VOLUME INDICATORS
Editor,
It was with considerable interest that I recently read the interview
with David Vomund in the October 1999 S&C. He brought to mind what
I felt was a balanced view of market efficiency. He also whetted my appetite
for volume indicators. I already use Marc Chaikin's money flow indicator
in the Australian market, without a great deal of success, and I am interested
in the volume accumulation percent indicator. However, as Vomund does not
mention how this indicator is calculated, and I have no other references
to it, I wonder if you can enlighten me?
Trevor T. Bestow, via e-mail
CANDLE PATTERNS
Editor,
In the January 2001 Letters to S&C, reader Dejan Milenkovic mentions
that upmoves end with a hanging man candle pattern and downmoves with a
hammer pattern. He is correct. This is a serious shortcoming often overlooked
in Japanese candle pattern analysis: You cannot have a bearish reversal
candle pattern in a downtrend, only in an uptrend. Similarly, you cannot
have a bullish reversal pattern in an uptrend. Reversal patterns call a
reverse of the very trend that helped identify them in the first place.
By the way, John, I didn't think anyone could replace Thom Hartle
as editor of S&C, but you are doing a superb job!
Greg Morris,
via e-mail
MurphyMorris, Inc.
Obviously, you are correct on all points. -- Editor
CANDLESTICKS REVISITED
Editor,
The article "Coding Candlesticks" in your November 1999
issue contains an interesting approach to quantifying candlesticks. But
after a second reading, my initial enthusiasm cooled off by quite a number
of degrees. It seems to me that the author has made a fundamental error
insofar as he confuses coding with valuating or weighing a candlestick.
The proposed "code" is a seven-bit number in which the
main characteristics of any single candlestick can be recorded -- that
is, encoded. Although Viktor Likhovidov explains the coding scheme, it
can only be the starting point for the part that matters: the valuation.
Here, Likhovidov just tells us that "the sense" of the coding
scheme consists of assigning the highest value to the most bullish candlesticks
and gives lots of examples.
In assigning weights directly to these binary codes, he runs into
difficulties. For example, different codes are assigned to white versus
black bodies (long white = "11," long black = "00"),
although color already has a value of its own: the seventh bit.
The most serious problem, however, arises out of the respective positions
of the upper and lower shadows in the code; a long upper shadow is assigned
a value of 12, a long lower shadow gets zero. This is where I get confused.
Evidently, lower shadows are more bearish than their upper counterparts.
Are shadows generally bearish?
I disagree. An upper shadow means the bulls tried to push prices
up but didn't succeed, so it is bearish; a lower shadow means the bears
had a try but got their noses bashed, and it should be considered bullish.
This commonsense interpretation is confirmed by Steve Nison in his book
Beyond Candlesticks.
This brings up the problem of valuation. For instance, a hammer appears
in a downtrend, but a hanging man (the same candlestick) appears in an
uptrend. This is a problem that remains inherently unsolvable with a one-period
indicator. You must consider the context when valuing candlesticks.
I'm sorry to say that I'm not yet done with my criticisms. I also
have to object to the method used to compute the values for the long, middle,
and short thresholds. In the article, Bollinger Bands around a moving average
are used. These are, as Likhovidov correctly notes, computed on the basis
of the "standard normal distribution." That's certainly not what
we have here. Candlestick lengths have a theoretical upper limit of infinity
and a -- very practical -- lower limit of zero.
This means we are dealing with a log-normal distribution that is
skewed to the left. Thus, using 0.x standard deviations below and above
the average value results in uneven bites taken from the total spectrum.
At a more basic level, this raises a question about the coding scheme
per se. Here, we have a complicated method that is used to divide the spectrum
of possible values into just three categories: small, middle, and long.
The fourth value, zero, is added to account for dojis and so forth. I believe
this is too stringent. Very small values should be counted as nil, too.
So for them, a fourth category is needed. Maybe we would get better results
using 10% percentiles.
Why use an artificial subdivision of values at all? Using the actual
values of bodies and shadows, multiply the body's value with a factor x,
add the lower shadow, subtract the upper shadow, and there is our indicator.
