The maximum expected utility reference loci that separate bullish and bearish
zones are unique statistical boundaries that are analogous to control chart
probability zones used in industry to statistically detect unnatural patterns.
(Small, 1957). In particular cases the bullish and bearish zone boundaries may
be adjusted by the analyst in order to investigate the effects of the boundary
locations on the signal generation process.
Let St = 1n(Pt) define the origin of the ladder generating process by S0 and
subsequent ascending and descending ladder values by SAi and SDj, respectively.
The indices i and j are required to keep track of where changes to ladder values
occur. At the close of each trading day ladder values are compared to the closing
price and adjusted according to the rules described below.
The first two rules are immediately obvious. If todays close is equal to or greater
than the current ascending ladder value then change the ascending ladder
value to equal todays close. Similarly, if todays close is equal to or less than the
current descending ladder value then change the descending ladder value to
equal todays close. These rules, in conditional form, are as follows:
1/2
The factors m and n are square root of time functions. If there is no change in the
value of the ascending ladder, i.e., if the ladder is inactive, then the value of m is
increased by one unit for each inactive day. Similarly, if the descending ladder is
inactive, then the value of n is increased by one unit for each inactive day. In all
instances (i + m) = (j + n) = t where t is the current time coordinate. As long as
a ladder level is active its m or n factor remains equal to unity.
As described in the example, at the end of each business day the closing price
falls into one of the three zones: the upper bullish, the lower bearish, or in
between. The fluctuations of interest are those that result in prices falling in
the gap that forms between the two maximum expected utility boundaries. It
is assumed that in order for a fluctuation to result in a closing price falling in the
gap between the ladder probability zones then either bullish or bearish price
pressures have eased, causing prices to move in the direction of equilibrium. The
values that appear in the intermediate probability zone between the maximum
expected loci are the balance points. The algorithmic procedure described next
uniquely identifies balance values.
Suppose the logarithm of todays closing price, St, is less than the logarithm
of yesterdays ascending ladder value, SA(t-1). Then St is located in one of the
three probability zones. By definition, the ascending ladder value must remain
above the bearish probability zone, i.e., the zone bounded by the bullish
expected value locus with the descending ladder value as origin. Thus, in order
to change the ascending ladder value St must be greater than SD(j+n) = SD(j)
+ 1n1/2. Otherwise there is no change in the ascending ladder value. An equal
but opposite argument applies to the descending ladder value. If the logarithm
of todays closing price is greater than the logarithm of yesterdays descending
ladder value, SD(t-1), then St is located in one of the three probability zones.
Again, by definition, the descending ladder value must remain below the bullish
zone boundary, i.e., the zone defined by the bearish expected value locus with
the ascending ladder value as origin. Thus, in order to change the descending
ladder value, St must be less than SA(i+m) = SA(i) - S1m1/2. Otherwise there is
no change in the descending ladder value. In conditional format these rules are:
R3. If St < SA(t-1) and St > SD(t-1) + 1n1/2 then SAt = St;
otherwise SAt = SA(t-1),
R4. If St > SD(t-1) and St < SA(t-1) - 1m1/2 then SDt = St;
otherwise SDt = SD(t-1).
These rules are sufficient to identify the occurrence of values that fall in the
gap between the upper bullish zone and the lower bearish zone. Rule 5 follows
immediately from the above:
R5. If SAt = SDt then St = S0.
With the occurrence of S0 the bullish ladder value is logically equivalent to the
bearish ladder value. This equivalence defines a transient balance between
bullish and bearish sentiment. S0 marks the origin of a new segment of the
Markov process. The terms balance values and signals are used interchangeably
in the following.
Price levels where a balance occurs are observed to be both anticipated and
recurrent. Such prices are considered important by a majority of market
participants as trading activity and fluctuations are influenced when these
levels are approached. Ladder values are affected by the perceived value of
the stock or index. Perceived value includes psychological factors; hence,
ladder values are to some degree a reflection of market psychology. If market
2014 . Issue 68
29
psychology is bullish new ascending ladder values are more likely to occur; on
the other hand, if market psychology is bearish new descending ladder values
are more likely. When balance occurs it is assumed, based on the probabilities,
that the bulls and bears have a similar perception of market value.
It is statistically possible for a sequence of fluctuations to occasionally close
the gap between the utility functions such that the bullish locus stemming
from the descending ladder is above the bearish locus stemming from the
ascending ladder. If this happens, depending on where the next close falls, both
ladder levels may become passive. Until the gap reopens market sentiment is
indeterminate. This condition is observed to be infrequent and, when it occurs,
short-lived.
Two types of errors are associated with balance value signals. The first type is
caused by random noise fluctuations instead of the assumed balance between
bullish and bearish sentiment. The second type of error happens if there is a
balance in fact that goes undetected. Reducing the risk associated with spurious
noise signals will be dealt with in a later section. Addressing the second type of
error is beyond the scope of this paper.
30
stocks or indices, each stock or index must be viewed as possibly unique with
its own Z values. In practice, we start with Z factors set to +/-1. Adjustment of
the Z values, if required, aligns balanced ladder values to support and resistance
levels believed by the analyst to be valid. The way this is done is straightforward:
Z values are simply reduced in magnitude in decrements of approximately 0.05
until signals appear on or near a selected barrier point or level. By decreasing the
Z values one increases the probability of a signal and decreases the probability
that the active level will become passive and conversely. The analyst must
review the results to make sure they satisfy common sense.
Consider the DJIA, an algebraic function of a collection of price sequences, each
of which is fluctuating in an unpredictable but generally correlated manner. It
is known that there are circumstances in which a prior price, or an anticipated
price, has an influence on market activity. Such circumstances have at least
one thing in common: a large number of active traders and investors attach
meaning to these values (Cassidy, 1997). Their collective buy and sell decisions
are influenced by, and have an effect on, fluctuations in the neighborhood of
these levels. Some of these levels are known to have psychological significance.
When the Dow Jones Industrial Average was first approaching the 10,000 level
in March of 1999, a great deal of publicity concerning this event occurred in
both print and broadcast media. Obviously, the 10,000 level had psychological
and technical significance to both the investment and trading communities.
A similar significance was also attributed to the 11,000 level. (Luccheti, 2005)
Such levels are often considered technical resistance and support price barriers.
The results of the application of the balanced ladder value algorithm to more
than 4 years of daily DJIA data, from May 1, 1997 through December 31, 2001,
are shown in Figure 11. The 10,000 and 11,000 levels are shown in red. Balance
values are indicated by yellow diamonds. Resistance and support barrier values
associated with signals are shown by blue dashes. Z values were reduced in
magnitude in order to produce the double signal shown in the first ellipse. Of
the 1174 daily closing values in this graph, 67 are balance values, i.e., 5.7%
of the total number of closing values tested. Between March 18, 1999 and
November 28, 2001 eight signals, shown within the six ellipses of Figure 11,
occurred in the immediate vicinity of the 10,000 level. The mean value of the
eight signals is 10,000.3. Similar results are found at 11,000 along with levels
intermediate to 10,000 and 11,000. These results suggest that it is plausible
to argue that the signals generated by the application of the balanced ladder
algorithm are indicative of values that have special significance, psychological
or otherwise, to a plurality of active market participants.
FIGURE 11
aPPlicaTion examPle
Figure 12 is an example of the application of balanced ladder rules to some of
the data of Figure 10. The solid red trend line satisfies the requirements of Rule
10. The trend line correlation coefficient, R2, is greater than 0.999, practically a
perfect fit. Three other trend lines, shown in black, are drawn parallel to the
red trend line. The dashed red trend line is a projection along the slope of the
principal trend. Arrows point out a few examples of valid signals and barrier
values. Note that prior balance values often define subsequent resistance and
support values. For example, consider balance and barrier values at the 9.3
level. The converse is also observed to be the case.
2014 . Issue 68
31
following the second signal, the off-balance volume statistics indicated very
bullish market sentiment.
viii. examPles
A. DJIA Prior Signal Levels: Fluctuation Effects
Figure 14 shows some of the effects that prior signal levels have on subsequent
price fluctuations. Levels A, B, C are extensions of earlier validated signals. Levels
A and B bracket fluctuations in the range between 8826 and 8691. Note that
the high the day before the second signal shown on the graph was 8850, just
FIGURE 12
FIGURE 14
above level A. At this point selling more than offset buying and as a result the
price was driven down sharply to 8612. A barrier at 8608 is associated with the
signal on level C. This graph provides further support for the assertion that prior
signals and barrier values influence subsequent buying and selling decisions
and thereby fluctuation statistics.
FIGURE 13
32
FIGURE 17
FIGURE 15
FIGURE 16
Figure 17 shows the off-balance volume sentiment chart. The red vertical
arrows locate the points of balance. After the second arrow significant selling
pressure becomes evident. After the third arrow, as the daily volume increased
the off-balance volume was negative until the prices encountered the support
level shown in Figure 16.
FIGURE 18
2014 . Issue 68
FIGURE 19
33
references
Figure 19 shows trends, levels and some tentative structural details for Amtech
Systems. Note that when resistance level B failed there was a sharp drop in the
price. Support was found at the level of two prior signals that are part of a head
and shoulders formation near the center of the chart. The resulting balance
value is validated by the signal to its left on the same level.
FIGURE 20
Figure 20 shows three anomalous spikes in daily volume. Associated with these
spikes are negative off-balance volumes. The last selling spike caused the failure
of support at level B shown on the previous chart. We might guess that some
large institution or fund decided that Amtech Systems had run its course and it
was time to exit.
ix. conclusion
The basic premise underlying this work is that there is an ongoing tension that
exists between bullish and bearish psychology, that ladder values associated
with a random walk are driven by ever changing market expectations and
psychology, and that a balance or relaxation of the tension between the
pressures of supply and demand can be detected by means of probabilities and
statistics. One of the innovative concepts in this paper is the identification of a
bullish locus with the bearish ladder and a corresponding bearish locus with
the bullish ladder and the classification of values that occur in the gap between
these boundaries as transient points of balance. The importance of preexisting
support and resistance levels in this process is recognized. These levels are
assumed to influence fluctuations in their vicinity in such a way that balance
values can be located using statistical analysis. Consequently, on the occurrence
of a balance value, it becomes possible to detect the presence of support and
resistance barriers. Trends are observed as a drift in balance values between
bands and channels defined by support and resistance barriers.
34
The Technical
analysis of sTock
markeT index Time series
a comparison
of support
andand
resistance
levels defined through balanced
ladder values and by Golden ratios
carl aspin
bioGraphy
Carl Aspin
Carl Aspin recently retired as President of Aspin Engineering Services in Prescott, Az. He received a degree in Engineering Physics from the
Colorado School of Mines. This was followed by post graduate engineering training at General Electric in Utica. NY. He was employed
at Motorola for more than 30 years in a wide range of R&D, engineering, and management positions. He is a graduate of the Motorola/
absTracT
Technical analysts utilize the golden mean or Fibonacci ratios to locate price
sequence support and resistance levels. Balanced ladder values have been
shown to identify such levels as well. The two methods, which are independent,
are applied to both stock and market index sequences and the results compared.
It is shown through examples that there is excellent agreement between the
support and resistance levels identified by balanced ladder values and golden
ratios. The two techniques are seen to be complementary.
inTroducTion
The paper The Identification and Utilization of Balanced Ladder Levels in the
Technical Analysis of Stock and Market Index Time Series appearing earlier in
this issue suggests that balanced ladder values correspond to prices and price
barriers that form technical support and resistance levels. Price analysts often
use Fibonacci ratios or the golden mean, (phi), to determine where stock
and market index values should find support and resistance (Brown, 2003).
The two methods for identifying possible support and resistance levels are
distinct and independent. If both methods are equally valid then the results of
each should be in reasonable agreement. Thus, the purpose of this paper is to
apply the balanced ladder algorithm to stock and market index sequences and
compare the support and resistance barrier values obtained to those associated
with golden means.
i. Golden mean
The golden mean or golden ratio, = 1.618..., and its association with Fibonacci
numbers is well known to technical analysts. Perhaps not so well known is the
fact that the convergence of Fibonacci ratios to is not unique. The Fibonacci
series is a particular case of the recurrence relation:
i > 2.
to the relative low, Y0, is equal to the ratio of Y0 to the magnitude of the trading
range between the high and low values. Then:
the phi levels depend only on n and are independent of any particular stock or
index. The baseline level, Y(0), is arbitrary and must be determined by means of
independent criteria. The criteria will be described next.
Yn+1 = Yn,
n => 0
The levels at n and n+1 are defined as the primary phi-based support and
resistance levels.
FIGURE 2
Figure 3 shows the natural logarithm of the closing prices for Intel from June 3,
2009 through January 14, 2013. The closing values lie in a range between 12
and 30. Superimposed on the price data are primary, secondary, and tertiary
phi levels. For this example n = 6. Examining Figure 3, we see that 18 and 29
are approximate support and resistance levels, respectively. The secondary and
tertiary phi levels also show possible regions of support and resistance.
FIGURE 1
It follows from Equation 5 that the primary trading range, Yn+1 Yn, is equal to
0.618 Yn, i.e., 61.8% of the primary support level. This range is too wide to be
generally useful for day-to-day or week-to-week analysis. Subordinate levels
are required. Secondary and tertiary trading ranges are obtained by again
assuming that between any two adjacent support and resistance levels the ratio
of the resistance level to the support level is equal to the ratio of the support
level to the trading range. Thus, the secondary levels, between Yn and Yn+1,
are 1.236Yn and 1.382Yn. Tertiary levels between Yn and 1.236Yn and between
1.382Yn and Yn+1 are, respectively, 1.090Yn, 1.146Yn, and 1.472Yn, 1.528Yn.
