February 2003 Letters To The Editor

or return to February 2003 Contents

The editors of S&C invite readers to submit their opinions and information on subjects relating to technical analysis and this magazine. This column is our means of communication with our readers. Is there something you would like to know more (or less) about? Tell us about it. Without a source of new ideas and subjects coming from our readers, this magazine would not exist.

Address your correspondence to: Editor, STOCKS & COMMODITIES, 4757 California Ave. SW, Seattle, WA 98116-4499, or E-mail to editor@traders.com. All letters become the property of Technical Analysis, Inc. Letter-writers must include their full name and address for verification. Letters may be edited for length or clarity. The opinions expressed in this column do not necessarily represent those of the magazine. -Editor

HOW RELIABLE ARE CANDLESTICKS?

Editor,

First, I want to compliment you on your excellent magazine.

Next, I would like to point out some faults in "How Reliable Are Candlesticks?", which appeared in the November 2002 issue of STOCKS & COMMODITIES.

In the article, author Giovanni Maiani uses definitions of candlestick signals without validating how correct these signals are. Remember, just because candle signals are in a software package, that does not necessarily make them correct. For example, to qualify a pattern as a hammer, there are four rules:

1. The color of the real body can be black or white.
2. A bullish long lower shadow that is at least twice the height of the real body.
3. It should have no, or a very short, upper shadow.
4. The market must be in a downtrend.

Rules 1 and 2 are easy to quantify. But how was rule 3 defined for the signal used in the article? (That is, what was the maximum length of the upper shadow for this to be defined as a hammer?) Or how about rule 4? How was a downtrend defined for the tests the author conducted in the article? A candle signal needs two criteria: The first is the shape of the line, the second is the trend. Thus, if there is no downtrend, there is no hammer - even if the shape of the line looks like a hammer.

The most serious error was basing the success of a candle signal on what happened the next session. Whether one trades based on a candle signal should be fully dependent on the risk/reward of the potential trade. This means that one should have a stop and a price target before placing a trade, and not just buy or sell because there is a candle signal. For instance, a stop should be placed under the low of a hammer. Unless one has a target that is a multiple of this risk, the hammer should not be used as a buy signal.

At my seminars, I have a series of pivotal rules. One of the most important is: Not every candle signal should be used to buy or sell. Always first consider the risk/reward. And this article doesn't take this vital risk/reward aspect into account. Indeed, there are times when a candle signal should be ignored. The author's study was based on just blindly buying or selling on every candle signal. Using a candle signal to trade without considering the risk/reward is like, as the Japanese proverb states, leaning a ladder against the clouds.

Steve Nison, CMT, President, Candlecharts.com

Thank you for sharing your comments and insight.

Note to readers: Steve Nison was a pioneer in bringing candlestick charting to this country and is a foremost authority on candlestick charting.-Editor

EDGAR PETERS INTERVIEW

Editor,

Thanks for the informative and timely interview of Ed Peters in your December 2002 issue. I'm a long-time fan of Peters' work and have read and reread his books, Chaos And Order In The Capital Markets and Fractal Market Analysis, many times over.

In the interview, Peters states that "...based on the relationship between stocks and interest rates, and a couple of other things, stocks are pretty severely undervalued..." I'm familiar with some approaches to determining a stock's value in comparison with the prevailing return on T-bonds; but I'm really interested in finding out about the "couple of other things" that Peters refers to in the interview. Would it be possible for Peters to give us some information on these other methods he uses to determine stock values?

Pat Lafferty, via e-mail, Pearland, TX

Thanks for your letter. We have forwarded your message to Peters with the hope that he will write an article about his methods.-Editor

TITANS OF TECHNICAL ANALYSIS

Editor,

I read with interest David Penn's article on the who's who of technical analysis ("The Titans Of Technical Analysis," October 2002 S&C). I would have included Perry Kaufman for his outstanding book, Commodity Trading Systems And Methods (Wiley, 1978); however, it is Penn's article.

