December 2003
Vic fired a shot across the bow regarding the value, nay the
legitimacy of technical analysis. It elicited one of the best
point-counterpoints we’ve ever seen on the subject. The contributors included
TA notables and founders, pure quants, almatarians, down-on-the-luck
day-traders, and money managers with
billions under management. We memorialize it below and invite reader reactions.
Send comments to ![]()
J.T.:
A light
bulb went off the other night when I was reading in the wee hours of the
evening. I came across a story of
Cortez and how after he landed at Veracruz, the first thing he did was burn his
ships. Then he told his men: "You can either fight or you can
die." Burning his ships removed a third alternative: giving up and
returning to Spain. A lot of times it
takes more effort and creativity to shed myself of excuses than it does to come
up with an empirically laid out speculative idea. I have bad habits of "fundamental" and
"technical" analysis and these ships are and have been my excuses for
awhile. I know I must burn these ships
because they stand in my way most of the times from making the right
choice. The edge lies in the facts and
this is where I can conquer and succeed. (12/4/3)
R.M.:
I don’t want to step on any toes here, something I have a bad habit of
doing at times, but to get anywhere you have to take on risk. My loss of faith
in the value of technical analysis can be pinpointed to a day. I had gone to
the downtown library to research how to calculate the ADX indicator (this was
before everything was on the internet) [a female technician money manager] had
been on CNBC that week promoting Street Smarts and had thrown everyone a bone
by revealing her "Holy Grail" trade, which was to buy when the ADX
was above some value after a small correction followed by a move up that broke
the previous down day’s high, or something like that. I figured I’d test this
system myself on a few markets. While searching for "ADX" I found
plenty of articles on the inventor the indicator, Welles Wilder; unbelievable
stories about the computer he sold for $300,000 that would "pinpoint major
market turning points", along with myriad other scams he had pulled. I was
a bit confused. How could the indicator of such a well documented con man still
be in vogue?
Of course, the faithful are always ready with unassailable
defenses. The argument, "Just because the inventor of an indicator was a
con-man doesn’t in and of itself invalidate the value of the indicator,” is not
illogical. But you would think the situation would at least raise a red flag or
two, wouldn’t cha?
Anyway I tested out the Holy Grail on the markets I had data
for and found that it had no value whatsoever. But then again I was only
testing it on ag markets. Perhaps I was just applying the Holy Grail to the
wrong markets. I’ve tested some of Tom Demark’s strategies too, which also at
least have the virtue of lending themselves to testing, since they are
concretely spelled out without any positive results. Again, maybe I m just
testing them on the wrong markets, or over the wrong time frames.
My quandary, though, regards [the female technician]
herself. Unlike most sellers of systems and advice, she actually manages a fund
and appears
one side of fence or the other.)
One skeptical theory is this: she started out as a market
maker and was very successful as a market maker, but on the sidelines has found
it easier to make money as a vendor than as a trader, despite continuing to
some degree as a trader. I simply have a hard time swallowing that someone who
is a great trader would make their system as transparent to the public as she
does. Wouldn’t that violate the theory of ever-changing cycles? I mean, if her
service worked well why couldn’t a hedge fund just
Victor
Niederhoffer: Subject: poseurs and promoters.
Date: 12/05/2003 12:14:30
Let us be omnivorous in our quest for knowledge and
hypotheses from the like of Heiby, Merril, Wilder and female technician money
managers. But let us also be skeptical of anyone who would widely disseminate a
system to make money as these are subject to ever-changing cycles, and
anticipation and front running by fast moving and flexible competitors. And let us be wary of attempts
by those without proper humility who endeavor to lead us down the road to
ruination with suggestions that are not
We must walk between the Scylla of excessive veneration of
these idiot savants and their like who have not the tools or the inclination to
keep up with modern techniques for unraveling decision making under
uncertainty, and the Charybdis of dismissing proper statistical analysis of the
raw data they attempt to fit lines to in their role as promoter or poseur.
It is helpful in this regard to revisit often the skepticism
which Dr. Miller applies to such attempts even in the seemingly well regarded
field of poker. (12/5/3)
Ross
Miller:
Now that the word is out that all academic finance research
is disinformation, I'm wondering about poker theory. While guys like Sklansky
and Brunson make some good points, there is a lot of bad or very incomplete
advice in their books. Texas hold'em (with limits) seems like a much better
vehicle for training traders than
Victor
Niederhoffer:
In checking one's premises about the superior results of
this one or that one, do not downplay that [the female technician money
manager] is not only excellent as a speaker but she was very effective in
trying, nay importuning, the relevant parties not to publish
Practical Speculation
on
grounds it was not fair to technical analysis as well as other weaknesses she
doubtless found. Regrettably she was unsuccessful so we had to suffer the
indignity of seeing it wallow into nothingness after hitting the sales room at
Amazon (it never made it onto the shelves of 75% of B&Ns as far as I can
reasonably determine). One must admire
the short skirt pattern and the ilk in the context of the buzz about the
superior profits. (12/6/3)
Yale
Hirsch:
How did you ascertain that Welles Wilder was not a genuine
article and that he had definitely pulled scams? This is certainly news to me.
Technical analysis is a discipline for many thousands of people. It fits the
way they wish to follow the market and individual stocks. Counting may not
appeal to them. (12/6/3)
Bbands:
I cannot comment on Mr. Wilder's character, but "New Concepts in
Technical Trading Systems", his 1978 book, provided some important tools
and did much
RE: [the female technician money manager]
RE: Technical Analysis
Philip
J. McDonnell:
Some might claim that Technical Analysis and Counting are
both based on using past price, volume and other technical market statistics to
predict future price changes and are thus similar or even the same thing. However there is a fundamental
methodological difference between the two disciplines.