The half distance between two adjacent tertiary levels is 2.8% Y0. Consequently,
further subdivisions are of little practical value. According to standard practice
a centerline at 1.309Yn is also included as a level. Once an initial value, Yn, is
selected all primary, secondary, and tertiary phi-based S/R levels are fixed.
An example is shown in Figure 2. Y(0) is the baseline support level. Note that
36
FIGURE 3
Consider Figure 4. In this graph balanced value signals and corresponding
barrier values are superimposed upon the sequence of daily closing prices. The
signals and barrier values locate support and resistance levels (Aspin, 2014). By
FIGURE 4
FIGURE 6
FIGURE 5
FIGURE 7
In order to compare the S/R locations obtained by the two analysis techniques
we must first specify a procedure for obtaining signal and barrier value averages.
Thus, let Y(0) = <Y> define a baseline and denote <Yk> as the average that is
to be compared to K*Y(0) where K = 1, 1.090, ,1.1618.
In order to obtain the average, <Yk>, a preselected range centered on K*Y(0)
needs to be defined. The width of this range is somewhat arbitrary. We want to
the mean value of the signals and barrier values between the range boundaries
is calculated.
The first result is shown in Figure 8. The mean < Yk > = 28.76; the corresponding
K*Y(0) value is 1.472Y(0) = 28.77. This process is repeated for the next level
with the results shown in Figure 9. Here < Yk > = 26.97 and 1.382Y(0) = 27.00.
2014 . Issue 68
37
FIGURE 8
FIGURE 10
FIGURE 9
FIGURE 11
This process is continued for the five remaining K*Y(0) levels. The results are
shown in Figure 10 and summarized in Table I. The weighted mean difference
between the K*Y(0) levels and the signal and barrier value averages is -0.02%.
Based upon this first example it appears plausible to argue that the balanced
value signals and associated barrier values are locating support and resistance
levels that are equivalent to phi levels, assuming that an appropriate baseline
value has been selected.
iii. examples:
Four examples follow: Wells Fargo, McDonalds, the S&P 500, and the DJIA. These
stock prices and index values cover four orders of magnitude. In each case it is
shown that < Yk > and K*Y(0) levels can be aligned to exhibit close agreement.
a. Wells farGo:
Figure 11 shows the closing prices, signals, and barrier values for Wells Fargo
between June 3, 2009 and January 14, 2013. The baseline S/R level is 27.35. Figure
38
b. mcdonalds:
Figure 13 shows the data and baseline for McDonalds. The baseline resistance
level is 89.37. The results are summarized in Figure 10. In this example, as the
baseline is set at the primary resistance level, the primary support level is fixed
at 55.23. The weighted mean difference between the K*Y(0) levels and the
signal and barrier value averages, Table III, is -0.02%.
c. s&p 500:
Figure 15 shows the data, baseline, and a summary of the results for the S&P
500. The primary support level is 914. The primary resistance level is 1479. The
weighted mean difference between the K*Y(0) levels and the signal and barrier
value averages, Table IV, is 0.13%.
FIGURE 15
FIGURE 12
FIGURE 16
FIGURE 13
FIGURE 14
It is plausible to argue that phi analysis and the balanced ladder algorithm
identify the same support and resistance levels. Insofar as this is the case the
correspondence further validates the results of the balanced ladder algorithm.
2014 . Issue 68
39
references:
Aspin, Carl H., The Identification and Utilization of Balanced Ladder Levels
in the Technical Analysis of Stock and Market Index Time Series, Journal of
Technical Analysis, 2014
Brown, Constance M., 2003, All About Technical Analysis, (McGraw-Hill, NY)
Livio, Mario, 2002, The Golden Ratio, (Broadway Books, NY)
TABLE III
TABLE I
TABLE IV
TABLE II
TABLE V
40
bioGraPhy
Stella Osoba, CMT
absTracT
Keynes insights into the behavior of market participants and the irrationalities
which shapes their behavior and informs their choices in a capitalistic society
continues to be as relevant today as it was in the days when he first wrote about
them. This article is an analysis of John Maynard Keynes contributions to the
fields of behavioral finance and behavioral economics, through a study of his
writings and an overview of his own speculating and investing behavior.
inTroducTion
It was in Japan in July of 2007, as worldwide fears mounted over the growing
problems in the US subprime mortgage market, loosening credit standards and
increasing signs of a housing bubble that Chuck Prince, then CEO of Citigroup
made a comment widely believed to be the most infamous of the financial crisis
of 2008-2009. He said, When the music stops, in terms of liquidity, things will be
complicated. But as long as the music is playing, youve got to get up and dance.
Were still dancing. (Financial Times, July, 2007)
Over 70 years earlier, Keynes wrote, This battle of wits to anticipate the basis
of conventional valuation a few months hence, rather than the prospective yield
of an investment over a long term of years can be played by professionals
amongst themselves. For it is, so to speak, a game of Snap, of Old Maid, of Musical
Chairs - a pastime in which he is victor who says Snap neither too soon nor too late,
who passes the Old Maid to his neighbor before the game is over, who secures a
chair for himself when the music stops. These games can be played with zest and
enjoyment, though all players know that it is the Old Maid which is circulating,
or that when the music stops some of the players will find themselves unseated.
(Keynes, 1936)
Perhaps, there can be no clearer illustration of the validity, in the real world,
of Keyness enduring description of the psychology of market participants and
proof of the validity of behavioral finance which underpins technical analysis
than Chuck Princes statement.
backGround
Much has been written about John Maynard Keynes. He was a towering intellect
of the twentieth century. Born in 1883, his father was a philosopher and an
economist at Cambridge university, and his mother; Florence Ada Keynes was
the first female councilor of Cambridge Borough Council and she became mayor
of Cambridge in 1932. Keynes studied mathematics at Cambridge, it was there
that one of his tutors, Alfred Marshall who became his mentor encouraged him
to turn his attention to economics. As an economist, Keynes contributions to
the formation of British economic thought was impressive. He worked as a
clerk in the India office, and wrote his first book, Indian Currency and Finance
which was published in 1913. Keynes left the India Office and joined the
Treasury Department. He was the senior member of the delegation to Paris
in 1919 that helped negotiate what became the Treaty of Versailles. Keynes
resigned and left Paris when his warnings were not headed about the dangers
the harsh reparations demanded on Germany by the allies could cause. He
subsequently wrote The Economic Consequences of the Peace (1919) which
turned out to be prescient. Keynes wrote The General Theory of Employment,
Interest and Money, (1936) considered his most important book. It spawned a
revolution in economic thinking, introducing the field of macroeconomics and
Keynesianism; a distinct school of economic thought. Keynes was also part of
the British delegation to the United Nations Monetary and Financial Conference
also known as the Bretton Woods Conference at New Hampshire in 1944 which
was responsible for the establishment of the International Monetary Fund and
the World Bank.
Keynes, J.M., (1972) The Collected Writings of John Maynard Keynes: Essays in Biography, My Early Beliefs, Volume 10, Pg 447-50, London, Macmillan.
Less well known, is Keynes contributions to the field of behavioral finance; the
study of the influence of psychology on market participants. Even less is known
about the fact that Keynes was in his day, a very successful money manager and
investor, operating an investment vehicle called The Syndicate, which were it
to be around today would be appropriately described as a hedge fund. Perhaps
were Keynes achievements in economics less monumental, more attention
would have been paid to his work on the psychology of markets and to his
record as an investor. Keynes highlighted the role of psychology in economics
long before behavioral economics and finance were formed as a distinct
field of study. (Shefrin, 2011) Prior to Keynes writings on the subject, a few
significant works on psychology and market participants had been published,
some of which were, Charles Mackays Extraordinary Popular Delusions and
the Madness of Crowds (1841) about crowd behavior and financial swindles.
Gustave Le Bons influential book, The Crowd: A Study of The Popular mind
(1895), which was about the social psychology of crowds. The Psychology of
the Stock Market (1912) by George Charles Selden, on the mental attitudes of
market participants and their influence on price movements especially in the
short term. But the dominant school of economic thought was the classical
school which gave us the rational man. That human beings were rational,
fully optimizing economic agents was, and continues, to be the accepted view
among mainstream economists. But Keynes disagreed. In My Early Beliefs
he states that as time went by, the falsity of the view that human nature was
reasonable became clearer to him. He states further that this view of what
human nature is like, both other peoples and our ownwas disastrously
mistaken. We are, he says eloquently, as water-spiders, gracefully skimming,
as light and reasonable as air, the surface of the stream without any contact at
all with the eddies and currents underneath. (Keynes, 1972) This paper will be
in two parts. The first part will be a discussion about Keynes main ideas on
the irrationality underlying much of human behavior and how that influences
investment decisions. The second part of the paper will focus on some of Keynes
speculating and investing behavior.
42
uncertainty; nor is the prospect of a Victory bond being drawn. Or, again, the
expectation of life is only slightly uncertain. Even the weather is only moderately
uncertain. (Keynes, 1973) The sense in which Keynes uses the term is that in
which the prospect of a European war is uncertain, or the price of copper and
the rate of interest twenty years hence, or the obsolesce of a new invention
about these matters there is no scientific basis on which to form any calculable
probability whatever. We simply do not know. (Keynes, 1973)
and tolerate uncertainty. He says, that these economists had not analyzed
carefully the psychology of how market participants actually behave in the
market place in forming their models. Their unwillingness to accept the fact
that our knowledge of the future is fluctuating, vague and uncertain (Keynes,
1973) renders their models useless in trying to predict human behavior in the
longer term. He says, I daresay that a classical economist has overlooked
the precise nature of the difference which his abstraction makes between
theory and practice, and the character of the fallacies into which he is likely to
be led. Keynes accuses classical economic theory of being itself one of these
pretty, polite techniques which tries to deal with the present by abstracting
from the fact that we know very little about the future. (Keynes 1973)
Keynes explains that in forming our expectations about those matters of which
we are uncertain, we are initially guided by facts about which we feel more or
less certain, which he calls, the state of confidence. We anchor our decisions on
those facts upon which we can attach a level of certainty to, even when they are
irrelevant to the decision about which we are forming opinions. We often will
look to those facts as they exist at present, which we know something about
and project them into the future in forming our long term expectations. We
only adjust by the extent to which we have reason to believe that we should
expect a change in future conditions. This state of confidence, Keynes argues
has not been studied by classical economists who instead have been content
to assume that they are able to use their models to calculate uncertainty in the
same way as they can calculate matters which are certain (Keynes, 1973).
Rational economic man, would, if he could, as has been assumed by classical
economists, optimize all his decisions by calculating probabilities. But human
decisions affecting the future, whether personal or political or economic, cannot
depend on strict mathematical expectation, since the basis for making such
calculations does not exist; and. it is our innate urge to activity which makes
the wheels go round, our rational selves choosing between the alternatives as best
as we are able, calculating where we can, but often falling back for our motive on
whim or sentiment or chance. (Keynes, 1936) Our tendency in decision making
as human beings is to ignore the laws of probability for the most part when
making decisions. In his Treatise on Probability Keynes stressed the necessity of
explicitly considering psychology to improve probability theory 2
Keynes observes that the vast majority of those who are concerned with the
buying and selling of securities know almost nothing about what they are doing.
They do not possess even the rudiments of what is required for a valid judgment,
and are the prey of hopes and fears easily aroused by transient events and as easily
dispelled. This is one of the odd characteristics of the capitalist system under which
we live, which, when we are dealing with the real world is not to be overlooked.
(Keynes, 1971) This causes markets to be precarious and subject to capricious
changes in sentiment. According to Keynes the following factors also increase
the precariousness of markets:
1) The majority of people participating in markets are ignorant of the real state
of affairs of companies in which they invest.
2) Noise or day-to-day or smaller fluctuations in stocks which are in themselves
insignificant tend to have an excessive and at times absurd influence on the
market.
3) Because large numbers of ignorant individuals arrive at a conventional
valuation of securities based on little more than on the maintenance of
convention, change is liable to be violent as a result of sudden fluctuations of
opinion due to factors which have little or nothing to do with the prospective
yield of the company, because there will be no strong roots of conviction to hold
it steady. (Keynes, 1983)
With imperfect knowledge, and in the face of uncertainty human beings
have devised conventions, which guide the decision making process. These
conventions are short cuts which allow us to pretend to ourselves that we know
more than we do and that we have some justifiable basis for our decisions
which can be relied upon. Since the convention is based on a somewhat
arbitrary selection of facts about which we have a level of confidence in arriving
at a projection for the future long term performance of a security, it is subject to
sudden and violent changes. new fears and hopes will without warning, take
charge of human conduct [when] the forces of disillusion may suddenly impose a
new conventional basis of valuation. (Keynes 1973) But even though when we
attempt to make predictions about the future, we know that the existing state
of affairs cannot continue indefinitely, nevertheless, says Keynes the idea of the
future being different from the present is so repugnant to our conventional modes
of thought and behavior that we, most of us, offer a great resistance to acting on
it in practice. (Keynes 1973)
Animal Spirits is a phrase that Keynes made famous when he used it in The
General Theory to apply to the spontaneous motivation in humans for action.
Keynes wrote that Even apart from the instability due to speculation, there is
the instability due to the characteristic of human nature that a large proportion
of our positive activities depend on spontaneous optimism rather than on a
mathematical expectation, whether moral or hedonistic or economic. Most,
probably, of our decisions to do something positive, the full consequences of which
Wesley, Pech. (2009), Behavioral Economics and The Economics of Keynes, Journal of Be-havioral and Experimental Economics (Formerly the Journal of SocioEconomics), vol. 38, issue 6, pp.2.