Penn credits John Bollinger with the invention of the Bollinger Band method of channelized trend trading. However, the earliest description of that methodology with which I am familiar appeared in a book by Chester Keltner. Keltner was a grain analyst in Kansas City, active in the 1940s through at least the late 1960s. His book How To Make Money In Commodities described the channelized trend method in 1960. He suggested a 10-day moving average for the base and the 10-day average trading range for the channel volatility bounds.

In 1971, a customer of mine from P&S Trading brought in the Keltner channel idea. We programmed it. We named it the Keltner channel method, shortened later to the Keltner model. P&S used the model for their customers. We optimized it and managed money with it from 1972 through 1988.

The Keltner model was never a secret. We publicized it and sold the data created to a number of firms. A local Chicago firm, Essex Trading, I believe, packaged the model and sold the software. I still occasionally hear from traders who are using the Keltner model for daytrading.

John Bollinger did not invent the channelized model, that is, a moving average base bounded by a volatility measure. So far as I know, it was Chester Keltner in 1960 who described the process long before Bollinger arrived.

Don Jones, via e-mail, Cisco Futures
dljones@cisco-futures.com

Thank you for sharing your comments, especially since we know you have a long history in the markets. It is important to know history and its contributors, because we build on the past. Regarding the channel method: While Bollinger may have developed and popularized his own version of the channel method, other channel methods also exist and indeed may have been around a lot longer. Thanks for pointing this out. We published an article on using Keltner channels in our December 1999 issue, available at our website, Traders.com.

Note to readers: See Donald Jones' recent articles in S&C on auction market theory in our June, July, and November 2002 issues.-Editor

THE FISHER TRANSFORM

Editor,

In the November 2002 STOCKS & COMMODITIES, John Ehlers presents an indicator called the Fisher transform. He states that if you take the 10-day momentum of the transform and multiply it by 10, you can create a major turning point indicator, which the article demonstrates.

My problem is this: Although Ehlers shows the code to create the Fisher transform, the EasyLanguage code of normalized prices and 10 times its rate of change are not shown. I attempted to recreate what was shown in the article with the following code but it won't plot properly in TradeStation: 

Inputs:  Price((H+L)/2), Len(10);

Vars: MaxH(0), MinL(0), Fish(0), FishMom(0), FishMomAccel(0);

MaxH = Highest(Price,Len);

MinL = Lowest(Price,Len);

Value1 = 0.33*2*((Price-MinL)/(MaxH - MinL) - 0.5) + 0.67*Value1[1];

If Value1 > 0.99 then Value1 = 0.999;

If Value1 < -0.99 then Value1 = -0.999;

Fish = 0.5*Log((1+Value1)/(1-Value1)) + 0.5*Fish[1];

FishMom = Momentum(Fish,Len);

FishMomAccel = 10*FishMom;

{Plot1(Fish, "Fisher");}

Plot2(FishMomAccel, "Trigger");


If I want TradeStation to plot only the momentum of the transform by itself, it will do so, but if I want it to plot both the transform and the momentum of the transform together on the same chart, the momentum of the transform becomes a flat line. I think that this is a scaling problem. As long as you're going to print the article, it would be helpful to publish all of the EasyLanguage code and not just part of it. It appears that when one puts both the transform and the momentum of the transform on the same chart, a scaling problem is created so that both of them won't plot correctly.

Can you please ask Ehlers specifically what he did to the Fisher transform in EasyLanguage to duplicate what he showed on page 42 of the November 2002 issue? Thank you for your assistance.

Scott Gibson, via e-mail

John Ehlers replies:

Thank you for studying the Fisher transform. I meant the momentum to cross the original indicator briskly, and that is the only reason for the times 10 multiplier. Rather than use TradeStation's built-in momentum function, I meant that the one-day momentum of Fish is just its current value less its value one bar ago. Thus, you can ignore the FishMom variable in your code and replace it with the following: 

 FishMomAccel = 10*(Fish - Fish[1]);


This will enable you to plot both indicators on a graph.

