December 1997
Letters to S&C
or return to December Contents

LETTER FROM JOHN BOLLINGER

Editor,

The conventional interpretation of Bollinger bands is not overbought at the upper band and oversold at the lower band, as the Tezels assert in "On using volatility bands" (S&C, October 1997).

Bollinger bands provide a relative definition of high and low that can be used in the interpretation of indicators to arrive at trading decisions. If a tag of the upper band is accompanied by confirmation from an appropriately specified indicator, higher -- not lower -- prices are called for. If that same tag were unconfirmed, then lower prices would be expected. The Tezels' assertion is insupportable; tags of the bands without reference to indicators are meaningless. Further, the Tezels' MAAVPB is nothing more than an 11-day, 1.7 standard deviation Bollinger band.

One last note: As the Tezels correctly observe, closes outside of the bands are continuation signals. This was covered in the earliest Bollinger band literature.

Thank you for writing. We'd love to hear more in the form of an article!?Editor

CONTINUOUS CONTRACTS

Editor,

I have just discovered your magazine. It's a good teaching tool, with many relevant articles in each issue.

The April 1997 issue referred to the "continuous T-bond contract adjusted for the rollover of each contract." I contacted some data providers, but their continuous contract data is not adjusted for the rollover. I am interested in US and international bond contracts. Do you know of any data providers that make such adjustments, or any software with appropriate algorithms?

For a discussion on the topic of continuous contract data, see the sidebar titled "Continuity" that appeared with the product review of Unfair Advantage by Technical Editor John Sweeney in the October 1997 issue of S&C.?Editor

BULL VS. BEAR MARKETS

Editor,

It is often said that an experienced trader can tell whether a chart is upside down or right-side up even if the axes are covered. The reason given is that up markets and down markets have distinct differences. However, most oscillators have equidistant overbought and oversold areas, and most trading systems are run coincidentally for long and short positions. Wouldn't it be better for system tests to be run separately for long and short positions?

Are you aware of any research or articles in your magazine that deal with this subject? Obviously, a market with a long-term upward bias such as the US stock market would have different parameters for long positions than short.

Ross Kovacs
via E-mail
When it comes to applying oscillators, you won't receive any argument from me that using a different set of overbought and oversold levels for bull markets than for bear markets is a wise approach. In fact, if you read my article last month, "Using Fibonacci ratios and momentum," you'll see my comments on this very subject.

Further, if you look at the work of Dennis Meyers, who has contributed more than a dozen articles to STOCKS & COMMODITIES on trading system design and who is the author of the software Dennis Meyers Systems and Indicators, you'll notice that all his systems use different parameters for his buy rules than for his sell rules. The rules are the same, but the parameters differ slightly, reflecting the unique dynamics of bull markets as compared with bear markets.?Editor

HOLY GRAIL SYSTEM

Editor,

In the October 1997 Letters to S&C, you went on record stating that a Holy Grail trading system cannot exist -- and then you attempted to justify the statement with false assumptions! You apparently assume that an infallible system must involve one or more of the current indicators/systems such as Elliott wave, Bollinger bands, MACD, delta phenomenon, cumulative volume or price averages, RSI divergence, and so on.
 

Further, you assume an approach where the indicator or system is designed, or tweaked, to match historical data. Well, if everyone thought like that, a Holy Grail trading system will never exist! And by the way, when one finally is developed, it certainly won't show up in your magazine or appear in anyone's mailbox as junkmail. The yet-to-be developed infallible Holy Grail system will probably use some of the mechanical techniques found in your magazine, but of course, it will use the correct independent variables!
 

After all this harsh criticism, let me add that your mag is really neat.

Frank Suler
via E-mail

Before we further our debate over whether a Holy Grail trading system can exist, let's define what a Holy Grail system would be, in theory, just to make sure we're all on the same playing field. When I hear the term "Holy Grail," I generally think of a divine, perfect, nearly unattainable ideal. Thus, to me, a Holy Grail trading system would be some kind of vatic system that calls every turn and doesn't miss a beat. In fact, you yourself use the term "infallible." On the other hand, maybe other traders think of the Holy Grail system as an approach that may call 50% of the market turns correctly over a period of several years, which would translate into a spectacular return to any money manager. It's certainly possible for a trader to develop a system with good probability for a period of time, and presenting the tools for traders to find an exceptional yet practical system is what this magazine is all about.

Further, what I imagine most traders think of when they conjure up an image of a Holy Grail trading system is a type of programmed, push-button approach to trading. In reality, trading is a lot of hard work. I believe it's possible for a trader to develop a system with an above-average success rate, but using it will involve more than spending 10 minutes a day in front of a computer. Instead, it will take a mature mindset, a commitment to spending the proper amount of time a day to use the system, the discipline to implement the signals and not second-guess them, and finally, the perseverance to adjust the system's parameters if the structure of the market should change.