2014 . Issue 68
43
2
will be drawn out over many days to come, can only be taken as a result of animal
spirits - of a spontaneous urge to action rather than inaction, and not as the
outcome of a weighted average of quantitative benefits multiplied by quantitative
probabilities. (Keynes, 1936)
According to Ackerlof and Schiller, animal spirits also refers to a restless and
inconsistent element in the economy, to our peculiar relationship with ambiguity
or uncertainty. Sometimes we are paralyzed by it. Yet at other times it refreshes
and energizes us, overcoming our fears and indecisions. (Ackerlof and Schiller,
2009) And it is to these animal spirits, to the spontaneous optimism of human
nature upon which enterprise depends. Animal spirits dim the thoughts of
ultimate loss when we embark on new endeavors and push us to take risks and
build enterprises which if successful will benefit the community as a whole. It is
animal spirits which propels the pioneer forward and which causes the healthy
man to put aside fears of death (Keynes, 1936).
Even though psychological factors underlie how market participants form
opinions upon which they base their decisions for predictions about the
future direction of prices, and even though this causes an acknowledged
precariousness, because it is human nature to become habituated to our
surroundings, Keynes says, we assume some of the most peculiar and temporary
of our late advantages as natural, permanent, and to be depended on, and we
lay our plans accordingly. (Keynes, 1919) Therefore, as long as the balance
of market participants accept the current basis for valuation this will cause
markets to show a degree of continuity and stability.
He who attempts [to be a serious minded investor] must surely lead much more
laborious days and run greater risks than he who tries to guess better than the
crowd how the crowd will behave; and, given equal intelligence, he may make
more disastrous mistakes. 3
Even the wisest and most astute of investors have to contend with the limitations
of the human mind. Their ignorance of the distant future is by far greater than
their knowledge of the present. In attempting to predict the future, they will
also find themselves disproportionately seeking clues in the present which they
will project into the future (Keynes, 1971). But what of the expert professional
investor? Can he not be relied on to correct the vagaries of the ignorant investor
with his greater skill, knowledge and judgment? Keynes explains that the
professional cannot be relied on to correct the irrationalities caused by the
masses, and may often exacerbate it. This is because no one wants to be left
holding a security which they believe the market will value lower three months
Keynes, J.M. (1936) The General Theory of Employment Interest and Money, Chapter 12 (5)
44
or a year from now. The professional investor is therefore engaged in the game
of trying to ascertain what public opinion will value a security at a few months
hence by looking for impending changes in the news or in the atmosphere, of the
kind by which experience shows that the mass psychology of the market is most
influenced. (Keynes, 1936) Professional investors find themselves therefore for
the most part engaged in a game of trying to beat the gun, to outwit the crowd,
and to pass the bad, or depreciating half-crown to the other fellow. (Keynes,
1936)
Keynes likens professional investment to a newspaper competition where
competitors have to pick out the six prettiest faces from a hundred photographs.
The prize is then awarded to the person who successfully picks out the face
which average opinion agrees is the prettiest. So the puzzle then becomes a
question of picking out of the bunch not the face that the competitor considers
the prettiest, but the face he thinks that average opinion will consider the
prettiest. In Keynes words, It is not a case of choosing those which, to the best
of ones judgment, are really the prettiest, nor even those which average opinion
genuinely thinks the prettiest. We have reached the third degree where we devote
our intelligences to anticipating what average opinion expects average opinion
to be. (Keynes 1936) Professional investing therefore be-comes a part of the
same game played by the public. But with their greater insight, the professional
plays the game by trying to beat the crowd. And as Keynes says, because crowd
psychology can be discerned, it is wise and natural that the professional investor
should be influenced by their expectations on the basis of past experience of
the trend of mob psychology. Thus, so long as the crowd can be relied on to act
in a certain way, even if it is misguided, it will be to the advantage of the betterinformed professional to act in the same way - a short period ahead. (Keynes,
1971) Therefore making the professionals actions somewhat rational (Winslow,
2010) while escalating the underlying irrationality in the market.
Chuck Prince came in for a lot of criticism for his dancing comments. But what
if he had chosen to ignore short term expectations and go against the street? If
this had caused Citigroup to underperform its peers, how would he then have
justified his position to his Board? It is likely he would have been viewed even
more harshly by his Board if in the midst of a raging bull market he was forced
to justify underperformance by pointing to the irrationality of others. And it is
possible that were he to be then relieved of his duties, his successor would have
joined in the game of musical chairs and got handsomely rewarded while the
game was being played. For as Keynes stated so presciently, worldly wisdom
teaches that it is better for reputation to fail conventionally than to succeed
unconventionally. (Keynes, 1936)
A good example of someone who sat out the game, is famed investor Warren
Buffett, widely acknowledged as the worlds most successful investor. But he
endured many years of ridicule during the dot-com bubble for some of his
investment decisions, most notably, his decision not to invest in anything he
did not understand, therefore standing aside during the market frenzy. As a
consequence, in 1999, Berkshire Hathaway underperformed the broad market
indexes, with many analysts dismissing Buffett and claiming that he had lost his
touch because he had missed the technology boom. When the market crashed
in the early 2000s with the Nasdaq dropping an intra day low of 1108.49 in
October 2002, from an all time high of 5,048 on March 10, 2000 Buffett was
vindicated. He said at the time, It was a mass hallucination, by far the biggest
in my lifetime. 4
Buffett, therefore plays to perfection the role Keynes ascribes to the serious
minded investor who, unperturbed by the prevailing pastime, continues
to purchase investments on the best genuine long term expectations he can
frame. (Keynes, 1936) But not everyone can be a smart investor, or will
be rewarded and not penalized for it. Keynes states some of the reasons
why smart investment strategies are so much more difficult than those
employed by the game-players. To stand apart from the crowd to is open
oneself to the possibility of making more disastrous mistakes. It needs more
intelligence to defeat the forces of time and our ignorance of the future than
to beat the gun. Human nature desires quick results, there is a peculiar zest
in making money quickly, and remoter gains are discounted by the average
man at a very high rate. The game of professional investment is intolerably
boring and over-exacting to anyone who is entirely exempt from the
gambling instinct. Also, an investor who decides to ignore near-term market
fluctuations needs greater financial resources and cannot afford to operate
with large amounts of leverage. (Keynes, 1936)
Finally, Keynes points out that this type of investor is likely to come in for the
most criticism, especially if he is managing funds for a committee, board or
bank. Since such an investor who goes against the crowd/consensus is likely
to be eccentric, unconventional and rash in the eyes of average opinion. If he is
successful, that will only confirm the general belief in his rashness; and if in the
short run he is unsuccessful, he will not receive much mercy. (Keynes, 1936)
It is likely therefore, that he will not remain in business long.
summary
Our need to predict the future comes out of our inability to tolerate
uncertainty, so we create conventions to help us to explain the future. Our
predictions are based on facts upon which we have a level of confidence
4
5
which we project into the future irrespective of whether or not those facts
have anything to do with the issue upon which we are attempting to form
predictions. When we are unable to form our own predictions, we accept
conventional judgment as our guide leading to herding behavior. When we
have arrived at our predictions, our animal spirits compels us to take some
action, any action. This behavior can cause precariousness in markets. But
precarious markets can trend for extended periods of time and then shift and
turn suddenly and unexpectedly, when the basis of conventional wisdom
changes resulting in booms and crashes.
Keynes, J.M. (1936) The General Theory of Employment Interest and Money, Chapter 12 (5)
Keynes, J.M. (1983) The Collected Writings of John Maynard Keynes: Economic Articles and Correspondence, Volume 12, 109, London, Macmillan
2014 . Issue 68
45
Keynes rebuilt his fortune the same way he had lost it, by speculating. Records
show that Keynes was an extremely confident speculator.6 By the end of 1922,
he had paid off all of the debts he had incurred for the Syndicate and in the
process became a substantial investor with records showing his net assets in
excess of 21,000.
It is not uncommon to come across superficial studies on Keynes where they
state something to the effect that he started off as a speculator/market timer,
learnt the error of his ways and then became a value investor. This is simply
not true. Apart from the earliest days when Keynes held individual stocks on a
modest scale, his speculating and investing activities overlapped for most of the
time he was active. Sometimes he speculated in the hopes of making a profit,
other times, he speculated to hedge investment positions.
By 1936, Keynes had become a substantial investor, managing money for
several entities, and he was still very active as a speculator. One incident will
illustrate his activities during this period. One day he was found measuring
the cubic capacity of Kings college chapel. When people enquired why, he told
them that he was about to take delivery of several loads of wheat. In fact, he
had speculated in what amounted to a months supply of wheat for the entire
country. He was measuring the chapel to see if it would hold the wheat when
he took delivery of it and was somewhat annoyed when he found that the
chapel was too small, only able to take about half of the wheat. Keynes then
hatched a plan. He decided that he would take possession of the wheat, but
as each cargo came in, he would object to its quality. He knew that each cargo
was coming in from Argentina and would likely need to be cleaned. He also
knew that the available machinery to clean wheat was limited and could only
handle one cargo at a time. It would also take some days to clean an entire
cargo of wheat. This is exactly what happened. As each cargo load came in,
Keynes would object to its quality and upon inspection, it was indeed found
that every cargo had to be cleaned. It took over a month for Keynes to clear the
position with no loss or profit to him for storing the wheat. 7
In a memorandum that Keynes wrote to the Estates Committee of Kings
College, Cambridge dated 8th May, 1938, he affirmed an earlier statement he
had made, when he stated that, it is safer to be a speculator than an investor in
the sense of the definition I once gave the Committee that a speculator is one who
runs risks of which he is aware and an investor is one who runs risks of which he is
unaware. (Keynes, 1983)
Keynes also wrote that, Speculators may do no harm as bubbles on a steady
stream of enterprise. But the position is serious when enterprise becomes the
bubble on a whirlpool of speculation. (Keynes, 1936) When speculators
dominate in markets over those involved in enterprise and markets become
little more than casinos, market instability is a result.
My central principle of investment is to go contrary to general opinion, on the
ground that, if everyone is agreed about its merits, the investment is inevitably too
dear and therefore unattractive. 8
Up until June 1919, when Keynes resigned from the Treasury, records indicate
that he operated his trading activities on his own account only, and on a
modest scale. At this time, though he did not trade for others, he did provide
investment advise to friends and family. Records show that Keyness first
purchase was on July 6th, 1905 for 4 shares in the Marine Insurance Company
(Keynes, 1983). Over the next few years, he bought more shares in a few other
companies, still on a very modest scale. During this period Keynes activities
seemed to have been limited to buying and holding small positions and adding
to those positions.
After June 1919, Keynes began to both speculate and invest in financial
markets on a much larger scale. Upon his return from Treasury to take up
teaching at Kings College, Cambridge, he was made Second Bursar of the
Colleges endowment fund. He immediately set upon influencing the College to
broaden its investment portfolio out of fixed income securities and real estate.
He succeeded in getting the Trustees to authorize the setting up of a separate
fund, which was the origin of The Chest.9 Through this fund, Keynes was able to
invest in foreign government securities, and other securities including shares,
currencies and commodities. Up until that time equities had been considered a
new asset class, and too risky for college endowments to touch.
46
In the early days, of his investing career, it would appear that Keynes employed
a top down macro investment strategy. Results of his performance are mixed
and it would appear that this strategy did not result in consistent success.10
Records show that between 1922 and 1929, Keynes performance was worse
than the Bankers Magazine Index (Keynes, 1983). He was a very confident
and extremely active investor in his early years, but also later on, with records
showing that within most of the years, between 1923 and 1940, the value of
securities he sold exceeded the market value of the securities he held at the
beginning of the year. An audit performed in 1945 showed that almost 30%
of positions were held for 3 months or less with only 15% being held for longer
than 3 years (Keynes, 1983). After 1929, Keynes investment record improved
considerably and he outperformed the index by a wide margin in 21 out of 30
accounting years (Keynes, 1983).
In 1924 Keynes was made First Bursar, of the Chest and he had complete
discretion over the investments in the fund. Keynes also held several other
money management posts including, Chairman of the National Mutual Life
Assurance Society (1921-38), a director of the Provincial Insurance Company,
and directorships of a number of investment trusts, namely; the Independent
Investment Company (1923-46); the A.D Investment Trust (1921-7); the P.R.
Finance Company (1924-36) and its Chairman (1932-6), and as if he did not
have enough of his plate, he ran the Syndicate; his hedge fund.
As a macro manager, Keynes, attempted to sell market leaders in a falling
market and buy them in a rising one. Keynes appears to have not been able to
do this successfully, saying, it needs phenomenal skill to make much out of this
strategy. He changed his investment strategy to a bottom up approach, and he
said, my alternative policy undoubtedly assumes the ability to pick specialities
which have, on the average, prospects of rising enormously more than an index
of market leaders. The discovery which I consider that I have made in the course
of experience is that it is altogether unexpectedly easy to do this. (Keynes, 1983)
Easier for him to implement, he said, than his credit cycling strategy, and it also
enabled him to take advantage of market fluctuations (which was the point of
credit cycling), although in a different way. With this new strategy, he could buy
when the market fell. Because in market crashes, only the astute investor is able
to see that bargains are to be had, as shares have gone on sale. As Keynes puts
it, It is largely the fluctuations which throw up the bargains and the uncertainty
due to fluctuations which prevents other people from taking advantage of them.
(Keynes, 1983) Keynes acknowledged that even though he had often been too
slow to sell his shares after they had finished most of their rise, it was better, in
his opinion to be too slow to sell, than to be too fast to sell in a rising market.
Because you would lose less (Keynes, 1983).