Jan Willem E. Roberts
Munich, Germany
Jan W.E. Roberts's letter stimulated Likhovidov to return to the subject
and elaborate. See his article "Coding Candlesticks (II)" in
this issue. Roberts's letter included many other ideas that I edited out
for brevity. I hope he'll write them up in a separate article. -- Editor
FORMULAS AND SPREADSHEETS
Editor,
I just read Arthur Merrill's letter in the December 2000 issue, in
which he discussed the problem of articles that do not provide formulas
or an Excel spreadsheet. I support the recommendations he made in his letter.
I was frustrated by John Ehlers's article on the squelch indicator because
it seemed to be a valuable indicator, but I could not check it out because
of the lack of basic formulas or an Excel spreadsheet. As a longtime subscriber
to your magazine, I have often had this frustrating experience. I suggest
that basic formulas and a spreadsheet example be requirements of having
an article accepted for publication.
Paul D. Baba, via e-mail
San Carlos, CA
Okay! Okay! We'll write spreadsheets for you guys! But we've got time
constraints, you know! -- Editor
EASYLANGUAGE
Editor,
Could you publish SuperCharts EasyLanguage for the Andrews pitchfork
featured in your article in the December 2000 issue?
Kent Siegrist, via e-mail
Please contact TradeStation Technologies, Inc. (formerly Omega Research,
Inc.), for code for use in their products; we don't specialize in writing
EasyLanguage. Check also the TradeStation website at www.TradeStation.com
for already-posted code, since they publish quite a lot of custom code
there.
In S&C, we try to provide the basic math, spreadsheet, or logic
for a concept, and readers may implement it in whatever product they happen
to use. We sometimes include the code for MetaStock or TradeStation if
the article's author works in that program and can provide it.
If TradeStation Technologies cannot provide this particular code and
if the Andrews pitchfork is not a built-in charting feature of your program,
you might try looking for third-party suppliers of custom code.-- Editor
NORMALIZING THE GAPO
Editor,
After reading about the Gapo index in the January 2001 issue ("Gopalakrishnan
Range Index"), I realized that it needed to be normalized to the underlying
securities' base price to have any interpretable meaning. For example,
if security A has just closed at 10,000 and its max-min for the past five
days is 500, then the Gapo value will be log(500)/log(5) or 3.86. In contrast,
if security B just closed at 10 and its max-min for the past five days
is 0.5, the Gapo value will be log(0.5)/log(5) or -0.43.
The interpretation in the article would lead one to think that security
A is dramatically more volatile and erratic than security B, even though
they both show the identical percentage price change (range of 5% over
the past five days). Comparing markets based on their relative volatility
and consistency of price behavior is a worthy goal. Unfortunately, unless
normalized for each security's base price, the Gapo does not accomplish
this task. A more tried and true measurement would be to compute the traditional
historical volatility for each security and use that to compare each security's
relative "erratic behavior."
Joseph M. Fisher, M.D., Ph.D., via e-mail
San Carlos, CA
Sounds like a good follow-up article. See also the response to the next
letter. -- Editor
MORE ON NORMALIZATION
Editor,
Unless I've overlooked something obvious in the formulation of the
Gapo index in the January 2001 issue, a one-point spread taken over five
days in a $10 stock would have the same implication as a one-point spread
in a $200 stock. If my interpretation is correct, the index is of questionable
use. On the other hand, if one were to determine an average price for the
issue (or index or commodity) for the period of interest, such as five,
10, or 21 days, and divide that value by the high-low range (or the inverse),
one would be on the way to determining relative (normalized) volatilities
that are comparable. I have developed and used similar methods for a number
of years.
M.G. Boobar, via e-mail
We think both thoughts are worth pursuing and invite an article on either
subject -- or both. -- Editor
FIBONACCI RATIOS
Editor,
I read Rudy Teseo's interesting article, "The Gartley Setup,"
in the January 2001 S&C. This pattern seems to be the same as that
shown in Larry Pesavento's Fibonacci Ratios With Pattern Recognition.