Finally, in response to your statement that I have only considered indicators that are currently in use: We don't only publish classic techniques of analysis but also publish new and interesting indicators and approaches. And in regard to your comment about designing a system to match current data, if you've been reading the magazine the past year, I'm sure you've noticed the ongoing discussions in this magazine about the perils of curve-fitting and about how to best use historical databases for testing without overfitting the parameters to past data.

Thanks for your comments.?Editor

 

TRADING WITH STATISTICS

Editor,

I have a few comments on two recent articles by Jeffrey Owen Katz regarding trading system evaluation ("Evaluating trading systems with statistics" and "Using statistics with trading systems," STOCKS & COMMODITIES, July 1997 and August 1997).

Katz recommends using the trade in question as the basic element of the population. Thus, the sample will be represented by the trades executed by the system in the chosen period.

This approach has a drawback if the parameters of the system to be evaluated are set to values that produce a small number of trades, and therefore a small sample (for example, less than 30). In this case, we know our confidence in the statistics computed from the sample will be rather low.

However, if we use the population of months (or days) instead of a population of trades in the chosen period and then compute the average and standard deviation of monthly returns, we can avoid the problems of systems with a small number of trades. The results of a particular system evaluation obtained from a t-test should be very similar in both cases (trades versus months).

This can be seen intuitively if one compares a system with conventional entry and exit signals to the same system with an extra rule that forces an automatic exit and reentry on the first day of each month: Both systems will have exactly the same overall performance, the same distribution of returns, but in the second case, the number of trades is equal to the number of months.

The probability, p(P/L0), should be the same in both cases.

In the case of in-sample optimized systems, the probability obtained from the t-test must be degraded by raising to the mth power, where m is the number of optimization steps that have been run. While this method seems intuitively correct, I could not find it described in any statistics textbook. Could you indicate where or how it was obtained?

In order to account for the correlation between various tests in an optimization, and reduce the amount of degradation introduced with the indicated formula, I suggest the following adjustment: Measure the correlation coefficient (r) between the series of returns obtained with the parameter to be optimized in its basic state and the series of returns with each new parameter value.

We shall therefore:

(a) compute m - 1 correlation coefficients: r12, r13, r14, ..., r1m, and

(b) compute (1-p)s, where s = 1+ sum(1- r21i )

Obviously, in all cases, except if all r1i are equal to zero, s will be smaller than m, and will decrease as the r1i coefficients increase; if all tests are uncorrelated (all r1i = 0) then s = m and we return to the original case described by Katz; if all tests are perfectly correlated (that is, the parameter under investigation has no effect on the result of the system), s = 1 and there will be no degradation; in all other cases, the degradation will be somewhere between the two extreme cases.

SOFTWARE COMPARISON TABLE

Editor,

I bumped into your magazine's Web site while looking for information over the Internet on technical analysis software. I found your Software Comparison Table interesting. Is it possible to get the conclusions from this study?

I have to say that I was happily surprised with the contents of your I-net pages. You are bookmarked!

Thank you for visiting our site! We are pleased to be a bookmark, and we hope our site content will continue to serve your needs.

The Software Comparison Table represents the results of a survey that we mailed out to various software vendors. We compiled the results into an abbreviated table and published it in our 1997 Bonus Issue. We then published the complete survey responses at our Web site, which doesn't have the same space limitations as the magazine.

The Software Comparison Table was designed to give investors a quick look at some of the software that's available. It is not a rating or ranking system, but rather a vehicle for comparing features side-by-side.

Thanks for writing.?Editor


JUST A FEW LITTLE REQUESTS

Editor,

We are full-time traders and our trading focus is on intraday (that is, trading 1% to 2% price swings) and short-term swing trading (that is, trading for two to 10 days and 3% to 10% price swings). We find very little coverage of our trading focus in your magazine and we would like to see the following in your forthcoming issues.

1. Please give us some good trading systems for intraday trading.

2. We'd like to see more coverage of trading systems. The November 1996 article "Dynamic multiple time frames" by Robert Krausz was great. You also did a good interview with Thomas DeMark back in May 1995. We want more on definite trading systems. Maybe interviews with traders such as Linda Raschke, Robert Barnes and others revealing their systems could be of great help.

3. We suggest you cover proven trading systems that aren't proprietary systems but in the public domain. The December 1996 article "The 2/20-day EMA breakout system" by David Landry issue was good but a little long-term in nature. We want to know more about commercial trading systems too.

S&C is a how-to magazine. We present ideas and techniques to help traders develop their own systems and techniques that work for them.

For articles on a short-term trading horizon, please scan our article abstracts at our Web site (www.Traders.com) for back issues. On the topic of intraday trading systems, perhaps Gary Smith's May 1997 S&C article, "Tape reading and daytrading stock index futures," will interest you.