The evolution of Keynes investment strategy is evident in a letter he wrote to
F.C. Scott dated 15 August, 1934, but, in so far as one is prepared to continue
to hold investments in the metal, I have decided for myself and for other accounts
for which I am responsible, to concentrate practically the whole of what I am
prepared to invest in this way in the Union Corporation, and then to hold the shares
obstinately for a period of years for a really large appreciation, - unless, as I have
said, the gold position as a whole shows signs of change. (Keynes, 1983) And
then he wrote, as time goes on I get more and more convinced that the right
method in investment is to put fairly large sums into enterprises which one thinks
one knows something about and in the management of which one thoroughly
believes. It is a mistake to think that one limits ones risk by spreading too much
between enterprises about which one knows little and has no reason for special
confidence. (Keynes, 1983)
In 1938, Keynes wrote a memorandum to the Estates Committee of Kings
College about the investments which he managed on behalf of the college. He
wrote in that report how his investment policy changed from 20 years ago when
he first persuaded the college to invest in equities. In this letter it is apparent
that Keynes had evolved and matured as an investor and the three principles
which he lays out in the letter are classic value investing strategies which are as
relevant to successful investing today as they were to investing in Keynes time.
According to Keynes, the principles to successful investing are:
(1) a careful selection of a few investments (or a few types of investment) having
regard to their cheapness in relation to their probable actual and potential intrinsic
value over a period of years ahead and in relation to alternative investments at
the time;
(2) a steadfast holding of these in fairly large units through thick and thin, perhaps
for several years, until either they have fulfilled their promise or it is evident that
they were purchased on a mistake;
(3) a balanced investment position, i.e. a variety of risks, in spite of individual
holdings being large, and if possible opposed risks (e.g. a holding of gold shares
amongst other equities, since they are likely to move in opposite directions when
there are general fluctuations). (Keynes, 1983)11 In this way, Keynes believed
that the skilled investor could succeed in achieving the social object of skilled
investment which should be to defeat the dark forces of time and ignorance which
envelope our future. 12
There still needs to be more research done into Keynes trading record for a definitive evalua-tion to be valid of his trading results.
Keynes, J.M. (1983) The Collected Writings of John Maynard Keynes: , London, Macmillan
12
Keynes, J.M. (1936) The General Theory of Employment Interest and Money, Chapter 12, Kindle Edn.
10
11
2014 . Issue 68
47
conclusion
There is much for the technician to learn from an analysis of the writings and
investment style of John Maynard Keynes. He was one of the earliest of the
modern day institutional and hedge fund managers. He traded a variety of
markets and in a variety of styles. Though he did not always succeed, he was
willing to take on risk and study the markets to glean what worked, and what did
not. He was not afraid to change his mind as we have seen through the evolution
of his trading strategies from top down to bottom up and from credit cycling to
value investing. He was able to speculate and invest without locking himself
into any one particular investment camp. And most importantly, perhaps from
our perspective as technicians is his contribution to our understanding of the
field of behavioral economics or the psychology of market participants, which
is the basis of technical analysis. As the Market Technicians Association says in
its explanation of what technical analysis is, In general, a technician believes
that people have predictable mental short cuts [when] reacting to action in the
markets (known as heuristics in cognitive psychology). Technicians seek to profit
by anticipating the mass psychological biases of buyers and sellers in a broad
range of markets. And this is exactly what Keynes sought to do through his
trading strategies.
references
Keynes, J.M. (1936), The General Theory of Employment, Interest and Money,
Chapter 12
Keynes, J.M. (1919), The Economic Consequences of the Peace, Chapter 1
Keynes, J.M. (1921) A Treatise on Probability
Akerlof, G.A. and Shiller R.J. (2009) Animal Spirits: How Human Psychology Drives
the Economy, and Why it Matters For Global Capitalism.
Keynes, J.M. (1971) The Collected Writings of John Maynard Keynes: Activities 1906
- 1914: India and Cambridge, London, Macmillan, Volume 5,
Keynes, J.M. (1971) The Collected Writings of John Maynard Keynes: The Treatise on
Money, London, Macmillan, Volume 6, p 323 - 324
Keynes, J.M. (1972) The Collected Writings of John Maynard Keynes: Essays in
Persuasion, London, Macmillan, Volume 9,
Keynes, J.M. (1973) The Collected Writings of John Maynard Keynes: The General
Theory of Employment, Interest and Money, London, Macmillan, Volume 7,
Keynes, J.M. (1972) The Collected Writings of John Maynard Keynes: Essays in
Biography, London, Macmillan, Volume 10, p 447 - 450
Keynes, J.M. (1983) The Collected Writings of John Maynard Keynes: Economic
Articles and Correspondence, London, Macmillan, Volume 12, p 4, 5, 10-12, 57, 82,
90, 109, 100-101, 107, 111
13
14
48
Keynes, J.M. (1973) The Collected Writings of John Maynard Keynes: The General
Theory and After: Defense and Development, London, Macmillan, Volume 14, p 112
-113, 114 - 115,124 - 125
Mackay, Charles (1980) Extraordinary Popular Delusions and the Madness of Crowds.
New York, NY: Crown Publishing Group
Tversky, A and Kahneman, D. (1974) Judgment under Uncertainty: Heuristics and
Biases. Science, New Series, Vol 185 pp. 1124-1131
Kahneman, D. and Tversky, Amos (1979) Prospect Theory: An Analysis of Decision
under Risk Econometricia, 47(2) pp.263-291,
Chambers, D. and Dimson, E. (2012) The Stock Market Investor. Journal of Financial
and Quantitative Analysis (Forthcoming)
Shefrin, H. Behavioral Finance in the Financial Crisis: Market Efficiency, Minsky, and
Keynes
Beachy, B, (2012) A Financial Crisis Manual: Causes, Consequences, and Lessons
of the Financial Crisis. The Global Development and Environment Institute (GDAE)
Working Paper No.1206, Tufts University
Fantacci, L., Marcuzzo, M.C., and Sanfilippo, E., (2010) Speculation in Commodities:
KeynesPractical Acquaintance With Futures Markets. Journal of the History of
Economic Thought, Volume 12, Number 3
Chua, J.H. and Woodward, R.S., (1983) J.M. Keynes Investment Performance: A
Note. The Journal of Finance 38(1): 232-235
Barberis, N., Shleifer, A. and Vishny R., (1998) A Model of Investor Sentiment,
Journal of Financial Economics, 49, pp. 307-343.
Wesley, Pech. (2009), Behavioral Economics and The Economics of Keynes, Journal
of Behavioral and Experimental Economics (Formerly the Journal of SocioEconomics),
vol. 38, issue 6, pp.891-902.
Winslow, Ted. (2010), Keynes on the Relation of the Capitalist Vulgar Passions to
Financial Crises, Studie Note di Economia, Anno XV, n. 3-2010, pp. 369-388
Ricciardi, V and Simon, H.K. (2000) What is Behavioral Finance? Business,
Education & Technology Journal, Vol. 2, No. 2, pp. 1-9.
Sewell, M., (2010) Behavioural Finance. University of Cambridge.
BOOVBM
TZNQPTJVN
#&:0/%5)&#&/$)."3,
"QQMJDBUJPOTGPS"DUJWF"TTFU.BOBHFNFOU
.BSDI
/FX:PSL$JUZ
&BSMZ3FHJTUSBUJPOJTOPX0QFO3FTFSWF:PVS4FBU5PEBZBU
IUUQTZNQPTJVNNUBPSH
bioGraPhy
Tom McClellan
Tom McClellan has done extensive analytical spreadsheet development for the stock and commodities markets, including the synthesizing of the
important market and economic data.
as an Army helicopter
absTracT
Natural phenomena have a much better correlation to stock prices than reason
or chance might allow. The linkage between the financial markets and sunspots,
rainfall, temperature, and other data likely runs through the agricultural sector.
Better or worse crop yields likely affect the financial markets more than is
appreciated by some analysts. Overwhelming evidence of a co-relationship exists,
even if we cannot establish a causal relationship. This information can be used to
aid market forecasting.
inTroducTion
When some technicians want to get answers ahead of time about what is going
to happen to the stock market, they turn for answers to the movements of the
other planets in our solar system. They think that they can find answers about
when stock prices will turn by looking at when planets other than Earth get into
certain specific orientations or relationships. But looking out into the solar system
can mean missing some really good information about what is happening right
here on our own planet. There are certain pieces of information about natural
phenomena here on Earth which have a much better correlation to stock prices
than reason or chance might allow. Something is clearly going on.
Among the interesting relationships between Mother Earth and the financial
markets are the following:
t4VOTQPUTBSFBTTPDJBUFEXJUIDZDMFTPGJOBUJPO
VOFNQMPZNFOU
TPDJBMVOSFTU
t3BJOGBMMJTCVMMJTIGPSTUPDLT
PSHPME
EFQFOEJOHVQPOXIFSFJUGBMMT
t&M/JPJTCFUUFSGPSDPSQPSBUFQSPUTBOEUIFTUPDLNBSLFUUIBO-B/JB
t"UPSOBEPXBTCBEGPS%PSPUIZBOE5PUP
CVUUIFZBSFHPPEGPSUIFTUPDLNBSLFU
t(MPCBMXBSNJOHJTBHPPEUIJOH
BUMFBTUGPSUIFTUPDLNBSLFU
t.PSFBSDUJDJDFNFBOTMFTT(%1
sunsPoTs
Sunspots are a symptom of greater activity in the sun, and their numbers wax and
wane on a fairly consistent 9-14 year cycle, with an average period of 11 years. As
far back as 1924, Alexander Chizhevsky had correlated sunspot activity to social
phenomena including disease outbreaks and revolutions. (citation). That revelation
HPU$IJ[IFWTLZJOUPUSPVCMF
CFDBVTFIJTFYQMBOBUJPOTGPSXIBUMFEUPUIF3VTTJBO
revolutions of 1905 and 1917 were at odds with what Joseph Stalin thought about
the reasons for those events. It resulted in Chizhevsky spending 8 years in a gulag
for espousing politically incorrect theories about the interrelationship between
geophysical and social phenomena.
Figures 1 and 2 show some up-to-date examples of what Chizhevsky was referring
to, with the ascending phase of each sunspot cycle seeing greater occurrences of
protest movements, revolutions, and other social unrest.
In Figure 1 it is not shown here, but is worth noting that sunspots saw a minimum in
BOEUIFOOVNCFSTTUBSUFESJTJOHJO
KVTUBTUIF"NFSJDBO3FWPMVUJPOXBT
getting underway. Similarly, the abolitionist movement reached a climax in the late
1850s, following a sunspot minimum in 1856 with rising sunspot numbers leading
VQUPUIFFMFDUJPOPG"CSBIBN-JODPMOJOMBUF GUQOHEDOPBBHPW
$PVOUJOH
forward from 2011, we can expect then next mass protest movement around 2022.
What jumps out to the naked eye is that the low points in the sunspot cycle
NJOJNB
UFOEUPDPJODJEFSPVHIMZXJUIJNQPSUBOUMPXTGPSTUPDLQSJDFT1FSIBQT
more importantly, the ascending phase of the cycle is associated with price
appreciation for stocks.
But it is still a really rough correlation, with some significant anomalies. This
means that the correlation by itself is not good enough to form the basis of an
entire trading system, even though over the decades it is evident that there is
something going on. Turning toward other types of financial market data, we find
better correlations to sunspot data, especially when an important adjustment is
made. Figure 4 shows a coincident comparison of the monthly sunspot number
UPUIFJOBUJPOSBUF
FIGURE 4 Notice that the peak in the sunspot cycle tends to coincide with a peak in the
JOBUJPO SBUF 5IFSF JT BMTP TPNF EFHSFF PG DPSSFMBUJPO FWJEFOU CFUXFFO UIF
TUSFOHUI PG UIF TVOTQPU DZDMF BOE UIF TFWFSJUZ PG UIBU JOBUJPO BU UIF QFBL
although that relationship is a little bit weaker. There have been only 10 sunspot
DZDMFTTJODFUIFDSFBUJPOPGUIF$POTVNFS1SJDF*OEFYCBDLJO
BOETPUIBU
is not enough to reliably correlate maximum sunspot number magnitude and
JOBUJPONBHOJUVEF
*UJTBMTPXPSUIOPUJOHUIBUTQJLFTJOUIFJOBUJPOSBUFIBWFPDDVSSFEBUPUIFSUJNFT
not suggested by the sunspot cycle. These are usually spikes related to actions by
governments, such as wars or commodity bubbles, as well as the notable example
PGUIFRVBTJHPWFSONFOUBMBHFODZ01&$NPVOUJOHBOPJMFNCBSHPJOUIFT
Once those exogenous forces are removed from the marketplace, we can see that
JOBUJPOHFUTCBDLPOUSBDLXJUIUIFTVOTQPUDZDMF
FIGURE 3 -
52
If we make one key adjustment, as discussed below, we can find that the sunspot
cycle will actually lead movements in other data. Figure 5 provides a comparison
of the sunspot number to the U.S. unemployment rate:
The idenTificaTion
and
of balanced
ladder levels
in
lookinG
TouTilizaTion
our oWn PlaneT
for markeT
insiGhTs
The Technical analysis of sTock and markeT index Time series
Normal rainfall for New York City is about 48 per year. Interestingly, that is more
UIBO4FBUUMFBOE5BDPNBHFU
JOUIFTVQQPTFEMZSBJOZ1BDJD/PSUIXFTU
XIJDIJT
discussed below. It is not the average rainfall that matters for the markets, but
rather the deviation from average.
