Chuck LeBeau's System Traders Club
BULLETIN Vol. 1 Number 4 July/August 98
Table Of Contents:
Informative Letter from Mr. Edward Borasky( Various Topics)
Traders Toolkit:Adaptive Channel Breakout Using Adx( with code)
Visit Us At http://www.traderclub.com
Our new web site is still under construction but several pages of it are already functional. We anticipate that the web site will soon replace the mail as our primary method of communication with club members. We have made it easy for you to keep track of the progress of our work at the new site. Just open the home page and then go immediately to the "What's New" page where you will find a daily log of our latest improvements and additions to the site.
The System Traders Club continues to grow by leaps and bounds and we now have more than a thousand members from nearly forty different countries. We have done very little advertising lately so most of our new members are coming from referrals. We wish to sincerely thank all of you who have told your friends and associates about the club. Because we have been providing all of the Club Bulletins and the "25 X 25" Bond system for free we have spent thousands of dollars on postage, printing and mailing expense. The enthusiastic interest in the Club is very gratifying but it has caused us to spend a great deal more money than we had ever anticipated. Communicating via our new web site will be cheaper and faster and should allow us to continue to provide many services for free. (Our sincere thanks to Robins Trading Company for generously offering to pay us for an ad on the back page of the Bulletin. Their ad helps to cover some of our postage expense.)
Perhaps the biggest advantage of the web site is that it will allow the members to easily and quickly communicate with us and also with one another in our Members To Members Forum.
Getting all of the web pages up and running will take some time and may in fact become one of those never ending projects. However, I think all of you will be pleased to see how much progress we have made in a very brief period of time.
If things go as planned you may now be reading the last Bulletin to be distributed by regular mail. Make sure we have your current email address because we plan to send the next Bulletin electronically. Most of you sent us your email address on your membership application. If there is any doubt just send us an email message containing your name, city, and preferred email address and we will promptly put it into the computer. If you don't have email please let us know and we will continue sending you our information by regular mail.
Please visit the web site soon and give us any comments and suggestions about what you would like to see there. This is your club and this is your site. Lets all work together to make it something really useful. Now is obviously the best time to give us your ideas.
Members To Members
The pipeline has been opened and the flow of information has begun. In this issue you will find several valuable contributions from club members.
Have you got an idea for a system? Do you have some useful information to pass along to fellow club members? Is there a product or service you would recommend? One of the universities here in California has a motto: "Freely you received; freely give." This motto should apply to the exchange of information within the club. The price we are asking for this valuable newsletter is your participation.
If you are a new trader you can participate by asking questions. There will be many other members who will benefit once we track down and publish an answer. If you are an experienced trader and systems developer please share some of your hard earned knowledge. We aren't asking you to give away your best trading secrets if you think it might have an impact on your trading results. Some idea or insight from you might be just what another member needs to solve a problem. Our members will be very grateful for your contribution.
The club is off to a magnificent start, lets all try to keep the momentum growing and establish a pattern of participation that will benefit us all in the future.
From Riccardo In Italy
I have been developing trading systems for more than 8 years and I think I know many techniques to exploit the market but many little details can make a big difference in our results. It seems that I never stop learning because the markets are such a challenge and I welcome every opportunity to participate in the sincere sharing of ideas. I appreciate your work is in this direction.
Here is an idea I would like to share. Have a nice day. Riccardo
Riccardo's system: This is a trend following system based on an adaptive channel breakout. It is the same system I teach in my technical analysis courses in Italy.
The length of the channel is a function of the trend: Strong trend = few days in channel. Weak trend (choppy market) = more days in channel. X represents the number of days of the channel breakout. X = 150/ADX(14 bars)
To implement the buy and sell breakout signals simply follow these instructions:
BUY HIGHEST(HIGH, X) + 1 POINT STOP;
SELL LOWEST(LOW,X) - 1 POINT STOP;
This is the most basic version of the system and it is employed as a reversal system without stops. There are just two variables and both the length of the ADX and the level of the 150 numerator can be optimized.