In traditional TA, charts are often used to prove an
hypothesis. Countists never use charts
as a form of proof. It simply isn't
allowed. In TA a few charts (culled
from thousands?) which
"illustrate" a principle are considered sufficient to
"prove" the system being claimed.
For countists the only acceptable proof is a study which uses the raw
numbers themselves and tests the validity of the hypothesis using valid
statistical tests. Ideally such tests would be non-parametric or bootstrap
methods which do not assume a normal distribution.
The reason charts are eschewed by countists is that the
human eye (and brain) has an amazing ability to discern patterns from visual
images. Sometimes these patterns are not even there. Perhaps the most famous example is of the two vases next to each
other, which can also be seen as two faces in profile looking at each
other. Which of the two perceptions is
right? I claim the third one - the
viewpoint of the countist is correct. They are simply data points - dots on a
graph.
This doesn't mean that countists never look at charts. I certainly look at many charts during my
trading day, but I never rely on them for trading. I use them only for a quick and dirty sense of what is going
on. From the occasional Bloomberg Grabs
posted by the Chair one would presume that he glances at charts from time to
time, but certainly not to prove the validity of a given trading strategy.
One significant problem with charts and chart patterns is
that some of the well known patterns are quite difficult to define in a
rigorous way. Consider defining a Head and Shoulders pattern for example. TA presumes it is a topping pattern. How wide should the left shoulder be in units
of time? How high in price? How high
the head, how wide in time and how high relative to the left shoulder and right
shoulder? What if the neckline is
sloping up or down, is there a slope that is too steep? What should the
TA would decide in hindsight that it was a top if the
telltale and eye attracting triple peaks visually stick out at the top of a
chart. If with perfect hindsight the
quadruple V's are sticking out the bottom the eye would classify it as a
quadruple bottom and never even notice the head and shoulders which is also
present. If you look at the questions I
raised above there are 12 in all. Most
of them require two answers - a minimum and maximum. That's about 20-24 variables in all. If you allowed me 20
To my mind that is the most persuasive argument against TA
chart analysis. There are simply too many subtle unrecognized variables in most
Chart
P.S. There are some good reviews of academic studies and
Chair's comments on
TA in Prac. Spec. p.95 et sequitur.
Bbands:
I am
afraid that I have to disagree. Even a cursory survey of the literature reveals
that charting is but one facet of TA. If you wish to address charting and
chartists--as they call them in the UK--fine, do so directly, but you should
not characterize the entire discipline for the action/beliefs of one branch.
The core of your post is an ad hominem attack on TA and is indefensible.
Labeling a package without a fair assessment of the contents is anathema that
deprives all of dignity and knowledge.
Dave g: The fundamental difference between TA and Counting goes deeper than the failure of TA to be "rigorous" and Counting’s reliance on statistical studies. Itis a right brain bias trader vs.. the left brain biased trader. The artist vs. the statistician. The TA artist will insist that rigorous definitions of patterns are not needed, all that matters is his / her ability to INTERPRET the forming patterns that will place the odds of success in their favor. The statistician countist, on the other hand, does not apply art or interpretation but rather hard data to place the odds of success in their favor.
It is an individual choice and these different philosophies
are what makes the market work. Here is
a link to an interesting MIT paper on the application of nonparametric kernel
regression on TA. The results are
indeed fascinating. (12/5/3)
http://web.mit.edu/alo/www/Papers/1705-1765.pdf
Bbands:
This is
patently absurd. There is nothing inherently right brain, left brain about TA
and in any case the right-left brain concept has been thoroughly
debunked elsewhere. There are many rigorous technicians, including systematic types that never employ charts. If you want a stalking horse you'll have to look elsewhere.
Philip
J. McDonnell: (an open post to) John,
I think your criticism of my post is fair in the sense that not all people who call themselves TA advocates employ visual chart interpretation. However I think it can be fairly said that chart pattern interpretation has traditionally been one of the cornerstones of TA and still has many adherents.
My main issue is that most TA types (but not all!) have been
content to accept a few chart examples as probative. Other issues include lack
of statistical testing, guruism (where only the guru has the knowledge to
interpret the tea leaves correctly) and the problem of very many variables
(which can seem to fit anything perfectly).
I believe there is a culture among many in TA of accepting weak
authentication and fundamentally spurious results.
However I do acknowledge some counter-examples to the
above. According to Prac Spec, Mr e has
made money using head and shoulders patterns.
Dr. Brett
John, as I have told you before I believe you are one of the
pioneers in our field with your development of the Bollinger Bands. As far as I know your
Omid
Malekan:
We spend a lot of time on this list arguing the same points
with different definitions, thinking that we are disagreeing. Perhaps if we
introduced some
I am a countist, and my personal experience with many
chartists, particularly the kind that is often criticized on this list, is that
they are mostly discretionary traders who depend on chart patterns as a tool of
communication. I.e., something tells them that its time to be bearish, so they
find a pattern that explains the bearishness. It is very possible for a
countist to look for patterns that only verify his outlook, but a proper
application of the scientific method usually eliminates that pitfall.
I am fairly convinced that given enough talent, experience,
time and capital, anyone can make money using any strategy, so far be it from
me to denounce any traders approach. At the end of the day and in the long run,
only the scoreboard can decide if a tennis player has the proper swing, and
only the P/L can declare someone a good or bad trader. However, there is a big
difference between using a strategy to make money, and teaching or selling it
to others. I may not have a problem with a religious zealot per se, but if he
comes to my door trying to convert me then I will probe, counter and question.