FIGURE 5 For this chart, the key adjustment is that the sunspot data have been shifted
GPSXBSECZZFBST NPOUIT
UPSFWFBMUIBUUIFVOFNQMPZNFOUSBUFGPMMPXTJO
UIPTFTBNFGPPUTUFQTZFBSTMBUFS
BUMFBTUGPSUIFNPTUQBSU
5IJT SFWFMBUJPO JFT JO UIF GBDF PG XIBU TPNF FDPOPNJD IJTUPSJBOT TBZ JT UIF
explanation for protest movements, revolutions, and other mass defiance
phenomena. There is a popular theory that it is unemployment and economic
suffering which lead to such protests and revolutions. But the mass protest
movements appear to be coincident with the rise in the sunspot number, whereas
UIFSJTFJOUIFVOFNQMPZNFOUSBUFUFOETUPPDDVSZFBSTMBUFS
"TXJUIUIFJOBUJPOEBUBSFGFSFODFEBCPWF
UIFSFBSFPDDBTJPOBMTQJLFTJOUIF
unemployment rate which are not explained by the sunspot cycle, but each can
be attributed to some type of governmental action. What can be said with greater
statistical reliability is that the rising phase of the sunspot cycle has always been
GPMMPXFECZBSJTFJOUIFVOFNQMPZNFOUSBUFZFBSTMBUFS
PSBUMFBTUUIBUIBTCFFO
UIFDBTFGPSBTMPOHBTUIF-BCPS%FQBSUNFOUTVOFNQMPZNFOUSBUFEBUBIBWF
FYJTUFE TJODF
5IJTTUSPOHMZTVHHFTUTUIBUTIPVMETFFBSJTFJO
the unemployment rate, which just recently has been trying to get back on track in
the wake of the Fed-induced housing bubble in the mid-2000s and its subsequent
CVTU*GXFEPOPUTFFSJTJOHVOFNQMPZNFOUJO
JUXPVMECFUIFSTUUJNF
JOPWFSZFBSTPGVOFNQMPZNFOUEBUBUIBUBSJTFJOTVOTQPUTXBTOPUGPMMPXFE
years later by a rise in unemployment.
rainfall
*GBOOVBMSBJOGBMMJO/FX:PSL$JUZJTHSFBUFSUIBOOPSNBM
UIFOUIBUTHPPEGPSUIF
stock market. That might not seem like a good place to look for a cause and effect
relationship, but getting stuck wondering about causes can prevent us from seeing
what is.
2014 . Issue 68
53
EPXOUSFOEXIJDIMBTUFEVOUJMUIFSBJOTUBSUFEDPNJOHCBDLJOMBUF"UUIFMFGU
end of the chart, we see that the market crash at the start of WWI coincided with
the start of a really dry period for New York City.
Why would that matter? Why is rain in New York City a good thing for the
economy, but dry periods are bad? It is not necessary have to answer the why in
order to notice the is. Our human brains like it better when we can understand
the why, but it is not essential. Until Copernicus came along, humans did not
understand the why of the sun rising each day, but they came to accept it given
the considerable evidence that it was so.
0OUIFPUIFSTJEFPGUIFDPVOUSZ
SBJOGBMMUPUBMTJOUIF1BDJD/PSUIXFTUNBUUFS
CVU
in a different way. Figure 8 compares rainfall deviations from normal in Tacoma,
Washington, to gold prices:
el nino
5IF&M/JPQIFOPNFOPOJOWPMWFTXBSNFSUIBOOPSNBMXBUFSJOUIF1BDJD
Ocean, and it has a big effect on global weather. The name comes originally
GSPN1FSVWJBOTIFSNFOXIPOPUJDFEUIBUBXBSNDVSSFOUTFFNFEUPBQQFBSJO
some years around Christmas. It is strongly correlated to what meteorologists
call the Southern Oscillation Index, which refers to the difference in atmospheric
QSFTTVSF CFUXFFO UIF JTMBOE PG5BIJUJ BOE %BSXJO
"VTUSBMJB 5IF 4PVUIFSO
0TDJMMBUJPOIBTCFFOLOPXOBCPVUTJODFUIFT
CVUJUTSFMBUJPOTIJQUP&M/JP
XBTOPUEFTDSJCFEVOUJM
XIFO1SPG+BDPC#KFSLOFTPG6$-"XSPUFBCPVUJU
The relationship is now referred to in the weather industry by the acronym ENSO
&M/JP4PVUIFSO0TDJMMBUJPO
GBDVMUZXBTIJOHUPOFEV
8IFO&M/JPDPOEJUJPOTBSFJOFFDU
XFBUIFSPCTFSWFSTSFDPSEQSFEPNJOBOUMZ
IJHIFSBUNPTQIFSJDQSFTTVSFMFWFMTJO5BIJUJUIBOJO%BSXJO*GUIBUDPOEJUJPO
occurs over a long period of time (several months), it tends to be associated
XJUISJTJOHTUPDLQSJDFT5IFPQQPTJUFDPOEJUJPOJTLOPXOBT-B/JB
XIJDIJT
generally bearish for stock prices.
GUQGUQDQDODFQOPBBHPWXEEHEBUBJOEJDFTTPJIJTBOEIUUQXXXDQDODFQOPBBHPWEBUBJOEJDFTTPJ
54
The idenTificaTion
and
of balanced
ladder levels
in
lookinG
TouTilizaTion
our oWn PlaneT
for markeT
insiGhTs
The Technical analysis of sTock and markeT index Time series
There are a couple of anomalous periods, but there is generally a strong positive
correlation. The tornado count drop suggested that the 1970 bottom should have
been more significant than the 1974 bottom, but we should recall that an oil
embargo and a presidential resignation in 1974 helped to push stock prices down
more than might otherwise have been the case.
Figure 12 provides a more up-to-date picture of that same relationship:
Tornados
The National Oceanic and Atmospheric Administration (NOAA) publishes monthly
totals for tornado occurrences in the U.S.2 These data reveal a cyclical nature in the
frequency with which tornados appear in the U.S. It turns out that these data give
about a 2-year leading indication for what stock prices are likely to do.
Figure 11 shows what this relationship between tornados and stock prices looked
MJLFGPSUIFQFSJPEGSPNUP
FIGURE 12 We can see that the big rally during the Internet bubble of the late 1990s followed
a similarly big increase in the tornado count (note that the tornado count plot is
shifted forward in this chart). And with the tornado count declining in 2000-2001,
JUJTVOEFSTUBOEBCMFUIBUBESPQJOTUPDLQSJDFTUPXBSEUIFCPUUPNXPVME
unfold to match the drop in tornados. A similar peak and decline also explain the
2007 top and 2008 bear market.
The period since the 2009 bottom is a little bit harder to explain according to this
NPEFM5IF'FETFPSUTUPTNPPUIPWFSBMMSJQQMFTJOUIFMJRVJEJUZTUSFBNJTQBSUPG
the story, as is a big anomaly in the tornado record during April 2011, when two
separate outbreaks hit the U.S. One was April 14-16, 2011, and it caused a string of
tornados stretching from Mississippi to Virginia. The other was April 25-28, 2011,
XIJDILJMMFEQFPQMF4. Since April 2011, the tornado totals have been lower
than normal, which suggests that if stock prices continue to correlate strongly with
tornadic activity, then we are likely to see a corrective period for stock prices.
At this degree of graphical resolution, it is still a lumpy looking correlation. Adding
BEEJUJPOBMTNPPUIJOHUPBNPOUIUPUBM
BTTIPXOJO'JHVSF
DMFBOTJUVQ
some more:
FIGURE 11 IUUQXXXODEDOPBBHPWTUPSNFWFOUTGUQKTQ
IUUQXFCBSDIJWFPSHXFCIUUQXXXTQDOPBBHPWQSPEVDUTXBUDIXXIUNM
4
IUUQXXXTQDOPBBHPWDMJNPUPSOGBUBMUPSOIUNM
2
2014 . Issue 68
55
raising interest rates, thereby further weakening the economy and the stock
NBSLFU4JODF
UIFEFUSFOEFEDPSSFMBUJPOJTCFUUFSBU
For most of the rest of this period, the advancing and retreating periods for
tornados have been matched 12 months later by corresponding advances
and declines in stock prices. This highlights one of the problems with
SFMZJOHTPMFMZVQPODPSSFMBUJPODPFDJFOUTUPEFUFDUSFMBUJPOTIJQT5IFZBSF
dependent not only upon timing but also upon magnitude of movements,
and sometimes the market does not work that way. Sometimes there is a
different magnitude of response, even though the timing of the direction
change is still there.
FIGURE 13 %PJOH B TJNQMF 1FBSTPOT $PSSFMBUJPO $PFDJFOU GPS UIF FOUJSF QFSJPE
shown brings a misleadingly high +0.88. It is misleading because there
is an upward slope to both series, and that throws off a simple correlation
DPFDJFOUDBMDVMBUJPO
If we detrend both sets of data by using a linear regression, we can calculate
FBDIEBUBQMPUTEFWJBUJPOGSPNUIBUMPOHUFSNUSFOE
UIFSFCZUBLJOHBXBZ
UIFFFDUPOUIFDPSSFMBUJPODPFDJFOUCSPVHIUCZFBDIQMPUTVQXBSETMPQF
Figure 14 shows that adjustment:
Global WarminG
The Southern Oscillation Index (SOI) discussed above also plays a role in
global warming, with a far closer correlation to temperature data than
is found with items like atmospheric carbon dioxide. Figure 15 provides a
DPNQBSJTPO CFUXFFO UIBU 40* JOEJDBUPS TIPXO BCPWF
BOE /"4"T (MPCBM
-BOE0DFBO5FNQFSBUVSF*OEFY
FIGURE 15 -
Oscillation Index
56
We can easily see that the biggest part of the rise in global temperatures
has been since the 1970s, a period that has also seen this SOI indicator rising
JOXBZTOPUTFFOCFGPSFUIFT4PIBWJOHNPSF&M/JPZFBSTJTSFMJBCMZ
associated with higher global average temperatures (AKA global warming).
5IBUTBMMSFBMMZJOUFSFTUJOHGPSNFUFPSPMPHJTUT
CVU'JHVSFIJHIMJHIUTXIZJUJT
important to those who use technical analysis for stock prices:
The idenTificaTion
and
of balanced
ladder levels
in
lookinG
TouTilizaTion
our oWn PlaneT
for markeT
insiGhTs
The Technical analysis of sTock and markeT index Time series
FIGURE 18 -
FIGURE 16 -
1980-2012
(MPCBM UFNQFSBUVSF EBUB IBWF B QSFUUZ OJDF DPSSFMBUJPO UP UIF BQQSFDJBUJPO JO
stock prices during the 20th century. It is not a perfect correlation, but it is wellcorrelated enough over this long time span to at least be worth knowing about.
5IF1FBSTPOT$PSSFMBUJPO$PFDJFOUGPSUIFZFBSQFSJPETIPXOJT
showing the strong positive correlation. The big rise in the stock market
during the 1980s and 1990s matched a general rise in global average
UFNQFSBUVSFT BCPVU ZFBST FBSMJFS &WFO UIF NBKPS EJQT IBWF CFFO PO
schedule, although the magnitude of the temperature dips has not always
NBUDIFEUIFNBHOJUVEFPGUIFTUPDLNBSLFUTSFTQPOTF5IFCPUUPNPGUIF
DSBTI XBT ZFBST BGUFS B CPUUPN JO HMPCBM UFNQFSBUVSFT TIPXO BT
CFJOHBMJHOFEJOUIJTDIBSU
EVFUPUIFZFBSPTFUPGUIFQMPU
5IFTUPSZJT
UIFTBNFXJUIUIFCPUUPN5IFTUPDLNBSLFUTEJQJOXBTNVDI
larger than this data implied it should have been, which just goes to show
what a big anomaly that financial crisis actually was.
The correlation since 2007 has not been as good as at other times, which
TVHHFTUTUIBUUIF'FEFSBM3FTFSWFIBTCFFOIBWJOHBOPVUTJ[FEJNQBDUPOUIF
NBSLFUTJODF%S#FSOBOLFUPPLPWFSBTDIBJSNBO*GUIF'FEFWFSSFUVSOTUP
BQPTUVSFPGIBWJOHBMFTTPWFSUJOVFODFVQPOUIFTUPDLNBSLFU
XFTIPVME
expect the correlation to improve again as natural market forces dominate
rather than governmentally induced artificial forces.
FIGURE 17 -
#ZTIJGUJOHGPSXBSEUIFHMPCBMUFNQFSBUVSFEBUBCZZFBST
BTTIPXOJO'JHVSF
17, we can clean up the slight correlation mismatch between this temperature
EBUBBOEUIFCFIBWJPSPGUIF415PTBZJUNPSFTJNQMZ
WBSJBUJPOTJOHMPCBM
average temperatures have been LEADING the movements of stock market
CZBCPVUZFBST
BUMFBTUEVSJOHUIFQBTUDFOUVSZQMVTGPSXIJDIEBUBFYJTU
Zooming in more closely, we can see how this relationship has acted over the
QBTUEFDBEFTJO'JHVSF
Now, here is an ironic twist to this relationship. There are many people who
are concerned about the risks to the planet (and humanity) from global
warming, and they are advocating for taking steps which they believe will
halt the warming trend and cool down the planet to some level that they
believe is more optimal. If one supposes that they are right, and if humanity
could somehow find a way to bring about a cooling trend for global average
temperatures over a period of several years, then the implication of the
correlation shown above is that such a change would also bring about a
multi-year bear market for stock prices. So if one is both an investor and
an opponent of global warming, then that involves mentally rooting for two
PVUDPNFTXIJDIBSFTUBUJTUJDBMMZJODPOJDUXJUIFBDIPUIFS
2014 . Issue 68
57
'JHVSF TVCTUJUVUFT /"4"T UFNQFSBUVSF JOEFY EBUB XJUI UIF 4PVUIFSO
Oscillation Index (SOI) data discussed above. It is not a perfect correlation,
but it is at least good enough to show is that there is something going on
there. The big growth years for stock prices in the 1980s and 1990s matched
BQFSJPEPGQSFEPNJOBOUMZ&M/JPDPOEJUJPOT
GBDUPSJOHJOUIBUZFBSMBH5IF
leveling off of the SOI indicator over the past decade has correctly foretold the
leveling off of the stock market over the same time period, again once we
BEKVTUGPSUIFZFBSMBH
FIGURE 20 -
58
FIGURE 21 BHSFBUFSJOVFODFPOUIFFDPOPNZ5IF'FEFSBM3FTFSWFJOUIFMBUFTXBT
USZJOH UP QSJDL UIF *OUFSOFU CVCCMF
BOE UP HIU JOBUJPO XIJDI XBT OPU ZFU
evident but which was thought to lie over the horizon6 . So corporate profits
BTBQFSDFOUBHFPG(%1TUBSUFEGBMMJOHBIFBEPGUIFQPJOUJOUJNFXIFOUIF
Southern Oscillation Index said such a drop should have started. But the
bottom for corporate profits in 2002 was right on schedule.