We need to use a more complex version of the system when X is less than 1 or greater than 40. Here is an example:
INPUTS: LEVEL(100), LENGTH(14);
VALUE1 = INT(LEVEL / ADX(LENGTH));
IF VALUE1 > 30 THEN VALUE1 = 30;
IF VALUE1 < 1 THEN VALUE1 = 1;
BUY AT HIGHEST(HIGH, VALUE1) + 1 POINT STOP;
SELL AT LOWEST(LOW, VALUE1) - 1 POINT STOP;
Again this is a very basic system that I am sending simply as an example to illustrate the concept of using one indicator to make another indicator adaptive. I like adaptive indicators and I hope that members will share some of their ideas for adaptive indicators with me. RICCARDO
Editor's Comment: Thanks Riccardo. I also like adaptive indicators and the logic of what you are doing seems very straight forward. The ADX is one of our favorite measures of trendiness and using it to vary the length of a channel breakout makes sense. The higher the ADX, the more trendiness in the market and the quicker you should react to potential entry signals. I had been giving some thought to a similar idea of simply subtracting the ADX value from a number (70 for example) to determine the length of the channel to use.
Maybe some of our more mathematically inclined members can give us some input on the best way to implement this logic. The ADX tends to oscillate between values of about 10 on the low side to as high as 50 or 60. In theory it can go higher or lower but it seldom does. (At least over the time frames that I monitor in the futures markets.)
However, I have strong reservations about reversing positions with this logic. If a market is trending strongly, I'm not so sure I would be too quick to go the other way. Seems to me we might want to quicken our entries and lengthen our exits. But perhaps that is just my trend following bias showing.
Riccardo has given us an excellent idea that might become the foundation of some worthwhile systems. Thanks for sharing.
Big Dipper idea from Franz in Austria:
Try trading two contracts per entry signal instead of one on the Big Dipper Bond System. Then take a 3 ATR profit on the first contract and hold the second contract as per original exit. This makes two systems out of one: One long term and one short term. The quicker exit has very high percentage winners.
Editor's Note: Good idea Franz, I'm passing it along.
A member in Chicago sent the following article (Thanks Charlie):
Economist Sees Severe Year 2000 Recession Probable
WASHINGTON (Reuters) - As the clock ticks toward the start of the year 2000, a leading Wall Street economist says the odds have risen that computer malfunctions will send the world into a severe recession.
"The fact is, there are only 550 days, and only 377 business days, until Judgment Day for our computers on January 1, 2000," Edward Yardeni, chief economist at Deutsche Bank Securities, said in a paper presented at the National Association of Manufacturers. "Progress is occurring, but not as fast as the year 2000 is approaching." Yardeni said the probability of a recession had increased to 70 percent from his previous forecast of 60 percent, and he said inaction on the part of global leaders and slow progress by the U.S. government had increased the likelihood of a crisis.
"I can no longer say with any confidence that there is enough time to avoid a severe global Y2K recession," Yardeni said.
The so-called "Millennium Bug," or Y2K (Year 2000) problem as it is known in the computer industry, stems from time clocks in computers that recognize years in only two digits and will not be able to differentiate between the years 2000 and 1900. Power outages and disruptions to telephone service and other vital services could result if the snafus were not fixed on computers that run those systems.
Yardeni said firms that have not addressed the problem could be cut off by others that have fixed it and could fail as a result. A panic could ensue if the problems were widespread and consumer confidence could plummet, he said.
President Bill Clinton has established a "Y2K Conversion Council" to look into how the federal government can address the problem. But Yardeni said a more aggressive effort is needed to set goals and priorities and establish contingency plans.
Yardeni, who was named by the Wall Street Journal as the top U.S. economic forecaster in 1997, is well known for his predictions of a Year 2000 recession. He said the downturn could be accompanied by deflation, or a cycle of falling prices.