I have to take issue with this idea that each approach is
based on a different side of the brain or someone's artistic vs analytical
skills. Pretty much anyone with either side of the brain could learn how to
measure a mean or standard deviation, but that's the easy part. We already
discussed how asking the proper question is far more important then finding the
answer, and to that I will add picking the right answer is far more important
than simply generating one. I am a countist, but for tomorrow I
Dave
g:
it was not my intention to offend anyone with my post. I appreciate and utilize the contribution
jab has made to understanding price movement and in
Non-parametric kernel regression is simply a fancy name/
framework for producing a moving average.
Instead of looking at the last 10 and weighing them equally or saying
I'll make the last one worth 10% and the previous one worth 90%---this provides
the freedom to specify any number of positions and a specific weighting to
each. One could, as an example, weight the fourth one back 22.2%, the fifth one
19.02 %, etc. The weightings chosen are
called the kernel. Andrew Lo decided to
use a bell curve as the kernel. One has to be aware of a deception inherent in
Lo's study---he is weighting the points after the current point as much as the
point before--so the study
Will:
I think
the most relevant question to ask about trading systems is not whether or not
they work, but why they work. This sort
of knowledge allows
The unscientific application of counting techniques will
produce a lot of data with little predictive value, particularly if one does
not understand what to count and why they should count it. Concerning chartists, big money can make
charts lie if there is an advantage in doing so -- not in highly liquid stocks,
but smallcap investors should beware.
Zussman:
Burns’ study showing that TA experts could not
reliably distinguish btw real stock charts and sham random charts does not
disprove that these individuals
(cf the recent white paper from Pat Burns, a smart guy with
AFAIK no ax to grind:
The Technical Analysis Challenge
Abstract: We report on a study of the ability of analysts to
distinguish an actual price series of an equity from random alternatives. Virtually all of the statistical tests on
the results support the hypothesis that no skill was exhibited in selecting the
correct response. Many of the analysts
were extremely over-confident about their ability to select correct answers.
http://www.burns-stat.com/pages/techanchal.html
D.F.:
First, what's the point of Burns' study, and does the fact that he was
able to generate random patterns (and also able to coax naive chartists into
trying to distinguish between the indistinguishable) make him a "smart
guy", or is there some other criterion used in labeling him as such?
Second, could not the same type of study be done with
counting, or would that somehow be different?
Subject: the scientific revolution
Might I suggest that the discussion we have had here in last
several days about technical analysis versus counting is the best discussion of
this that has ever
been held and also the same for the hunters versus the
prey in the market. Both were
engendered by noting a series of defects and anomalies in the usually accepted
ideas on this subject. Then a strident new idea on the subject was introduced
by a few of us. It was reformulated, attacked vigorously, went thru the "
It's ridiculous, it explains a lot, it deserves to be considered, let's fill it
in with facts, it's completely tautological, everyone knew it anyway” process.
This, apparently, according to Kuhn, is the way many new
scientific ideas of substance seem to evolve (albeit that should be tested).
I have always felt my ideas about deception being the key
mechanism in markets, and this enabling the strong to prosper and the market to
continue illumined in
EdSpec was
somewhat unappreciated and unexplored. In fact the reviews of
EdSpec
are
almost as uniformly negative as those not written by my friends for
PracSpec..
I wonder if the Kuhnian paradigm can be used to gain insight
into trends in industry or stock rotation. Vic (12/6/3)
Dr.
Brett Steenbarger:
This is a very interesting idea. The heart of the Kuhnian account of scientific progress is that
the rules and understandings guiding normal inquiry (the "paradigm")
are relatively resistant to change until a sufficient body of anomalies
accumulate. At that point, there is a
qualitative shift to new rules and understandings (the "revolution").
If Kuhn's account holds for markets, investors should be
relatively slow to identify anomalies in their assumptions of bullishness or
bearishness for the broad market, particular industries, or individual
stocks. This would create exploitable
inefficiencies in the market. It would
also fit with some of the early behavioral finance research on under-reactions
to initial earnings surprises.
A few years ago, a chance observation by my daughter Devon led
me to look at individual stocks that were making an X period low during a
period of generally rising prices for the SP (and vice versa). When there was only one such stock, it was
often company-specific news only creating the lagging performance. But when
several stocks in my basket of 40 were making new X period highs/lows during
times of falling/rising market prices, this seemed to be a true anomaly--and it
wasn't too long that the rest of the market school started to follow these
"lead fish" in a trend "revolution".
Friday AM, on the heels of a seeming breakout around 11:30
ET, we saw such an anomaly. The SOX and
small cap stocks (SP 600, Russell)--frequent lead fish--should not be hitting
new multiperiod lows in a viable uptrend.
Those seeming breakouts--the bane of system/trend traders
everywhere--are beautiful examples of market deceptions that part visual
traders from their capital.
George
R. Zachar:
So others won't have to duplicate my google:
http://en.wikipedia.org/wiki/Paradigm_shift
Kuhnian Paradigm Shifts
A epistemological paradigm shift was called a scientific
revolution by epistemologist Thomas Kuhn.
A scientific revolution occurs, according to Thomas Kuhn,
when scientists encounter anomalies which cannot be explained by the universally
accepted
paradigm within which scientific progress had thereto been
made. Once new discoveries are made that cannot be reconciled with a current
paradigm,
and these results are repeatedly independently confirmed by
other scientists, then the scientific community is forced to create a new
paradigm in line with the evidence. This is a key difference between religion
and science (and generally to science and other belief systems); adherents of
the scientific method are generally willing to change their beliefs when new
facts and compelling logic are presented.
Classic examples of paradigm shifts include:
Galilean relativity.
idea of the ether and of absolute motion.
previous ideas of space and time.