In a similar way, corporate profits started falling in 2006, ahead of this
TDIFEVMF
BOEBUBUJNFXIFOUIF'FEFSBM3FTFSWFXBTBUUFNQUJOHUPQSJDL
the housing bubble. The Fed succeeded in that effort, and then in 2008
when this Southern Oscillation Index model said profits should fall, the
EFTDFOUXBTNBHOJFE1SPUTIBWFCFFOSFCPVOEJOHTJODFUIFCPUUPNJO
The idenTificaTion
and
of balanced
ladder levels
in
lookinG
TouTilizaTion
our oWn PlaneT
for markeT
insiGhTs
The Technical analysis of sTock and markeT index Time series
2PG5IFDPOUJOVFESFCPVOEJOQSPUTBTBQFSDFOUBHFPG(%1EVSJOH
2011-12 comes at a time when this model says that they should still be under
downward pressure, and that continued rebound is likely a product of the
'FETFBTZNPOFZQPMJDJFT 2&
FUD
DSFBUJOHGVSUIFSJNCBMBODFTJOUIF
financial markets.
arcTic ice
5IF/BUJPOBM4OPXBOE*DF%BUB$FOUFSIBTCFFOQVCMJTIJOHEBUBPOCPUIUIF
area and the extent of arctic sea ice since 19787. Not surprisingly, the ice
waxes and wanes on an annual cycle, with the peak for both ice area and ice
extent usually coming in the month of March, as seen in Figure 22.
conclusions
The relationship of the financial world to the geophysical world has a more
significant linkage than many believe, if we are to accept the message
of the data presented above. The linkage between the financial markets
and sunspots, rainfall, temperature, and other data likely runs through
the agricultural sector, with better or worse crop yields likely affecting the
financial markets more than is appreciated by some analysts. There is even
BO BSNBUJWF TDJFOUJD CBTJT GPS UIF SFMBUJPOTIJQ CFUXFFO TVOTQPUT BOE
SBJOGBMM
BTSFWFBMFECZUIFSFTFBSDIPG%BOJTIBUNPTQIFSJDSFTFBSDIFS)FOSJL
Svensmark. To overly simplify his findings: More sunspots equals less cosmic
rays hitting the upper atmosphere, which then equals less rain, which affects
crop production, food prices, wealth, liquidity, and ultimately the financial
markets. The bountiful stock market returns of the 1980s and 1990s were
associated with a huge solar and atmospheric anomaly, causing a big upward
deviation from the bigger scheme of how average global temperatures look
PWFSUIFSFBMMZMPOHSVO %JTDPWFS.BHB[JOF
Even if we cannot establish a causal relationship, we can at least accept the
overwhelming evidence of co-relationship, and infer that there must be
some underlying causal relationship. Better still, we can use that information
to aid in market forecasting.
2014 . Issue 68
59
references
Bureau of Economic Analysis
http://www.bea.gov/iTable/iTable.cfm?ReqID=9&step=1
Cycles Research Institute
http://cyclesresearchinstitute.org/cycles-history/chizhevsky1.pdf
Discover Magazine
http://discovermagazine.com/2007/jul/the-discover-interview-henriksvensmark#.UN6QbawVXB0
Government Pringing Office
http://www.gpo.gov/fdsys/pkg/CHRG-108shrg96536/html/CHRG108shrg96536.htm
Kessler, Billy
http://faculty.washington.edu/kessler/occasionally-asked-questions.html#q2
National Oceanic and Atmospheric Administration
ftp://ftp.ngdc.noaa.gov/STP/SOLAR_DATA/SUNSPOT_NUMBERS/MONTHLY
National Oceanic and Atmospheric Administration
http://www.ncdc.noaa.gov/stormevents/ftp.jsp
Snow and Ice Data Center, University of Colorado Boulder
ftp://sidads.colorado.edu/DATASETS/NOAA/G02135/
60
Parameter-results stability:
A New Test of Trading Strategy Effectiveness
connors research
bioGraPhies
Larry Connors
How
Markets Really Work, Short Term Trading Strategies That Work, High Probability ETF Trading,
Guidebook High Probability Trading with Multiple Up & Down Days.
ETF Trading with Bollinger Bands, Options Trading with ConnorsRSI, Trading Leveraged
ETFs with ConnorsRSI, and ETF Scale-In Trading.
absTracT
This study provides the results of a trading strategy that enters positions
when a stock becomes oversold and exits when the oscillator used to define
the oversold condition returns to a neutral level. A method of analyzing test
results from a practitioners perspective will also be presented. This practical
approach to strategy analysis could be adopted by traders seeking to verify
their strategy has a high probability of being as profitable in the future as it
was when applied to back tested data.
Few studies address short-term trading strategies from a traders perspective.
Because the traders perspective is different from the researchers perspective,
there is a need for new ways to evaluate the effectiveness of trading
strategies. While the Sharpe Ratio, excess returns and other statistical tests
offer important information, parameter-results stability is also important.
Parameter-results stability provides traders with a means to evaluate the
robustness of the general idea underlying the trading strategy. This paper
explains the relationship between parameter selection and strategy results
while demonstrating that the relationship should be predictable.
Parameters are the numeric variables of a rule that is part of a trading strategy.
For example, many strategies rely on moving averages and the number of days
or months used to calculate the moving average is a parameter. It is possible to
test a wide range of parameters for a given strategy to find one that works, a
problem known as data mining. Strategies based on data mining are unlikely
to work in the future since they are finely tuned to the past. A sound strategy
with a high probability of future success should be based on parameters that
are selected with a rational and logical approach. The parameters defining
a sound strategy should also exhibit a degree of stability, meaning a small
change in the parameter will result in only a small change in the test results.
To extend the moving average example, a sound strategy would be one that
showed little variability in performance for a moving average calculated for 10
months, 9 months, 8 months, 11 months and 12 months. The small change in
the value of the calculation should have a minimal impact on profitability. If
the 10-month moving average is profitable but the nearby parameters of 8, 9,
11 and 12 months are not, the strategys success is most likely due to luck rather
than a sound idea. This strategy would be unlikely to be profitable in the future.
A parameter can be considered to be stable when small changes in its value
have only a small impact on the strategys performance. While the topic
of parameter stability should be important to traders, many traders use
optimization to select strategy parameters. Optimization involves finding
the parameter that meets a limited objective. Optimization might be used to
maximize the percentage of winning trades, for example. Optimized strategies
may not enjoy stable results in the future because they are finely tuned to the
price history used in the testing. Rather than optimization, traders should
consider testing that offers insights into how small changes in the strategys
parameters impact key performance metrics. If small changes in parameters
result in relatively small changes in performance, the strategy should be more
likely to be stable in the future since this approach is not finely tuned to the past.
when a stock becomes oversold and exits when the oscillator used to define
the oversold condition returns to a neutral level. A method of analyzing test
results from a practitioners perspective will also be presented. This practical
approach to strategy analysis could be adopted by traders seeking to verify
their strategy has a high probability of being as profitable in the future as it
was when applied to back tested data.
i. liTeraTure review
Traders and researchers have different perspectives and those differences are
apparent in their published work. Short-term trading is largely unexplored in
the academic literature. Conrad and Kaul (1998) ignore short-term strategies
after demonstrating that trend following fails to capture profits with a oneweek holding period. This result is the most statistically significant finding in
their landmark paper but has been overlooked by most researchers. Instead
the focus has been on the papers conclusion that trend following strategies
can deliver statistically significant profits over time periods ranging from three
to twelve months.
The fact that trend following failed to work over one-week time periods leads
to the question of whether contrarian, or mean reversion strategies, could be
profitable in the short term. This question has not been the subject of much
research.
Practitioners have devoted some attention to the question of what works in
short-term trading but it is important to note there are significant differences
between academic research and practitioner-published research. There is
generally no peer review of practitioners work and there is no requirement
to include test results in the work. This leads to well-selected examples being
offered as the only proof of many ideas or test results that fail to cover a
sufficient time period. A final difference is that while academic literature will
include a number of papers explaining ideas that fail to work, practitioners will
generally only publish successful ideas.
Practitioner studies related to mean reversion strategies with a holding period
of approximately one week do not seem to have been widely published.
Among notable exceptions are Connors and Raschke (1996) and Connors and
Alvarez (2008). Although a number of other books have been written about
short-term strategies, they tend to rely on the well-selected example as proof
of success rather than quantified testing with a large sample size.
This study provides the results of a trading strategy that enters positions
62
Parameter-results stability
A new Test of Trading Strategy Effectiveness
connorsrsi
bucket
5-day return
30
ConnorsRSI and the 2-period Relative Strength Index (RSI). The 2-period RSI, or
RSI(2), is more responsive to price changes than the traditional 14-period RSI.
Figure 2 summarizes the test results and shows that ConnorsRSI identifies
trading opportunities that have a larger average return. This is true for both
oversold and overbought stocks.
Figure 1 displays the frequency of occurrence for each bucket. It can readily
be seen that ConnorsRSI rarely reaches extreme values. ConnorsRSI readings
below 5 occurred just 0.63% of the time in testing. Readings below 15, the first
three buckets shown in Table 1, indicate a stock is oversold and occurred 6.46%
of the time. Readings above 85, the last three buckets in Table 1, indicate a
stock is overbought and occurred 6.39% of the time in the test.
Testing has demonstrated that ConnorsRSI is reliable and testing has also
revealed that trade signals will occur infrequently, an important consideration
for traders who have limited resources. Additional testing is needed to
determine whether or not ConnorsRSI offers any benefit relative to more widely
followed indicators. This test was completed by Radtke (2014) and compared
the five-day average return for buckets of stocks with identical readings of
2014 . Issue 68
63
the next day. Trades will be simulated assuming entries are made 2, 3 and
5% below the previous days close. All trades will be closed when ConnorsRSI
returns to a neutral level. Two levels, values of 50 and 70, will be tested.
By using a variety of parameters we will be able to evaluate whether or not
the underlying idea that oversold stocks can profitably be traded on the long
side is valid. We would expect to see only small changes in the results as the
parameters change if the idea is valid. This test will focus on the percentage
of winning trades. A parameter will be considered stable if small changes in
the parameter have only a small impact on the percentage of winning trades.
In summary, testing will be done on the setup condition, the entry level
and the exit to identify how robust the strategys rules are to changes in the
parameters. Results will be quantified with performance metrics to analyze
whether or not parameter selection plays a predictable role in a well-designed
trading strategy.
Performance metrics will include the total number of trades, the percentage
of winning trades and the average percentage return of all trades. These are
metrics that are of significance to traders.
Risk is also an important metric and that is incorporated, from a traders
perspective, into the average percentage return of all trades. A high percentage
return indicates that the size of the winning trades is cumulatively larger than
the losing trades. The Sharpe ratio and profit factor are also provided. To find the
Sharpe ratio, the average percentage return and standard deviation of returns
is calculated. These two figures are annualized. The risk free rate of return is
subtracted from the annualized average return and this value is then divided
by the annualized standard deviation of returns to obtain the Sharpe ratio. The
profit factor is the dollar amount of the profit from winning trades divided by
the total dollar value of all losing trades. These values were calculated with
AmiBroker, the software used for this test.
iii. resulTs
In this section, we present the results of back testing. We expect to see that
more restrictive strategy rules decrease the frequency of trading activity
but increase the percentage of winning trades. More restrictive exits should
increase both the percentage and average size of winning trades. If the
strategy is based on a sound idea, these changes should occur in a nearly
linear fashion and the results of all trading strategy variations should be fairly
consistent. From a traders perspective, there is no need to develop an objective
definition of nearly linear and fairly consistent. Since the goal of a trader
is to make money, consistent profitability is the only standard the results
must meet. Unstable results (for example, a single parameter set that results
64
Parameter-results stability
A new Test of Trading Strategy Effectiveness
#
Trades
Avg %
Profit /
Loss
Avg
% of
Trades /
Winners
Year
5,690
1.96%
67.1%
2,789
3.14%
941
4.72%
Sharpe
Ratio
Profit
Factor
CRSI
Buy
#
Trades
Avg %
Profit /
Loss
Avg
% of
Trades /
Winners
Year
429.4
1.18
1.78
15
941
4.72%
72.1%
68.7%
210.5
1.69
2.64
10
947
3.95%
72.1%
71.0
2.73
3.41
1,674
Sharpe
Ratio
Profit
Factor
CRSI
Buy
71.0
2.73
3.41
70
70.6%
71.5
3.80
4.10
50
2.51%
68.0%
126.3
1.66
1.97
70
1,683
2.20%
67.6%
127.0
2.54
2.30
50
2,493
1.64%
66.7%
188.2
1.16
1.51
70
2,508
1.56%
67.3%
189.3
2.09
1.92
50
One factor not shown in the table is that these are short-term trades with an
average holding period of less than five trading days. This general holding
period is applicable to all test results presented in this section.