The Fisher transform is important to establish nearly Gaussian probability functions so that the calculation of the one-sigma or two-sigma standard deviations have real meaning. A number of indicators use multiples of the standard deviation for their operation. The disadvantage of the Fisher transform is that the generating function must be normalized to range between +1 and -1. I used a 10-bar price channel as an example in the article. However, other indicators, such as the RSI and stochastic oscillator, also have normalized ranges. Generally, the normalization ranges between the limits of zero and 100. To use the Fisher transform on these indicators, the translation is easy. In terms of EasyLanguage code, the translation for RSI, for example, is: 

  NormRSI = 2*(RSI/100 - 0.5);


Since the original RSI ranges between zero and 100, the NormRSI will range between -1 and +1. You then can take the Fisher transform of NormRSI.

I encourage all readers to experiment with the Fisher transform to find the true standard deviation value.

PS: Some readers have had problems reproducing Figure 7 because they used the built-in TradeStation function to compute the rate of change. The following code, replacing the Trigger, yields the results of Figure 7:

 Plot2(10*(Fish - Fish[1]));

This code plots 10 times the one-day rate of change of the variable Fish.

ERRATA: HOT ZONES

Editor,

I started to read "Hot Zones" in the December 2002 S&C with some anticipation, but found Figure 3 confusing. Hope you or the author can clarify the following:

1. The open/close matrix in Figure 3 is described as "Columns representing opening zones" and rows, the closing zones. Then the text says, "Looking along the second row (since the market opened in zone 2), the market...." If the columns represent opening zones, shouldn't we be looking along column 2, not down row 2?

2. Why is there no data in row 6?

3. The Opening Zone portion of Figure 3 also has an unclear description ("The first line is the total number of key reversal up patterns"), and the number of patterns (whole numbers) are shown to two decimal places. Very confusing, but I now see what the table is trying to show.

I look forward to S&C every month, and offer these comments in the interest of maintaining the high quality of your magazine. Keep up the good work!

Bob Hansman, via e-mail

We regret some errors in the caption for Figure 3 are making it unclear. It should have read, "Columns representing zones," not "Columns representing opening zones." In addition, in the Opening Zone portion of Figure 3, it should have read, "The first column is the total number of key reversal up patterns," not "The first line is the total number of key reversal up patterns."

Regarding the data in row 6, the reason you're looking along row 2 is to determine which zone the market will close in.-Editor

THE STOCK MARKET, THE CALENDAR, AND YOU

Editor,

I enjoyed the article "The Stock Market, The Calendar, And You" (December 2002 S&C). However, I believe that the period starting October 1, 2002, and ending December 31, 2003 (that is, the Presidential election cycle), is 15 months, not 14 months, unless the number of months in a calendar year has changed recently.

Jim Orr, via e-mail

RESULTS OF A SHORT STRATEGY?

Editor,
I have a question directed toward Jay Kaeppel, author of the recent article "The Stock Market, The Calendar, And You" (December 2002 S&C). I would like to know if it would be possible to hear from him on another iteration of the "+2" strategy.

Suppose the days when the seasonal index reading was a zero, you were short the Otc with 2-to-1 leverage. A reading of 1 would indicate a cash position. A reading of 2+ would be long with 2-to-1 leverage.

I would be interested in the results a short strategy would produce, including whether the extra gains (if any) would be enough to outweigh the extra commissions.

Rick Shoup
rick_shoup@msn.com

We have forwarded your letter to Jay Kaeppel at Essex Trading. Perhaps that will entice him into writing a follow-up article!-Editor

BACK ISSUE SEARCH

Editor,

I read your magazine every month and really find it useful.

I am writing because I cannot find my issue from a few months ago. In it there was an article that mentioned a link to a website where you can type in a stock and it will tell you what stocks are statistically correlated to that stock. If it would be possible for you to e-mail me that link, I would be very grateful.

Dennis Young, via e-mail
Atlanta, GA

Readers wishing to look up past articles might try using the search engine at our website, Traders.com.-Editor

MATRIX ANALYSIS?

Editor,

I have been reading your magazine for a few months now and have been impressed to the point that I ordered a five-year subscription. Your magazine has information that is useful for beginners and the most dedicated programmers. Also, I like that you seem to have the lowdown on everybody offering services in the industry. Your magazine is a strong part of my plan to make money trading.