A common misinterpretation of Kuhnian paradigms is the
belief that the discovery of paradigm shifts and the dynamic nature of science
is a case
for relativism, i.e., "All kinds of belief systems are
equal", so that say, magic, religious concepts or pseudoscience would be
of equal value to true
science. Kuhn vehemently denies this interpretation and
states that when a scientific paradigm is replaced by a new one, albeit through
a complex
social process, the new one is always better, not just
different.
These claims of relativism are, however, tied to another
claim that Kuhn does at least sometimes endorse. This is the claim that the
language and
theories of different paradigms cannot be translated into
one another or rationally evaluated against one another; he says they are in
these cases
incommensurable. This idea gave rise to a lot of talk of
different peoples and cultures having radically different worldviews or conceptual
schemes--so different, anyway, that whether or not one was
better they could not be understood by one another. However, the philosopher
Donald
Davidson published a highly-regarded essay in 1974, "On
the Very Idea of a Conceptual Scheme", arguing that the very notion that
any languages or
theories could be incommensurable with one another was
itself incoherent. If this is correct, Kuhn's claims must be taken in a weaker
sense than
they often are. (12/6/3)
Chess Grandmaster Nigel Davies (www.tigerchess.com):
There is a way that technical analysts could be tested. Sit
them down, showing them one bar at a time with different sorts of data; some of
it genuine (but
Over enough samples they should produce superior performance
with the 'real' data if their claims are valid. (12/6/3)
Fitz:
You are all railing at windmills with this pointless exercise. The goal
of trading is to make money. Period. It should not matter whether the data is
real
>...my contention is that even the most competent counter
could not determine, using the most sophisticated mathematics and quantitative
methods, between a
By inspection?
Dr. Theodore P. Hill asks his mathematics students at the
Georgia Institute of Technology to go home and either flip a coin 200 times and
record the
results, or merely pretend to flip a coin and fake 200
results...
http://256.com/gray/info/benfords.html
Nigel
Davies (www.tigerchess.com)
I think this EXPERIMENT would be quite revealing. And
countists could be tested similarly, given 3-5 years of 'data' and then asked
to trade the same 'market'
Of course we don't know if the exercise is pointless unless
it is actually conducted. The goal of trading may be to make money, but this
can't be accomplished
just by stating it. A METHOD is required. (12/6/3)
Fitz:
Here is my last word on the subject, and then I will go back
to the sidelines where I belong. The experiment would not be revealing at all
because of the
imvho, much on this list is self-congratulatory heat, not
light. I gratefully read because every once in a great great while I learn
something of use. But it
J:
For what
it's worth I don't see a major differences in the kinds of data that the
typical technical analyst uses versus a counter. The difference as far as I can
tell is primarily in methodology. The
counter seeks to understand the prospective expectation around a proposed price
pattern/event by testing it historically, by hand, or with a computer. However,
the patterns for both are often some formulation of price action.
We have worked with many chartists/TAs, trying to help them
quantify their patterns into systematic rules.
Our most typical finding is that the
To me this is a big disadvantage for most Chartists, and
lends itself to cherry picking the historical results that fit the desired
outcome. The quant or counter knows that only 1 out of 100 (perhaps much less)
ideas will pan out historically, not to mention prospectively. While it's tough to know if a pattern will
work, it's helpful to know if it ever worked. If you can get the Chartist to define the method(s) you can test
and at least know the latter.
Big
Al:
I think you're touching on a very important point: To what extent do traders of any persuasion
actually do what they say/think they are doing? Versus: To what extent do we react emotionally to specific
situations and then rationalize it later into a strategy?
Nigel
Davies (www.tigerchess.com)
I must admit that I would also have great difficulty in
giving you the 'rules' for why I make a particular chess move, so inevitably I
have some inbuilt
So I don't understand why there are all these discussions
about which methodology is valid; if people claim they can make money from the
markets using a particular methodology, then let them show it by either making
their trading statements/results public, publishing their predictions or
entering into an experiment such as the one I have described. If they do this
they have my greatest respect, if not I figure they're just wannabe gurus who
want some cash to explain what they can't do.
Does anybody have right to exist in the investment business
if they perform worse than a tracker fund? I wonder what people would think
paying for my chess lessons if I couldn't beat a random move generator!? (12/6/3)
Adi
Schnytzer:
Nigel's idea seems right to me. Give a group of chartists
and TA folk a hundred series of prices and volumes and get them to predict.
Until it's been done, we simply won't know which works. (12/6/3)
Charles
Pennington:
Subject: Take the Challenge
Technicians, chartists, countists, Ba'al-worshippers, step
up to the plate.
Attached is:
1) a plot, and
2) an text file
each containing 3 years of daily closing data for both a
real MYSTERYSTOCK, whose identity will remain a secret (though it is heavily
traded and on the NYSE),
and for SIMUSTOCK.
For SIMUSTOCK I generated the daily changes in a purely random way, so
that there should be no predictability, but beyond that I won't
tell you. I did not
attempt to make SIMULASTOCK look similar to MYSTERYSTOCK except that I tried to
make the magnitude of their typical daily fluctuations the
same.
I haven't labeled which one is MYSTERYSTOCK and which one is
SIMUSTOCK. The data sets are just
labeled as "1" and "2".
Those bold enough to take the challenge
have to decide which is which.
GOOD LUCK!
Subject: Take the Challenge
OK, I'll put my foot in it. Data set 2 "looks"
more like a real stock to me. Obviously I'm coming from the chart approach.
Phil
McDonnell: Countist that I am I used the head and shoulders approach
to get my answer.
Subject: Take the Challenge
Whatever is beyond the chart on the right-hand side is in
the dark. It cannot be seen. However, other things are going on in the
investing/trading
environment that will impinge on the last data point.