Editors Comment: This discussion may seem a bit off topic and we are all probably getting as bored by Y2K articles as we are by El Nino. However, I worked with Ed Yardeni for several years when he was the chief economist for E. F. Hutton and I have the highest respect for him and his work. When Ed speaks, people should listen! Maybe the problem really is bigger than most of us realize. Scary thought.
Suggestion For Current Topic
To help get more dialogue started I would like to suggest some topics for members comment. Your response can be as brief or as lengthy as you wish. Send it to us by email if you like. All letters we publish will be edited so don't worry about format, grammar, spelling or typos. Make it easy on yourself. Here are some possible topics:
1. Please tell us what method, strategy, or indicator has solved a particular problem for you. Tell us the nature of the problem and how your solution came about.
2. Please tell us about your favorite market. Why is it your favorite? What strategies do you think work best in this market? What markets do you prefer not to trade? Did you have bad experiences? Can you offer any advice to members on what are good or bad markets?
3. Do you need help solving a particular problem? Be specific. Good solutions are usually dependent on a clear statement of the problem.
4. Are you willing to help other Traders Club members in any particular field? Can we post a best time and method to contact you? Do you want to exchange ideas on a particular market or trading strategy?
5. There are also some topics that we would like to hear about but do not intend to publish. If you have had a good or bad experience with a product or vendor we would like to know about it privately. We may share this information if one of members should make a personal inquiry. We don't want our publication or web site to become a vendor bashing forum. There are already other publications for that.
Questions From Members
Question regarding multiple signals: Perhaps some of the members could give me their opinions on this question: If one is trading multiple, separate long and short trading systems on one market and receives
(a) both long and short entry signals the same day, does one execute both signals or does one let them cancel each other? and
(b) more than one entry signal in the same direction (long or short) on the same day, does one execute only one signal or does one treat it as an opportunity to "double up"?
Editors Comment: Tough question. Not sure there is any right or wrong answer. Do any of our members have an opinion they would like to share with us?
Your question reminded me of a paragraph in an interview in Futures magazine. (July 98 page 98.) Here is a quote from this interview with Dan Byrnes, a very successful CTA:
"Byrnes' system includes six trend-following components, such as volatility-adjusted breakouts and methods that enter existing trends on momentary pullbacks. Each component acts as a mini-system, giving signals to buy, sell or be flat. To determine actual positions, the signals are netted out. For example, if a strategy trading two contracts says buy and a strategy trading three contracts says sell, Byrnes sells one contract. If both say buy, he buys five."
I personally would like to have a filter or tie breaker for this specific situation that would only allow trades in one direction on any given day. Our plan for multiple systems is to design them so that they turn on and off automatically so that there is only one direction being traded at a time. For example let's say that we select the 20 day moving average as our filter. If the prices are below the MA then the short system is turned on and the long system is turned off.
Regarding (b) where you have multiple signals in the same direction on the same day, I would assume that the systems are designed to be traded over different time frames and I would take both the trades with the understanding that one position will have a short term exit strategy and the other will have a long term exit. It is quite possible that the short term trade will be a winner and reduce the risk on holding the long term trade. This seems to me to be an ideal situation rather than a problem. If both systems trade the same time frame I would probably pick the system with the highest percent winners and skip the other signal.
I would be very interested in hearing from our members on this thought provoking question. I hope many of you will respond because this is a question that needs to be properly resolved if we are to implement our plan of trading multiple systems per market.
Question regarding system results: When developing systems, is it important that the system is useful in every market? I feel that reliability of a system over different markets is a sign that it is better "prepared" for the future than one created on the historical data of a specific market.
Editors Comment: Remember that having sound logic in a system is often more important than the details of the historical performance. When trying to take a system designed for one market and apply it to another market first take a look at the logic in the rules of the system and see if that logic should also apply to the new market.