For my interview with Linda Bradford Raschke, see the September 1993 S&C, as well as her article, "Historical volatility and pattern recognition," in August 1996. Not many top traders will reveal their system to you. We hope that readers will get at least one good trading idea from each interview or issue of STOCKS & COMMODITIES.

We present reviews of software in each and every issue. As we've said in these pages many, many times before, we generally stay away from reviewing commercial systems for two principal reasons: first, as an educational, how-to publication, we usually don't review blackbox products for which the rules and logic aren't revealed; and second, we can't track systems for six months and report the results; that isn't part of our editorial scope.

We've often reviewed software packages that offer compilations of trading systems and have reviewed newsletters that present trading systems. Again, please check the article abstracts at our Web site for such products.

Thanks for all your suggestions.?Editor

OPTIONS OPEN INTEREST

Editor,

I have been subscribing from Germany for two years now and like the ideas I get from S&C.

I trade index options and I'm wondering if you've published any research about using open interest data from the options markets to forecast where the index will close on expiration day.

Over the past two years I did my own research, and I discovered that, by looking at whether more calls or puts are being sold, I could figure out where the index didn't close on the third Friday. I have used the DAX options for my research, and this specially developed indicator failed only in two months out of 24.

Now I'm the head trader of a private investment club and we use this technique every month to write options on the DAX index. My research indicated that this same indicator could also work on the OEX. Have there been articles in past years about this subject, or am I the only one who uses this approach to squeeze some money out of the market?

I also want to tell you that the interview with Constance Brown in the June 1997 STOCKS & COMMODITIES was the best in the last two years because it focused on the psyche of the trader rather than on indicators.

Thomas Bopp
Niddatal, Germany
Sounds like you have the makings of a good article. I encourage you to visit our Web site for our Author Guidelines. In addition, see my interviews with Bernie Schaeffer (S&C, November 1997) and Larry McMillan (S&C, December 1996) on the topic of options trading.?Editor

PRODUCT DECISIONS

Editor,

I have been trading relatively successfully for the past year. I subscribe to your magazine, but I'm still a bit confused on the question of software, real-time quotes and technical analysis. Let me explain. I've been looking into software to buy, and I've found that some companies offer a real-time package for a monthly fee, some offer a program to purchase outright with an ongoing monthly fee for data, and others offer end-of-day packages with no monthly fees.

I'm confused because I simply don't understand the distinctions between the packages and which direction to go in.

You didn't indicate whether you are a real-time trader or end-of-day trader, which would obviously affect whether you select a real-time package or end-of-day package. Please visit the Back Issue Archive area of our Web site (www.Traders.com) and browse our article abstracts for past product reviews.?Editor

MORE PRODUCT DECISIONS

Editor,

Certainly, you must hear this all the time, but I'd like to comment on the high quality of your publication. Please continue to produce such an informative publication.

My reason for writing is to seek advice on software. I have consulted your 1997 Bonus Issue for the Readers' Choice Awards and Software Comparison Table as well as your Web site for the Software Comparison survey results. Unfortunately, after contacting the companies whose software interests me, I am still somewhat perplexed.

After having used an end-of-day charting program for nearly three years, I am considering moving up to something more advanced. I am considering a real-time package, a program with an expert system or a program with system-testing capability. Can you tell me which one I should buy and what programs are worth their price?

Chad Larson
via E-mail
Products' features will range with their price, and you're going to know what features you need and want better than we will. Keep reading the product literature to decide what features you need.

We try to present readers with information about software packages in the way of monthly product reviews, Quick-Scans, Trade News & Products, and the Bonus Issue; the actual buying decisions must be left up to the individual to suit individual needs.

Thank you for your complimentary feedback.?Editor

S&C INFO
For subscription service questions, address changes or ordering information, call 800 832-4642 or 206 938-0570. Or write to STOCKS & COMMODITIES magazine, 4757 California Ave. SW, Seattle, WA 98116-4499. Or visit our Web site at https://www.Traders.com.

Single back issues from the current year (subject to availability) are $8 prepaid. Prior years are available in book format (indexed and without ads) or in CD-ROM format. Currently, Technical Analysis of STOCKS & COMMODITIES Volumes 1 through 14 are available for purchase (see page 11 of this issue for ordering information).
 

To provide us with feedback on the articles in each issue, locate and tear out the yellow card that can be found in each issue, check off your favorite and least favorite articles and drop the card in the mail.
 

Any reader interested in submitting an article to STOCKS & COMMODITIES may request our Author Guidelines, which describe categories of topics accepted, style and submission requirements, copyright information and remuneration. Write to Author Guidelines, STOCKS & COMMODITIES, 4757 California Ave. SW, Seattle, WA 98116-4499, or call 206 938-0570. Or view the Guidelines at our Web site at https://www.Traders.com.

Back to Contents