It is not just that last data point that is definitive for
our solution. Take an example. If this
is a gold chart what is the dollar doing, what is
inflation doing, what is current interest rate environment,
what is public sentiment, what is Fed policy, what wars are going on, how is
oil doing, are
beans in a crisis, are there droughts across the land, do we
have high employment, low employment, is the investment community under attack
from
government legal sources, are there scandals, is Soros in
heavy on this item, who is the ax. And
so on. Then a better prediction could
be made -
not by me, but by some of the sophisticated, talented,
bright participants on this list.
Subject: Take the Challenge--We have a winner
The winner of the challenge is Alix Martin, who I think
would be considered a countist. I won't
reveal yet how he did it, because I want to give those
remaining a chance to come in second and third. But I'll reveal that apparently it was easy
for Alix, for him somewhat like choosing between the real da Vinci
Mona Lisa and the Warhol silkscreen version. I learned a lot from his explanation of how
he did it, which I'll eventually post (unless he'd prefer
otherwise).
The first column of numbers is random in some
BAALWORSHIPPING sense. The second is not quite. At least that's what I figure,
making 2 the MYSTERYSTOCK. If I am right, I will be happy to explain how I did
it. If wrong, I will quietly hide in the corner!
Subject: Take the Challenge--Final Challenge Results
Data set 2 was MYSTERYSTOCK, and it was Ford. Data set 1 was SIMUSTOCK.
Congratulations to Brian Haviland, Alix, Adi, and Alex
Castaldo! Of these 4 Brian is the lone
self-proclaimed chartist.
One respondent gave an incorrect answer, but he acknowledged
that he was guessing tentatively. He
thought that DataSet1 was more aesthetically pleasing
and suspected that it was the real stock. Lesson: stocks are ugly.
Alix looked at, among other things, the behavior of the two
"stocks" immediately following big down days, declines of more than
5%. Ford/MYSTERYSTOCK/DataSet2 had, on
average, a nice recovery following 5% drops, but SIMUSTOCK/DataSet1 followed
with random behavior.
There were no responses from aeiethieeiestic Baal
worshippers or NPR pledge-week contributors.
I'm planning another Challenge for the near future.
Alix
Martin: I also looked at the histogram of price changes for both
datasets. The first one was very symmetrical, apparently doctored to match the
stock's
average return and standard deviation. The stock's data
was more natural looking, with fatter tails, asymmetry, and a
rough looking distribution. The excel file with the histogram is
available on
http://alix.martin.free.fr/speclist/Challenge.xls
Thanks Charles for an interesting experiment.
Subject: Take the Challenge--Final Challenge Result
I have just realized that the Ford residuals' behavior
demonstrates that there is heteroskedasticity in there; i.e., that the
volatility is not
constant. I guess Engel deserved his prize!!!
Yale
Hirsch: John Wiley publishes the work of Thomas N. Bulkowski and in
his "Trading Classic Chart Patterns" and his "Encyclopedia of
Chart Patterns" he provides a
thorough and statistical assessment of all the various
chart formations. I had never seen this done before. A review appeared
in the 2003 Almanac. Also, are we technical analysts when we are on
Yahoo and enter a stock symbol and then look at price action on its
chart for today, the last three months or five years? Technical analysis
is used by all of us to some extent or other. When the chair predicts we
will likely see 10000 on the Dow before we see 420 on gold, isn't that
sort of a technical prediction?
Nigel
Davies (www.tigerchess.com):
It may be impolite of me, but I'd ask Mr Bulkowski if his
results could be independently verified and then I'd want to see his trading
statements...
There are one or two guys who claim to be good chess
teachers despite having low/non-existent ratings. They talk 'authoritatively',
but then they don't seem to be able to apply their brilliant teaching methods
to their own game! Usually they figure out some excuse such as 'retirement'. I
even know one chess teacher who claims that he doesn't play because he's
worried about becoming paranoid!
I saw an interview with Van Tharp once in which he made out
that he was a great coach but it just so happened that he didn't trade. And my
respect for him went rapidly to zero...
(12/6/3)
Hany
Saad: Let me suggest a fun
nonscientific way of testing technical analysis` premises from someone who took
and passed the three levels of the chartered market
technician exam before deciding for good that the backbone logic on which all TA premises is based, namely "the trend is your friend" is inherently flawed...
The idea occurred to me one day crossing the street as the
signal was flashing "Don’t Walk" while I was daydreaming (a too
frequent occurrence) by mistake in
Paris in one busy street. To my amazement all pedestrians
followed blindly without checking the light either with cars honking behind me
in a chaotic scene.
I was so amazed by this coincidental finding that I spent
the rest of the day playing this game (crossing on a Don’t Walk signal),
getting the exact same result.
The moral of the story is that it’s human nature to follow the
leader. People feel more safe in company.
What the hell does this have to do with TA???
Throughout history, wealth distribution remained in the area
of 20 to 80 with 20% controlling 80% of the wealth and vice versa. While the
mkt isn’t a zero sum game as you undoubtedly all know due to the vig, the
aggregate wealth sure is. Now, all the people in the 80% (the majority) league
must be part of a herd and must be taking the other side of those in the 20%
league by doing something COMMON like FOLLOWING THE LEADER-- "me" in
the case of the “Don’t Walk” signal, or buying while others are on up days and
selling while others are on down days a la Mr Livermore in the case of markets.
This is clearly a case where the herd loses to pay the vig, the friction costs
for market upkeeps and the 20% handsomely.
What i am trying to say here is that people by NATURE are
trend followers which puts them in the majority league which puts them by
definition on the losing
side.