Sharpe
Ratio
Profit
Factor
CRSI
Buy
Once again, a clear pattern emerges. Restrictive rules reduce the frequency
of trading relative to less restrictive parameters and more restrictive rules
generally increase the probability of success. The least restrictive entry (a limit
price 2% below the previous close or Y = 2 in Rule 3) enjoys slightly more
winning trades than the more restrictive exit for that entry parameter. This fact
does not materially impact the conclusions of the testing since the relationship
between parameter selection and results is still relatively stable. Of all the
parameters tested, the exit parameter has the least impact on trade results.
188.2
1.16
1.51
70
All eighteen possible permutations of the trading strategy are shown in Table V.
68.0%
126.3
1.66
1.97
70
72.1%
71.0
2.73
3.41
70
Table II holds the ConnorsRSI buy level constant at 5 (X = 5 in Rule 2) and retains
the same exit threshold of Z = 70 from the previous test. The percentage of the
limit price used to enter a trade is varied in this test (Y in Rule 3).
#
Trades
Avg %
Profit /
Loss
Avg
% of
Trades /
Winners
Year
2,493
1.64%
66.7%
1,674
2.15%
941
4.72%
Results again confirm expectations. There are more trade signals with a less
restrictive entry parameter but those signals have a lower probability of
success when compared to signals given by more restrictive signals.
In Table IV, a summary of the impact of the exit parameter is presented. This
test holds the ConnorsRSI buy level constant at 5 (X = 5 in Rule 2) but varies
both the percentage of the limit price used to enter a trade (Y in Rule 3) and
the value of the threshold trigger used in the exit rule (Z in Rule 4). Results are
sorted by the value of the entry limit price and then the sell threshold.
Table V: All trade results. This table shows the results of all permutations for all
parameters.
Table V is sorted by the percentage gain of the average trade. Strategies with
the most stringent setup conditions, entry rules and exit criteria consistently
outperform strategies with less strict criteria. All strategies display a
remarkably consistent percentage of winning trades and trading frequency is
directly correlated with the strictness of the strategy parameters.
Using the results from Table V, we can answer the question of whether or not this
strategy is useful to traders. We can conclude that this strategy is robust since
small changes in parameters result in small changes in performance. This was
confirmed by running a series of different tests that vary individual parameter
and observing a high degree of results stability. With high parameter-results
stability, we conclude the strategy is robust and tradable.
It is important to note that actual performance would depend upon trading
costs. Trading costs may vary for each individual but French (2008) found that
2014 . Issue 68
65
#
Trades
Avg %
Profit /
Loss
Avg
% of
Trades /
Winners
Year
941
4.72%
72.1%
947
3.95%
2,789
Profit
Factor
CRSI
Buy
71.0
2.73
3.41
70
70.6%
71.5
3.80
4.10
50
3.14%
68.7%
210.5
1.69
2.64
70
1,674
2.51%
68.0%
126.3
1.66
1.97
70
2,828
2.43%
68.1%
213.4
2.29
2.70
50
1,683
2.20%
67.6%
127.0
2.54
2.30
50
5,690
1.96%
67.1%
429.4
1.18
1.78
70
5,760
1.80%
67.9%
434.7
1.19
2.06
70
5,916
1.74%
68.0%
446.5
1.83
1.88
50
2,493
1.64%
66.7%
188.2
1.16
1.51
70
2,508
1.56%
67.3%
189.3
2.09
1.92
50
5,875
1.48%
67.7%
443.4
1.79
2.21
50
12,430
1.31%
67.7%
938.1
0.95
1.60
70
8,919
1.25%
67.8%
673.1
0.97
1.74
70
13,027
1.10%
67.3%
983.2
1.48
1.70
50
9,135
1.07%
67.0%
689.4
1.53
1.93
50
19,582
0.95%
67.6%
1,477.9
0.76
1.51
70
20,730
0.80%
66.7%
1,564.5
1.26
1.57
50
the average cost of active trading is 0.67%. It is reasonable to assume that the
costs would be lower now since commission rates have dropped since that study
was published. However, using the value French reported, each variation of this
trading strategy would be profitable for a trader with the lowest performing
variation of the strategy offering an edge of 0.13% after trading costs under
Frenchs model. The best performing variation would have an historical edge of
4.05% per trade.
aPPendix: ConnorsRSI
CLOSINGPRICE
$20.00
$20.50
$20.75
$19.75
-1
iv. conclusions
$19.50
-2
This paper demonstrates that a sound trading strategy should provide results
that vary slightly when the strategy parameters are varied by a small amount.
In doing so, this paper adds to the body of knowledge of technical analysis by
providing a framework to test a trading strategy for robustness. This type of
testing would be conducted by practitioners who should ensure that a high
$19.35
-3
$19.35
$19.40
66
STREAK DURATION
Parameter-results stability
A new Test of Trading Strategy Effectiveness
The second aspect of the solution is to apply the RSI calculation to the set of
Streak Duration values. By default, ConnorsRSI uses a 2-period RSI for this
part of the calculation, which we denote as RSI(Streak,2). The result is that
the longer an up streak continues, the closer the RSI(Streak,2) value will be
to 100. Conversely, the longer that a down streak continues, the closer the
RSI(Streak,2) value will be to 0.
The result is a very robust indicator that is more effective than any of the three
components used individually. In fact, ConnorsRSI also offers some advantages
over using all three components together. When we use multiple indicators to
generate an entry or exit signal, we typically set a target value for each one.
The signal will only be considered valid when all the indicators exceed the
target value. However, by using an average of the three component indicators,
ConnorsRSI produces a blending effect that allows a strong value from one
indicator to compensate for a slightly weaker value from another component.
A simple example will help to clarify this.
The final ConnorsRSI calculation simply determines the average of the three
component values. Thus, using the default input parameters would give us the
equation:
ConnorsRSI(3,2,100) = [RSI(Close,3) + RSI(Streak,2) + PercentRank(100)] / 3
Lets assume that Trader A and Trader B have agreed that each of the following
indicator values identify an oversold condition:
t34* $MPTF
t34* 4USFBL
t1FSDFOU3BOL
Trader A decides to take trades only when all three conditions are true. Trader B
decides to use ConnorsRSI to generate her entry signal, and uses a value of (15
+ 10 + 20) / 3 = 15 as the limit. Now assume we have a stock that displays the
following values today:
t34* $MPTF
t34* 4USFBL
t1FSDFOU3BOL
t$POOPST34*
Trader A will not take the trade, because one of the indicators does not meet
his entry criteria. However, Trader B will take this trade, because the two low
RSI values make up for the slightly high PercentRank value. Since all three
indicators are attempting to measure the same thing (overbought/oversold
condition of the stock) by different mechanisms, it makes intuitive sense to
take this majority rules approach.
The default Percent Rank look-back period used for ConnorsRSI is 100, or
PercentRank(100). We are comparing todays return to the previous 100
2014 . Issue 68
67
references
Connors, Laurence and Alvarez, Cesar, 2008, Short Term Trading Strategies That
Work (TradingMarkets, Jersey City, NJ)
Connors, Laurence and Raschke, Linda Bradford, 1996, Street Smarts: High
Probability Short-Term Trading Strategies (M. Gordon Publishing Group, Jersey
City, NJ)
Connors Research, LLC; Connors, Laurence; Alvarez, Cesar and Radtke, Matt,
2014, An Introduction to ConnorsRSI, 2nd edition (TradingMarkets, Jersey City,
NJ)
Conrad, J., and Kaul, G., An Anatomy of Trading Strategies, Review of Financial
Studies (1998) 11 (3): 489-519.
French, Kenneth R., The Cost of Active Investing (April 9, 2008). Available at
SSRN: http://ssrn.com/abstract=1105775
Jordan, Douglas J., and Diltz, J. David, 2003, The profitability of day traders,
Financial Analysts Journal, 59(6), 8594.
Radtke, Matt. (January 2, 2014). Compare ConnorsRSI to RSI2. InTradingMarkets.
com. Retrieved July 25, 2014, from http://www.tradingmarkets.com/
analytics/how-does-connorsrsi-compare-to-rsi2-1583255.html.
68
bioGraPhies
G.L. Biff Robillard III CMT. President & Co-Founder, Bannerstone Capital Management.
Biff Robillard is a Chartered Market Technician, portfolio manager and a principal of Bannerstone Capital Management, LLC.
Bannerstone is a registered investment advisor in Deephaven, Minnesota.
Thomas C. Pears
absTracT
inTroducTion
Leveraged exchange-traded funds (LETFs) were first introduced in 2006
(McCall, 2011). By holding swaps, futures, and other derivatives of a given
index, LETF managers are able to create products that provide daily returns
equal to an intended multiple of that underlying indexs daily return. Long,
or bull, LETFs generate two or three times the underlying indexs daily return,
while inverse, or bear, LETFS provide a leverage factor of negative two or
negative three. Since these financial products are engineered to deliver
amplified daily returns, the multi-day return of an LETF will usually differ from
the return of the underlying index times the leverage factor. During periods
of high volatility, an LETFs return will often fail to achieve the intended
multiple. This is because the effects of volatility drag are amplified by
leverage. Volatility drag is the price decay that occurs during volatile periods
with equal daily percent gains and losses for instance, if an investment
gains 10% and then loses 10% the next day, the overall return is not 0%, but
1%. If you have ever taken the average of your investments daily returns
and found that your investment actually performed much worse, you know
exactly how detrimental to return this decay can be.
LETF underperformance comes from the unique way in which these
products are levered. In order to achieve the same multiple of daily returns
every day, LETFs leverage needs to be rebalanced in between trading
sessions. Understanding the difference between this type of leverage and
conventional margin is essential to understanding the inherent decay risk of
LETFs. Levering with 50% margin creates a 2X-levered investment initially,
but once the price changes, the leverage changes as well. In this way, a static,
conventional margin loan creates variable leverage that changes from day to
day. Suppose an unlevered index ETF is priced at $100. We buy one share of
this ETF, using $50 of equity and $50 of debt. If the ETF rises to $125 (a 25%
gain) our equity will increase by $25 (a 50% gain). We achieved twice the
return of the index, but now we have $75 of equity invested in our $125 ETF;
our leverage factor is now 1.67.
Conversely, LETF managers rebalance leverage between each trading
session. In this way, one could consider LETF leverage, adjusted to maintain
a particular leverage factor, to be fixed leverage. To achieve fixed leverage,
managers increase their exposure when the LETF rises and decrease their
exposure when it declines; this can create a buy high, sell low pattern that
reduces return during high By 1936 Keynes volatility periods. For instance,
when the index goes up during the day, a long LETFs exposure is increased
that night, essentially doubling down the traders directional bet (the same
effect would be achieved by buying the index ETF on margin and adjusting
the leverage at the end of every trading session to maintain the same
leverage ratio). This style of leverage allows investors to stay leveraged as
the underlying index trends upwards, similar to the trading technique of
pyramiding. However, if the index then reverses, the increased exposure,
achieved by re-leveraging at a higher price, works against the investor, and
often causes the leveraged ETF to decay while the index experiences nondirectional volatility.
day
100
100.0
100
105
5.00%
115.0
15.00%
85.0
-15.00%
100
-4.76%
98.6
-14.29%
97.1
-14.29%
105
5.00%
113.4
15.00%
82.6
-15.00%
100
-4.76%
97.2
-14.29%
94.4
-14.29%
105
5.00%
111.7
15.00%
80.2
-15.00%
100
-4.76%
95.8
-14.29%
917
-14.29%
105
5.00%
110.1
15.00%
77.9
-15.00%
100
-4.76%
94.4
-14.29%
89.1
-14.29%
105
5.00%
108.6
15.00%
75.7
-15.00%
10
100
-4.76%
93.1
-14.29%
86.5
-14.29%
70
FIGURE 1 - ABC, BULL & BEAR price over the 10-day period.
brought ABC to a value of 105, but it only took a percent decrease of 4.76% to
bring it back down to 100.
The relationship between equivalent percent gains and losses is expressed by
the following equation:
EQUATION 1:
1
D+1
Where:
U = Percent change on the day the price goes up
D = Percent change on the day the price goes down
purposes we simply subtract the expense ratio for the period, E, from our
n
, where
return equation. We define E with the formula
252
e is the annual expense ratio of the LETF. For a full explanation of how we
derived this formula, see Appendix A.
To model the relationship between LETF return and volatility, we now express
index volatility in terms of A when return is R. For this study, we used
realized volatility, as opposed to the more widely-used historical volatility.
Historical volatility is more appropriate for directionally trending markets,
while realized volatility is more appropriate for volatile, non-directional
markets; we concluded that the latter would be more appropriate for a
study of volatility drag. Nonetheless, we did not find a significant difference
between our hypothetical historical volatility and realized volatility results,
and we believe that a practitioner wishing to apply our models to an
investment strategy could use the historical volatility tool that is available on
most financial software. The realized volatility daily formula, as described by
the VolX Group (2013), is displayed on the following page.
Our example demonstrates that decay in the value of leveraged ETFs is directly
correlated with index volatility: higher volatility leads to higher decay. Now we
will explore the mathematical relationship between index volatility and LETF
return. To do this, we build equations for volatility and LETF return that are
based on our 10-day hypothetical scenario (Figure 1 and Table I). To simplify,
let us reduce our reference period to two days. Since the 10-day simulation was
simply a 2-day pattern repeated five times, this will not affect the relationship
between volatility and LETF return.