If the logic should apply and the system doesn't work, then look for some bias in the price data. Compare the price data of the first market with the data of the second market. Does one market have a strong bias to the upside while the other has a bias to the downside? Does your system have a direction bias that causes it to work in one market and not in the other? Is one market extremely volatile while the other market is orderly? Try to see on a trade by trade basis what is happening in each market and understand why the system works or fails on each trade.
You should also review each element of your system and see if there is any way to make the system more adaptive to changing market conditions. For example is there any parameter expressed in dollars that could be changed to a percentage of the price or to a multiple of the Average True Range?
Examine the details of the rules of the system and make certain that all of the parameters are as robust as possible. Does the system perform well only if the moving average is exactly 34 bars or does it perform well with any moving average between 10 and 50? If the parameters are very robust with a wide margin for error the system should perform well on any market with similar data.
Consider the second market as your out of sample testing. Don't expect the results to match very closely but they should both be profitable. If not, perhaps there is a flaw in the logic of the system or it has been overly fit to the first set of data.
Let's not forget that markets can be substantially different in nature because of the variety of forces that influence prices. A large contract that is an index of many prices like the S&P would obviously be expected to have substantially different characteristics than a small contract single commodity like corn. I wouldn't lose any sleep if my corn system didn't work in S&Ps and vice versa Hope this helps.
A Lengthy But Very Informative Letter From M. Edward Borasky
Dear Mr. LeBeau:
First of all, I'm honored to become a charter member of the System Traders Club. Your book is the best book on trading system design on the market. (Editors note: Thanks Ed. Flattery will get your letters published.)
Second, I'd like to invite your members to visit my web page, http://www.teleport.com/~znmeb especially my bibliography at http://www.teleport.com/~znmeb/biblio.shtml
And third, you have my permission to print all or parts of this letter in your bulletin. I hope I can contribute some ideas in the future as well.
A little background: I am a computer scientist / applied mathematician with over 35 years of experience. I got interested in computational finance in the early 1980s. I started with options pricing theory, then portfolio theory and finally mechanical technical trading systems. I was a fairly active stock and stock option trader in 1991 and 1992, but my approach was fundamental rather than technical.
Currently I design and analyze systems and I enter contests from time to time, but I'm not an active futures trader. My current "trading" is monthly asset allocation in my 401(K) at work.
My favorite trading system: point and figure charting. I have my own variant which I call micro point and figure. At one time I had automatically generated point and figure charts on my web page, but I shut the program down when the CFTC proposed requiring system vendors to register as CTAs. Even though I was technically exempt because I was not charging a fee, I did not want to risk a lawsuit. I have also done extensive simulations of Donchian's N-day channel rule.
Hardware and software: I don't own any software specifically for trading. I prefer to write my own software in conventional high-level programming languages or spreadsheets. My main computer is an HP100LX Palmtop PC, an 8 Mhz 80186 running DOS 5.0. Most of my programs are in Perl. For example, the Donchian simulations were done in Perl on the HP100LX. These simulations were described in the Spring 1997 Perl Journal; the code is available on their web site at
For high-speed simulations on large data sets I use Forth, a very fast and compact environment well-suited to the HP100LX. And for plotting results I use the HP100LX built-in Lotus 1-2-3 spreadsheet. For historical simulation, I use Pinnacle Data linked contracts, and for my trading contest entries I use free end-of-day data from the Internet.
Measures of system performance: ah, where to begin? For managed money, I'll defer to Jack Schwager's outstanding book, "Managed Futures." For my own simulations, I use a two-step process. First, I simulate the system taking into account as best possible limit days, gap openings, unables and rollover trades. However, I do *not* include an allowance for commission and slippage in this step. I collect a trade log for each simulation which gives for each trade the position, long or short, the date and price the trade was entered, the date and price of exit, how many trading days the trade was in place and the profit or loss as a dollar value. Technically this last result is unnecessary; it can be computed in the post-processing step from the other numbers. Then I post-process the trade logs extensively.