Consider this: Contrary to popular belief, cougars are the
most successful predators of all...not lions, not wolves. While cougars differ
widely in their hunting techniques, they do not lurk in the treetops waiting to
ambush passersby....they hunt on the ground and ambush their prey from behind...they
are generally nocturnal and solitary hunters...the success of the hunt depends
solely on the element of surprise... they are classified as a stalking predator
rather than a pursuit predator like the wolf...a fatal bite below the base of
the skull, resulting in a broken neck, is their preferred method of killing
prey...it very likely for a person to live in the same area as a cougar without
every seeing one...while scientists attribute the cougars’ success to physical
abilities and techniques, I suspect that SOLITUDE in the hunt is the main
reason for a cougar’s success. A lion on its own can hunt more successfully
than a herd of lions. This is another example where being a member of a herd (a
trend follower) can put you automatically on the losing side.
One way to make a devoted technical analyst lose faith in TA
is to play his own game (which I did with floor traders). TA books are loaded
with specific examples proving certain specific patterns like head and
shoulders, ascending triangles, blah blah, ignoring the rest of the sample
where the same specific patterns clearly didn’t work. It’s easy to generate
countless charts with specific past dates where the trend was clearly in a
certain direction before reversing abruptly: Make sure the trendline isn’t too
far away from the 50- or 200-day moving average (a major objection by technical
analysts that it’s a sign of reversal). Stop the time frame on the chart right
before the reversal and ask a technician to give u a prediction on the chart.
Then show him the chart in real time. While this experiment is in no way
scientific, it’s only fair to play an unfair game the unfair way.
Philip
J. McDonnell: Sorry, I couldn't let this one go...
Hany wrote:
>While the mkt isn`t a zero sum game as u undoubtedly all
know due to the vig, the
aggregate wealth sure is<
The market and wealth are variable sum games. As the market rises wealth is created out of
thin air. Wealth evaporates when
markets fall.
Microsoft has around 11 billion shares outstanding. When MSFT rises by 1 point $11 billion in
wealth is created.
Otherwise Hany great post, I especially liked your street
crossing discovery - very enlightening.
Debashis
Basu:
Might it be useful to consider that our entire focus should not
be on discussing systems or methods that work, but on what do we do in a live
situation when we find that a pre-tested pattern (TA or counting) is failing.
WHEN do we determine that failure and WHAT do we do about it? Take the small
loss or double up on the other side?
There are 1,000 times more advice in the world on WHAT to
buy than on WHEN to sell. Similarly, there is more discussion on what TA
patterns and counting work
than on what to do in case they don't.
Out of this, is there a third way possible, one that focuses
more on the point of failure of a method rather than its success? In other
words, if we know when
the system failing and take the loss, so as not to be part
of the 80%, will we enter the select 20% riding on the shoulders of sheer randomness
-- NO MATTER
WHAT the method was?
David
Higgs: Do you call it a "cycle change" when tested
pattern no good anymore? Who got wise when your share of chip allotment
exceeded..."X"... as there always seems a tendency for to keep investors in the 80%
group...........Yes, where is......... "here's how ya do
it"................... isn't is a simple thing?
And for those that walk when sign say don't walk, well they
the ones that probably belong in the 80% group. (12/6/3)
So we don’t get lost in forest for the trees, let us recall
that many technical analysts are charlatans who don’t have a clue as to how to
trade. They encourage
others to overtrade and sell worthless systems, thereby creating
much misery and dead weight cost. There is such a large random component in
stocks that is
extraordinarily difficult to tease out a method of making
money from it. Certainly the nonsystematics of the vast majority of technical
analysts not sufficient to
beat buy and hold. After a period like this year when
throwing darts at any low-priced stock with insider buying made 75% or so, many
tend to forget that just
by being long, one has to make money over most periods. Then
the technicians who say make 10% in a year like this on average, give back
hundreds of billions,
Ross
Miller: Technical analysts are the Ramones of the financial
markets. (12/6/3)
Robert
Mahan: Subject: Punk finance.
I take issue with your insult of the Ramones. The Ramones
may not have been great musicians, but they certainly weren't phonies and
neither deceived themselves nor others as to their abilities. I'd say technical
analysts are the Yanni of the financial markets.
In fairness to all wouldn't they be the MILLI VANILLI ???
James
Altucher: In terms of TA versus counting I really wonder if the
argument is simply entertainment. I don't really believe either side. Do
countists take money from
In the spectrum of fact on one side and fiction on the other
we only really know two things on the 100% fact side:
While we know that a financial time series in general
subscribes to the properties of a lognormal distribution and hence statistical
testing can be of some use, we also know that at unfortunate moments this
relationship breaks down and goes from poisson to poison for its practitioners.
For me, the benefits of counting is in the construction of a plan. When
everything goes according to plan and the counting works like magic (since
ultimately science = magic in the eyes of a child) then all is good. When
things don't go according to plan, then one can get the hell out - the world
has become different from the past. Having a plan, to me, being ultimately more
important than the method of construction. That counting is my method I feel
strongly about, but religion is a personal choice.
We can see from the IASG site that the real key to success
is not the method but the staying power. How many of the funds tracked on that
site suffered disastrous losses even in their early years, only to stick
around, persevere, and five years later end up managing a billion or more. (Roy
Niederhoffer’s 33% loss in 1999 comes to mind. I'll be honest and say I don't
know if I have the psychology or the temperament to have been able to handle
that and yet here he is, today, managing 10 times as much as he was managing
then. To me that is admirable, not necessarily the method used to get this
year's 5% return).
But, as my daughter Josie told me this morning,
"4-year-olds know more than daddies." And I have to assume she is
right.
reversal and a macroscopic tendency to reversal. In some of
our recent work, we have found that markets can be modeled as victims of an
epidemic with
susceptibles, and immunes. After a certain period, markets
become immune.