I have been reviewing numerous indicators, oscillators, price relationships, and so on. I have found that it is easiest for me to classify things using a numeric coding system and to place them in a matrix format. The problem is, I do not know of any past work that has been done in this area. Could you recommend any books or resources that I could look into to help me refine my ideas? Any help you can provide is greatly appreciated.

David Hodge, via e-mail

I have not heard of any work that places indicators in a matrix format. I recommend you go to the Traders.com Advantage portion of our website at Traders.com for a list of indicators. In addition, here are two more resources for a listing of indicators.-Editor

Achelis, Steven B. [1995]. Technical Analysis From A To Z, Probus Publishing.
Colby, R.W., and T.A. Meyers [1988]. The Encyclopedia Of Technical Market Indicators, Dow Jones-Irwin.

WAITING FOR THE FED

Editor,

In "Waiting For The Fed" in the August 2002 S&C, David Penn discusses the system of Mark Boucher, which bought in 10/95 and sold in 9/98. In the Letters to S&C column the following month, Penn discusses the calculation in more detail, stating that the absolute value of the year-over-year rate-of-change (ROC) of three-month Treasury bills had to be less than 6%.

Using his data source, the numbers  are:

Feb 1995  5.94
Mar 1995  5.91

and

Feb 1996  5.15
Mar 1996  4.96

This leads to an absolute ROC of:

16.5%
13.7%

which should have resulted in a sell. Has something been omitted from the discussion?

Lawrence Kirsch, kirsch@brandeis.edu
Newton, MA

David Penn replies:
It appears you are right that something was left out of the discussion - and, perhaps, out of Mark Boucher's analysis as well. Here are the buy/sell signals I get since the 11/95 long signal. The entries represent: long/sell, effective date, the year-over-year rate of change in three-month T-bill, and the actual yields of the bills.

Long 11/95
Year-over-year ROC 1.3%
5.45 to 5.52

Sell 12/95
Year-over-year ROC 8.2%
5.76 to 5.29

Long 7/96
Year-over-year ROC 5.2%
5.59 to 5.3

Sell 8/96
Year-over-year ROC 6.8%
5.57 to 5.19

Long 9/96
Year-over-year ROC 3.5%
5.43 to 5.24

Sell 11/96
Year-over-year ROC 6.3%
5.52 to 5.17

Long 12/96
Year-over-year ROC 4.7%
5.29 to 5.04

Sell 9/98
Year-over-year ROC 6.7%
5.08 to 4.74

Long 8/99
Year-over-year ROC 3.4%
5.04 to 4.87

Sell 10/99
Year-over-year ROC 23%
4.07 to 5.02

Long 1/01
Year-over-year ROC 3.8%
5.5 to 5.29

Sell 2/01
Year-over-year ROC 12.6%
5.73 to 5.01

According to the model, we would still be out of the S&P 500 as of 11/02.

Thanks for reading Working Money and STOCKS & COMMODITIES - and for spotting what I agree appears to be an oversight in Boucher's otherwise compelling timing model.

WRITING GUIDELINES

I'm a teacher at National University Lvivska Politechnika in Ukraine. Can I publish my article in your magazine?

Ihor Vysotskyy, via e-mail

Thank you for your interest in submitting an article to STOCKS & COMMODITIES. You can find our author guidelines at our website, Traders.com, in the editorial department area. The guidelines describe what we look for in an article and how to go about submitting one.-Editor

ERRATA: McSPREAD

Editor,

We just received the November 2002 S&C. Our software program, McSpread, was mentioned on page 144, but it has a wrong picture with it. Track ECN has nothing to do with McSpread.

Ron Schelling, www.2hedge.com

We regret that the pictures were displayed out of order on this page of our Trade News & Products column. The correct picture appears in this month's Trade News column.-Editor

ERRATA: TRUTH ABOUT TRENDLINES

Editor,

I have a correction to my article "The Truth About Trendlines" as it was published in the October 2002 S&C. On page 41, Figure 3, the last line should show "True" for both table entries. The words are missing.

Tom Bulkowski

We regret this omission in the table. Thanks for pointing it out.-Editor

Back to February 2003 Contents