Regarding this pattern structure (named Gartley 222), the book writes
at point 6: "If point X is exceeded the trend will continue to move
down to at least 1.27 or 1.618 of the X to A move."
How does this reconcile with what is written in Teseo's article?
Figure 1 clearly shows that the CD leg can be 1.27/1.618 of the BC leg,
and that in any case all the retracement AD must not be greater than 0.786
of XA (that is, D must not be less than point X). I think that some misunderstanding
is occurring with this pattern. Would it be possible to get some clarification
from Rudy Teseo?
Guglielmo Marco, via e-mail
Carmagnola, Italy
Rudy Teseo replies:
I have not read Larry Pesavento's book and have no idea what coverage
he has given to all the possible harmonic patterns. However, my article
only applied to the ideal bullish Gartley. What you are referring to, I
believe, is the bullish butterfly. In this pattern, the ratios are somewhat
different. The AB ratio is 0.786, the BC ratio is 0.618./0.786, and the
CD ratio is 1.27/1.618. Again, these are ideal figures. I trust this answers
your question. Thanks for your interest.
ASTROLOGY AND THE MARKETS
Editor,
It was a pleasure to read Arthur Merrill's letter in the December
2000 issue regarding his study of the relationship between full moons and
the Dow Jones Industrial Averages (DJIA). I have great regard for Merrill's
work, and I still treasure an autographed copy of his classic Filtered
Waves, Basic Theory.
First, if one really wants to know whether there is a correlation
between astrology and financial markets, then one must attempt to understand
the nature of astrology. It is not a study that lends itself well to empirical
science or statistics, which will almost immediately cause a skeptic to
assume that the method is of little value.
Merrill concurs with Brooks Rimes, who questioned the place of astrology
in technical analysis in a prior letter. He states that he tries to keep
an open mind to the subject, but to date has seen no evidence that demonstrates
the validity of astrology as a market analysis tool. As one of the pioneers
in this field, I would like to comment.
Merrill's full-moon study concurs with quantitative studies I have
conducted on solar and lunar eclipses, which are considered even more potent
to most proponents of astrology than full moons (The International Astrologer
Journal, Spring 1998, Volume XXVII, No. 2). Again, there is no evidence
of a statistical correlation to crests, troughs, or price direction in
the DJIA when plotted anywhere from zero to five days to either side of
either eclipse. The results did not disturb me, for I had observed as much
over the past 20 years. However, I was concerned because so many financial
astrologers frequently claimed that there was a correlation and implied
that it was one of the most important astrological indicators, yet they
never cited evidence or even anecdotal examples to back up these claims.
Here is where knowledge of astrology can be invaluable to one who
wants to determine whether there is any correlation to financial markets.
Most astrologers understand that lunar cycles are relatively mild in their
correlation to cycles of human activity in comparison to planetary cycles.
Yet most academic researchers try to prove or disprove astrology (especially
in financial markets) through statistical studies involving new or full
moons.
A new moon is simply a moment in time where the moon and sun appear
to be in alignment, as seen from the Earth. It occurs about every 29 days.
It is not a rare event. Likewise, a full moon occurs approximately every
29 days, when the sun and moon are on opposite sides of the Earth. Yet
planets appear together or in opposition to one another too, and far less
frequently, which would seem to be a more intriguing basis for a study
than something so transitory as the moon in its orbit around the Earth.
My own quantitative studies on these astronomical events, as published
in several books, suggest a far stronger correlation to financial markets
than either new or full moons, or solar or lunar eclipses. I would be more
than happy to send Arthur Merrill a copy of these published studies related
to gold, silver, or US stock market indices if he were interested.
The problem with doing any kind of statistical study to validate
astrology lies in the mechanics of both astrology and market activity.
For instance, specific astrological signatures rarely correlate with only
a crest or a trough in a given financial market. The theory behind astrology
is that it marks a change in collective psychology or investor sentiment.