Good tests for all these can start with Leo Goodman’s work
on movements and comovements between dependent time series. (12/7/3)
Yale
Hirsch:
I don't see how there can be a legitimate conversation on
technical analysis without a definition of terms, which are then agreed upon by
both combatants.
Also to be considered is that there are those that may have
a proclivity to one discipline over the other. Other considerations are the
natural intelligence one
brings to the table and the willingness to work hard and
spend years developing the tools of whichever trade is chosen.
There's no real reason for one group to be intolerant of the
other. There are always charlatans among every group. What were the recent
"market timers" in
mutual funds who have disgraced us all? Were they TAs or
countists, or just thieves? I would have thought I would be seeing a fantastic
quote by Henrik
Ibsen written a hundred years ago repeated time and again in
this climate:
"Those heroes of finance are like beads on a string.
When one slips off, the rest follow."
Does anybody have a definition of the two disciplines that
we can all agree on? If not, we should move on. (12/7/3)
A Fundamental Analyst looks at company numbers and market
information and the economy and global politics and whatever else, and then
condenses all that
information down into an order to buy or sell some order
size of stocks, options, futures, bonds.
When the Fundamentalist places his order, his decision
is stripped of all the information that went into it. The market receives only price and order
size.
Efficient Market Theory says that all the analysis done by
all the buyers and sellers of a given security (or commodity, etc) is reflected
in the data points comprising the current spread for that security. Weak form EMT says that no additional
advantage can be gained from studying historical price/volume data. Now, however, even Sharpe admits there is
evidence of price momentum at certain times.
Both Technical Analysis and Speculation
("counting") work on the implied assumption that, traders having sent
signals to the market with buy/sell orders, the market sends signals back to the
trader via price and volume data, and that this price and volume data can be
analyzed to provide new information predictive of future price moves.
Speculation allows that most of the signals sent out by the
market are simply noise. However, one
may be able to spot "tells" or temporary patterns that indicate some
underlying dynamic. Perhaps some
particular group of securities always loses about 3% the week before each Fed
meeting, and then regains the 3% the week after the meeting, no matter what the
outcome of the meeting. This indicates
an underlying reality, that some big players are lightening up before the Fed
and then reloading afterwards.
Speculators know that as a pattern becomes widely known, it should be
traded away (ever-changing cycles). Also, the underlying dynamic may simply change.
Technical Analysis seems to believe that the signals sent
back by the market are much richer in information and also that many
interpretations of those signals are true most or all of the time (i.e., not traded
away). So, resistance/support,
head-and-shoulders, breakouts, etc., are valid from stock to stock and from
situation to situation. However, most
TAs argue that their analysis is simply a tool and takes experience to
interpret correctly. Whereas the
patterns that Specs look for should be traded away as they become known, TA
should work in the opposite fashion, i.e., TA should have at least some
self-reinforcing qualities. For
example, the more people believe in
support/resistance, the more of a support/resistance
"effect" one is likely to see.
Adi Schnytzer: And, of course, if you believe in the efficient market hypothesis then Charles's puzzle is silly because there should be no difference between real share prices and numbers going on a random walk with or without drift! The reason I believe markets are inefficient and that the inefficiencies are NOT bid away quickly are several:
1.
Not all traders are rational and different biases give rise
to different free lunches among those who aren't beset by them.
2.
Not every one have equal ABILITY when it comes to data
analysis, be it econometrics, counting, TA, studying the books, whatever. Thus,
we will
all come to different conclusions about the identical data
sets. Why are there bulls and bears out there?
3. We don't have identical data sets
anyway.
4. And even
for the same degree of 1, 2 and 3, we won't have the same amounts of money.
There's no point telling me that I can gain a net
pure arbitrage of 0.001% on a certain deal if I only have $1000
to invest. I'd be wasting my time.
5. Inside
information matters and ultimately destroys the efficient market hypothesis.
Victor
Niederhoffer:
The Don de Almantaria proposes that we define technical
analysis before positing that it is mystical and pseudo science or a valid
technique that simply cant be quantified. I google under “technical analysis
definition” and came up with the first 7 definition based on the google
analysis. I give each of the top 7 below. The first one captures the essence of all of them: TA is "a method of
evaluating securities by relying on the assumption that market data, such as
charts of price volume and open interest can help predict future market trends.
Tools used by technical analysts are advance/decline line, head and shoulders,
moving averages, point and figure charts, resistance, support analysis
ascending bottoms, descending tops, Eliot wave theory, momentum indicators,
MACD, on-balance volume, trendline, vertical line charting, arms index, double
top
Alternate definitions:
1.
"Technical analysts are market timers who believe that
trends once established have a tendency to continue until something occurs to
change the balance of supply and demand.”
2.
“Technical analysts are market observers who believe that
the 50- and 200-day moving averages form lines of resistance that are difficult
for a stock to penetrate
on the way up. Once these lines of resistance are
penetrated, however, technical analysts believe that stocks have a clear shot
at rising toward new highs."
3. Technical analysts believe
that trends once established have a tendency to continue until something occurs
to change the balance of supply and demand. Tech
analysts use charts
to uncover price and volume patterns. Indeed. "TECHNICAL ANALYSTS RELY SO
HEAVILY ON CHARTS THAT CHARTIST AND TECHNICIAN ARE OFTEN USED
SYNOMYMOUSLY.”
3.