Using an independent variable A to represent the daily index price changes,
we construct an equation that gives us LETF return in terms of these daily
price movements. Equation 2, below, gives LETF return over n number of
days when the index follows our 2-day example with a 2-day price change
represented by the parameter R. This formula takes the geometric mean of
two factors: the percent increase on the first day, and the percent decrease
on the second day. Multiplying one plus part one by one plus part two
gives us our 2-day return, plus one. Because of the commutative property
of multiplication, the order in which the price movements occur does not
affect return. Hence, the product of these two halves, minus one, represents
the n-day return of an LETF whose underlying index has a return of R and
average daily price movements of A.
Fund expenses must also be considered when investing in LETFs, and
generally cost 90110 basis points per year. Expenses reduce the net asset
value of the LETF every day after close. Thus, the effect of the expenses is
not significantly dependent on the indexs price movements, and for our
EQUATION 2:
Where:
= LETF return over the reference period
A = The indexs price gain on day one
L = The leverage factor of the LETF
n = The number of trading days in the reference period
E = The expense ratio of the LETF for the period
R = Index price change over the 2-day period, as calculated by the formula:
Where:
EQUATION 3:
Where:
V = Realized volatility
252 = The approximate number of trading days in a year
t = A counter representing each trading day
n = Number of trading days in the reference period
Ln = Natural logarithm
= Index price at close on day t
= Index price at close on the day preceding day t
The equation below gives us realized volatility in terms of daily price change,
A, when the indexs 2-day price change is R. For a full explanation of how we
2014 . Issue 68
71
EQUATION 4:
72
from 0 to 5. The more values you use, the smoother the curves will be
increments of .02 should be sufficient. Next to the A column, create a column
to hold volatility values for the given A value, using Equation 4. Equation 5,
below, shows one way of writing this as an Excel formula.
EQUATION 5:
Where:
Column B contains the corresponding realized volatility values and B6 is the
cell containing this specific formula.
Column A contains the A values, and A6 contains the A value for which this
formula is calculating a realized volatility level.
is the fixed cell containing the indexs annual return.
The formula for BULL return should be written in column D, and is based off
Equation 2. As an excel formula shown on the following page.
EQUATION 6:
Where:
C is the column containing BULL return values and C6 is the cell containing
this specific formula.
is the fixed cell containing the leverage factor for the LETF.
is the fixed cell containing the annual expenses of the LETF.
In column D, the formula for BEAR return should be the same, but should
reference column D instead of column C. The cell containing the leverage
factor, in this case
, should be negative.
You can also create a column for the non-levered ABC return, which always
equal index return minus index expenses. The formula in this case would be
. Notice that the formula does not refer to the independent
variable A, since ABC return is unaffected by volatility. Table II, below, shows
the cells that contain the various formulas above and the referenced user-
2014 . Issue 68
73
defined parameters.
Graphing BULL, BEAR, and ABC return (columns C, D, and E) in terms of realized
volatility (column B) will produce graphs like Figures 25. You can use these
graphs to find the theoretical return for LETFs at different index return and
volatility levels. You can also use the LINEST function in Excel to extract the
coefficients of a trendline, which will allow you to create a formula for LETF
return in terms of index volatility. A fourth-degree polynomial trendline
usually has an R-square in excess of 0.999, and can be used to easily and
precisely find LETF return for a corresponding volatility level.
Remember that these return levels are theoretical, and tracking error can
cause your actual return to differ from the model. Nonetheless, simply
studying the relationship between volatility and LETF return for various index
return and leverage parameters is extremely valuable for the LETF trader,
and allows for a better understanding of the unique behavior of financial
products that are re-leveraged on a daily basis.
74
dropped by more than 16%. Volatility rose considerably, breaking 20 for the first
time that year (Figure 6).
The graph above illustrates the estimated return for a 3X long LETF, given
a 30-day index return of 5%. Notice that if we simply wanted our bull to
outperform the index for this given index return, we would only need volatility
to be less than 54. However, since we desire additional compensation for our
added risk, we will instead find where the return curve intersects with our
break-even line. They cross when volatility is 34; we will use this volatility
level to build our oscillator.
Figure 9 reveals that volatility was indeed very high for several weeks,
causing our oscillator value to stay below zero. If we were not to adjust our
previous index return forecast, we could wait until volatility seems to be on a
downward trend. By examining the oscillator on a daily basis, we might have
determined that by about September 21st, volatility had declined to such a
level that use of LETFs to achieve our bullish return objective could then be
appropriate. If we suspected an index rally, we could go long a 3X bull ETF
after we see our oscillator cross the zero line. In our example, we buy shares
of UPRO, a 3X S&P 500 ETF, just before the market closes on September 22nd.
From September 22nd to November 3rd, the S&P 500 went from 1129.56 to
1261.65, an increase of 11.7% over the period of 30 trading days. 30-day
realized volatility for the index on November 3rd was 29.8. UPRO, our 3X
LETF, had a 30-day gain of 35.4%, just over three times the indexs return
(our model from part I predicts a 35.0% LETF return for these index return
and volatility parameters).
FIGURE 10 - S&P 500 and UPRO closing prices for our examples
holding period.
FIGURE 11 - S&P 500 and UPRO closing prices for August 15th
through November 3rd.
Figure 10, two above, plots the prices of UPRO and the S&P 500 over our
holding period. For comparison, Figure 11, above, shows their performance
starting on August 15th, when we may have invested if we had not consulted
2014 . Issue 68
75
the Voloscillator. UPRO still realizes substantial gains, but the high volatility of
August through mid-September causes noticeable decay in the value of the
LETF during that period. The tableIII compares our hypothetical investments
return for the two periods.
UPRO PRICE
9/22/11
1129.56
22.79
11/3/11
1261.15
30.85
Return
11.6%
35.4%
UPRO PRICE
8/15/11
1204.49
28.24
11/3/11
1261.15
30.85
Return
4.7%
9.2%
Average
Voloscillator
Value
Achieved LETF
Return
Multiple
Average
Voloscillator
Value
Achieved LETF
Return
Multiple
4.427
3.04
-0.696
1.96
TABLE III - Price and return for S&P 500 and UPRO for the two
investment periods. The lower part gives the average Voloscillator
value for the two periods, as well as the multiple of index returns that
UPRO achieved.
For Period B, the period starting on August 15th, the average Voloscillator
value was below zero, and UPRO return was 1.96 times index return, lower
than our selected risk margin of 2.2. Period A, on the other hand, had an
average oscillator value well above zero, and UPROs return multiple
actually exceeded the 3X daily leverage factor. By applying the Voloscillator
to this strategy, we avoided a highly volatile period that would have been
detrimental to our investment return, and our investment ended up
outperforming our objective.
Note that this example is an instance when the oscillator was effective
not as a return prediction tool, but as a mechanism to discern when LETF
investment was appropriate. While Period A happened to have a higher S&P
500 return than did Period B, this was a coincidence our tool does not help
the user find periods of high return, but rather periods where volatility is
76
aPPendix
A. Multi-day return can be expressed as the product of daily percent changes.
If we break down our 10-day example to a 2-day scenario, there are only two
days we have to consider: the day that the index price moves up, and the day
the index price moves back down. Thus, if we know an LETFs daily percent
changes for these two days, we can calculate the return of that product.
The equation below shows the relationship between 2-day return and daily
percent change.
Where:
r = 2-day return
u = Percent change on the day the price goes up
d = Percent change on the day the price goes down
Since LETF daily percent changes are simply multiples of index daily percent
changes, we can write a return equation for an LETF in terms of the index
daily percent changes based on the relationship we just defined.
Where:
Now that we have an equation for the return of the LETF in our 2-day
scenario, we need to find a way to relate our return equation to a realized
volatility equation. We will do this by defining the variables U and D in terms
of an independent variable, A.
Where:
Where:
R = Index price change over the hypothetical 2-day period
= Index return over the entire reference period
n = Number of trading days in the reference period
We rearrange this formula to find R in terms of r:
Where:
A = The indexs price gain on day one
Substituting our new values for U and D, we can get LETF return in terms of
our new variable, A.
We now have LETF return for when index return goes up by A on day 1 and
down by A on day 2. To make the model applicable to scenarios of more than
two days, we will raise the 2-day return to the power of /2.
Where:
= LETF return over the reference period
A = An independent variable representing the indexs price gain on day one
L = The leverage factor of the LETF
n = The number of trading days in the reference period
R = Index price change over the 2-day period, as calculated by the formula:
2014 . Issue 68
77
Where:
= Index return over the entire reference period
B.
We start with the realized volatility formula:
Where:
= Annualized LETF return
A = An independent variable representing the indexs price gain on day one
L = The leverage factor of the LETF
n = The number of trading days in the reference period
R = Index price change over the 2-day period, as calculated by the formula:
Where:
= Index return over the entire reference period
Where:
V = Realized volatility
252 = A constant representing the approximate number of trading days in a year
t = A counter representing each trading day
n = Number of trading days in the reference period
Ln = Natural logarithm
= Index price at close on day t
= Index price at close on the day immediately preceding day t
Since realized volatility is derived from the sum of squared daily returns, for
our 2-day scenario we can eliminate the sigma from the equation and simply
add the squared daily returns for the two days.
We can use the variable A to express the index price at the end of day 0, 1, and
2. Substituting in A and setting n to 2, we get:
78
Note that we take the realized volatility of the index, not the leveraged ETF,
so the L variable is not used.
references
McCall, Matthew, 2011, Leveraged ETFs: Are They Right For You?, Investopedia,
Web.
The VolX Group Corporation, 2013, RealVol Daily Formula (Realized Volatility
Formulas), Web.
$IBSMFT)%PX"XBSE
3 & $0 ( / * ; * / ( & 9$ & - - & / $ &
/PX"DDFQUJOH4VCNJTTJPOTGPSUIF$IBSMFT)%PX"XBSE
5IF$IBSMFT)%PX"XBSEIJHIMJHIUTPVUTUBOEJOHSFTFBSDIJOUFDIOJDBMBOBMZTJT5IF"XBSEFNCPEJFTFYDFMMFODFBOE
DSFBUJWJUZJOUIFFMEPGUFDIOJDBMBOBMZTJT8JOOJOHQBQFSTIBWFDSFBUFETVDDFTTGVMUSBEJOHTZTUFNT
JOTJHIUTJOUP
UIFPSJFTPGIPXNBSLFUTGVODUJPOBOEIBWFSFQSFTFOUFEUIFSJDIOFTTBOEEFQUIPGUFDIOJDBMBOBMZTJT
0GUIFBVUIPSTDPBVUIPSTXJOOFST
FJHIUIBWFHPOFPOUPQVCMJTICPPLTCBTFEPOUIFJSTVCNJTTJPOTUPUIF$IBSMFT
)%PX"XBSE8JOOFSTIBWFQSFTFOUFEBUNVMUJQMF.5"FWFOUTBOEXFSFNFOUJPOFEJOWBSJPVTNFEJBPVUMFUT
JODMVEJOH#BSSPOT
#MPPNCFSH/FXT
$/#$
$//.POFZ
*OWFTUJOH%BJMZ
.JOZBOWJMMF
5IF4USFFU
5IPNTPO3FVUFST
BOE5IF8BMM4USFFU+PVSOBM
5IFDPNQFUJUJPOJTPQFOUPBMMQSBDUJUJPOFSTBOEBDBEFNJDT5IFTVCNJTTJPOXJMMCFKVEHFECBTFEPOJUTBCJMJUZUP
FOIBODFUIFVOEFSTUBOEJOHPGNBSLFUBDUJPO
UIFDPODFQUTPGUFDIOJDBMBOBMZTJT
BOEUIPSPVHISFTFBSDI'PSUIF
DZDMFXFBSFBDDFQUJOHQSFWJPVTMZQVCMJTIFEOPODPNNFSDJBMXPSL'PSBEEJUJPOBMJOGPSNBUJPOPOUIF4UBOEBSETPG
+VEHNFOUJODMVEFEXJUIJOUIF(VJEFMJOFTGPS4VCNJTTJPOTWJTJUIUUQHPNUBPSH%PX"XBSE(VJEFMJOFT
5IFXJOOJOHTVCNJTTJPOJTEJTUSJCVUFEUPPVSWBTUNFNCFSTIJQTQBOOJOHDPVOUSJFTBOEBQQSPYJNBUFMZ
NFNCFST5IF"XBSEDBSSJFTBQSJ[FPG
BOEJTQSFTFOUFEBUUIF.5"(BMB"XBSET%JOOFSIFMEJO/FX:PSL$JUZ
JO.BSDI'PSNPSFJOGPSNBUJPOPOUIF$IBSMFT)%PX"XBSE
QMFBTFDPOUBDU"KBZ(+BOJ
$.5BU
%PX"XBSE!NUBPSH
'JOBM4VCNJTTJPO%FBEMJOF
+BOVBSZ
0VUMJOF4VCNJTTJPO%FBEMJOF
/PWFNCFS
4VCNJUUJOHBOPVUMJOFJTPQUJPOBM5IFPVUMJOFQSPDFTTXJMMIFMQDMBSJGZBOEPSHBOJ[FUIFDBOEJEBUFTUIPVHIUTXIJMFQSPWJEJOHUIFGPVOEBUJPOGPSMPOHFS
SFTFBSDI5IFDBOEJEBUFXJMMSFDFJWFDPNNFOUBSZPOUIFUPQJD
JUTTVJUBCJMJUZ
BOEUIFBQQSPBDIUIFDBOEJEBUFJTUBLJOH
Issue 68