For example, in one instance, a sun opposite Jupiter (which occurs about
once every 13 months) might correspond to a multi-month crest in US stocks,
but in the next instance, it might correlate to a multi-month trough. The
point is that certain astrological signatures (but not all) have a correlation
to market reversals, or a change in investor sentiment. How long that change
will last depends upon a multitude of other factors, both astrological
and mundane, such as the technical or cyclical condition of the market.
The matter is compounded by the fact that single astrological signatures
rarely occur. They are often present with other signatures that occur nearby
in time. For instance, the sun may oppose Jupiter every 13 months. But
in one year, Mars may conjunct Uranus two days earlier, and in another
year, Venus may conjunct Mars one day later. In another year, both of these
signatures (or others) may occur within the same week, while in yet another
year, no other significant signatures occur nearby. Merrill might find
the same thing even in his full-moon study. That is, those full-moon dates
in which stocks seem to respond sharply may have coincided with periods
in which other geocosmic signatures were occurring within a day or two.
By the way, I noticed that Merrill's study cites 60 full moons between
1992 and 1999. Since there are 13 full moons per year, this suggests about
30 instances are missing. Was this perhaps due to the fact that he did
not consider full moons that occurred over weekends? Just a question in
the spirit of research, as I don't think the results would have changed
significantly anyway, due to my belief that lunations have very little
correspondence to trend changes in US stock indices.
The mechanics of financial market prices also present challenges
to a proper statistics study. Let's say one wishes to analyze the dates
from which 4% or greater filtered waves are defined (to use one of Merrill's
technical tools, which I use to conduct studies relating astrology to financial
markets). Let's say there is a crest from which prices decline more than
4%. However, before the greater part of that decline commences, the market
trades for a week or so in congestion, forming a double top one week later.
The second crest is just a tick below the first crest. Which crest do you
use? The first one, which records the actual high tick, or the second one,
which commenced the more severe decline? In attempting any kind of statistical
research involving any market timing indicator (that is essentially what
astrology is), one has to define a variety of criteria to be used in order
to create results that are useful to actual traders or investors.
I would like to thank Arthur Merrill -- and Brooks Rimes -- for opening
up this discussion on astrology and financial markets. Unlike so many authorities
on financial markets, Merrill at least tries to keep an open mind on the
subject. That is greatly appreciated by serious market students like myself.
Most scientific people tend not to give any credence to astrology
for only one illogical, unscientific reason: it shouldn't work,
therefore it doesn't work.
Raymond Merriman, CTA, CAP
Editor, The MMA Cycles Report
West Bloomfield, MI
Rudy Teseo replies:
Thanks to all the readers who contributed to our discussion of astrology
and trading, since it was prompted by our review of Ganntrader 3.0 in the
July 2000 issue of S&C. -- Editor
WORTHLESS OPTIONS
Editor,
I would like to respond to the article "How Many Options Actually
Expire Worthless?" by Lawrence McMillan in the January 2001 issue.
I am quite surprised by the percentages mentioned. I am active in the European
market and have done significant research, and I have a different conclusion.
I have several years of data for the the Amsterdam general index
(AEX) -- the Dutch Dow, if you will. For example, the bottom chart on http://www.night-trading.com/openinterest/oi_aex.html
shows the percentage of options that will expire worthless given the value
of the AEX, if expiration were today. The number is above 60% with the
current index of around 630. Historically, it's around 70% on expiration.
This also applies to individual stocks. I'm surprised that the figures
for Chicago are so different from those for Amsterdam.
The chart itself is quite interesting. The price tends to go to the
top of the chart, where the percentage is the highest, and thus contains
some predictive value toward the price goal at expiration.
John Phoenix, via e-mail
AUTOMATIC TRADING SYSTEM
Editor,
Has anyone ever demonstrated or sold a trading system that can be
set to operate on full automatic -- that is, make real trades with real
money for the investor, without constant supervision and without the investor
having to okay each trade? Are there any systems like this currently available,
or any being discussed that you can talk about?
Larry Burford, via e-mail
Yes, some brokerage houses integrate trading systems with their brokerage
services. Two examples are Zap Futures and Lind-Waldock -- Editor