The Society of Technical Analysts gives diploma and defines
TA as” "TA's believe that the
balance of supply and demand is the
root of price setting. A thing is worth what it can be sold for. The averages
discount everything.... their taxonomy is bar charts, gaps, islands, moving
averages, candle charts, Dow theory , chart patterns, e.g., triangles, flags,
pennants....” Also, from the Society’s trading glossary: “Technical analysis a
pattern that results where a stock price reaches a peak and declines, rises
above its former peak and again declines, and rises a third time but not to the
second peak, and then again declines…. TAs generally consider a head and
shoulders formation to be a very bearish indication."
Based on this, charting and trends continuing would seem to
be common to all of them as well as the techniques such as Eliot waves listed
above. And yes, I claim that all this is mumbo jumbo, that none can possibly
come up with valid predictions above chance from such desiderata, and that the
only hope for anything useful from the above is to state a hypothesis, quantify
something, starting with the standard statistical tools such as serial
correlation, and runs, and then to test the hypothesis, ... then to take
account of how to revise it, then to try to understand the factors that would
prolong its ever so miniscule departure from randomness. Vic (12/7/3)
D.B.
I'll add my $.02 since I know TA
and a little bit of math/stat. (BA Econ., BS Math).
I'll respectfully disagree that all of TA is mumbo jumbo.
I'll agree that some (maybe most) of the generally accepted ideas of TA (see
books by Pring, Murphy
However, in my opinion, TAs do have a lot of tools (mostly
price derived indicators) that can and should be analyzed scientifically. I
think it would be
FYI: I begin work on my MS in financial math next semester.
I am making the transition from a TA to a QA.
Mr.
E: I have watched this discussion for a while and allowed all sides to
express themselves like the other major capital market bonds, equities also has
it own term structure...and in a few minutes I’ll describe this concept which
explains all the confusion between those who want to deride TA and those who
don’t.
By way of history, I have watched the changes in money
management styles for over 35 years, from the perspective of several major
league buy sides, bulge
bracket sell sides, and of course many years at a world
class performing hedge fund.
I began my career in the employee benefit trust at Bankers
Trust in the late 60s, which parenthetically was then the 2nd largest pension
fund manager in the
world, and I can report that senior portfolio managers with high
irregular pension contributions from the entire ATT, IBM, Standard Oil of
California/New
Jersey/Ohio and other big funds USED TA for timing
assistance of what new oe existing positions to add or initiate… And further
THEY INVITED Shaw from Smith
Barney and ONEIL who subsequently founded INVESTORS DAILY to
give regular presentations to the entire pension and personal trust portfolios
and analysts.
The point was that, no matter what data may have been
available academically to undermine the credibility of TA, the PRACTICE of the
most sophisticated VERY
LARGE MANAGERS was to be guided by the charts and its
various forms ...
The term structure of bonds is nothing more than the CAUSE
of the SHAPE of the YIELD CURVE, now and in the future...the forecasting technique
to determine the
term structure is simple and yet highly complex...at every
point on the curve, say the two-year area, and the five-year area, WHATEVER
dominant part of the
customer base, say property and casualty companies that have
positive cash flow at that time, well this cash NEEDS TO BE INVESTED, and if it
is more than the
available supply of new or second hand bonds, THEN THE YIELD
CURVE will INVERT AT THAT POINT UNTIL
THE INDUSTRY IS OUT OF CASH...alternatively, if the US Ttreasury or Fannie Mae
were to finance a huge deficit at an average life of say 10 years, well the
curve might BLIP UP if there wasn’t the cash flow from real or
sovereign money investors...
Fast forward to stocks, if the majority of active managers
read charts and listen to TA analysts, and if they use those approaches to time
stock purchases (and it is mostly buying by the way not selling) then over time
the market because of the utilization will develop patterns that irregularly
can be mined for profit.
But just like the bond market, when the segments of the
stock market who create these patterns RUN OUT OF CASH, so to will the BLIPS in
the term
structure of bonds and stocks quickly disappear...
We have 22 years of running models that have produced excess
returns across the macro world of stock bonds currency and commodities which
means that TA is used by almost all classes, except INDEX PLAYERS, who will
either extend or detend a TA anomaly...
To the extent that academics try to PROVE that TA isn’t
important is WONDERFUL, because in the real world TA is almost required
knowledge among the BIG and
The reason that TA appears to be random has more to do with
the TERM STRUCTURE changes.
Steve
Wisdom:
I'm hesitant to weigh in after hearing from the likes of
Chair, E, & Charles, but I have a bit of personal experience.
Not too long ago I worked at a dignified Wall St firm under a
trader who read charts.
He'd often make statements of the form 'XYZ has a good story.. but the chart is horrible!', or 'ABC has
a great chart.. I'm getting a woody!'
And his 'read' on the chart would be shared by nearly
everyone he spoke with, buy- and sell- side.
Indeed I was embarrassed to be unable to
The unanswered questions are (1) whether chart-reading is
really just shorthand for an effect such as 3- 6 month momentum that's been
written of
ps/ As the years go by on Wall St the colorful old-time
chart-readers are gradually giving way to faceless quants.. After they're all
gone, I think Chair will miss them.
M.K.:
Of course Technical Analysis doesn't work in any systematic
way. But then fundamental analysis doesn't work that way either. If it did you
wouldn't be able to replace 70% of fund
managers with chimpanzees without anybody noticing any difference (except
possibly their spouses).
To wax epigrammatic, TA may be said to try to predict the events of tomorrow by
studying what people are doing today while FA tries to predict what people will
do tomorrow based on the events of today. I have no axe to grind either way
though I suspect many are drawn to TA as a way of providing false certainty, as
a kind of pseudo religion. In evaluating its techniques, as with any strategy,
one has to be aware of asymmetric payoff - it's the expectation, not the
probabilities that are important. A method might work nine times out of ten but
the tenth time gives it all back and more.