The article in the Perl Journal shows some simple statistics. Commissions and slippage are deducted in the analysis phase; that way I can test a number of different values without having to rerun the time-consuming simulation. The main performance measure I look at is the histogram of the profit/loss values after deducting the commission and slippage. This is probably the most informative picture of performance.
For example, from the mean and standard deviation of the profit/loss values adjusted for commissions and slippage, I can determine the probability that the system actually is profitable -- that it has a positive expectation. Another process I sometimes apply to the trade logs is called bootstrap simulation. With bootstrap simulation, I can
estimate the probability distributions (histograms) of important parameters like drawdowns.
One comment on the bulletins: they should be dated. I'm still thinking about your change in philosophy noted in Bulletin #2. While the case for multiple *diversified* systems is a strong one (again, see Schwager's "Managed Futures"), it's not clear to me without seeing the details exactly how well diversified your systems are. There are only four of them and three trade the same contract!
I've spent a lot of time dissecting Jack Schwager's recent writings, especially "Managed Futures" and his interview in the Commodity Traders Consumer Report (June 1996). His approach makes a lot of sense to me as a statistician. I think for a large CTA like Wizard Trading a diversified portfolio of market-system combinations is the only way to go.
The last time I looked at their web page, Wizard was trading 60-odd contracts. They're a little less specific about what systems they use but I believe they're mostly long-term trend-following. Louis Lukac, the other principal at Wizard, is a big fan of Donchian's channel rule; it always seems to test well against other well-known techniques.
One thing Schwager said in the CTCR interview is that he tests systems on broad ranges of parameters over many years of data for many contracts, then trades all of them, except perhaps the obvious stinkers like a two-day Donchian channel. You are right that this produces a "system that barely works in many markets yet isn't outstanding in any."
But it does produce a system with a positive expectation. If he's got the computer power and the programming talent he ought to be able to make this approach work. In addition, he views managed futures as part of a portfolio that includes stocks and bonds.
When we look at the task of a lone individual trader with, say, $25K or less, however, a very different strategy is required. As Kelly Angle pointed out in "Technical Analysis of Stocks and Commodities" some time ago, a "small" trader can't really diversify as much as needed. The only thing that really makes sense at this end of the scale is something like Alexander Elder's Triple Screen.
In Elder's system, you trade only in the direction of the long-term trend, enter on a stop in the direction of the trend after a reversal with a
protective stop very close to your entry. Your 25x25 system conforms to this philosophy to a certain extent. I would call it a quadruple screen; it trades only long because of the long-term uptrend in T-bonds, trades when the ADX indicates an uptrend and enters after a short-term decline on a buy stop.
Incidentally, the long-only approach based on the "historical" uptrend in bond prices (downtrend in long-term interest rates) is a perfect example of hindsight bias. You're betting that this trend will continue and given a ten-year backtest it will take ten years for you to find out that this trend has changed!
When it comes to mixing systems, for example the 25x25, Big Dipper and Little Dipper, what matters is the *correlations* between them. You designed them "so they are unlikely to enter the market at the same time," It takes more than that, though, to have uncorrelated systems. For example, they are all trading the same contract and backtested on the same data.
One of the things I'm currently working on is a correlation analysis package for my own use. I'm not sure when it will be ready, but I'm willing to offer it to the club. You will need Perl, which is readily available (and free!) for both Windows and UNIX. Trying to code something like this in Easy Language is a major task; I'm not even sure it's possible. There are so many things you can do in a conventional programming language compared to a trading package that I'm frankly surprised that these packages sell as well as they do.
Other random musings and opinions:
1. I like the System Designers Tool Box. Since I like to design my own systems, it is a real help. I think there are better tools for trend detection than ADX, but it does appear to be useful.
2. Although I don't currently trade options or use them, I am *convinced* that traders who don't use options are missing out. At the very least, implied volatility from options prices is an important and worthwhile indicator.
3. My sense as a mathematician is that there *has* to be a better way to analyze futures trading systems than brute-force simulation on past price data. After all, futures prices satisfy the same partial differential equations that are used in options pricing theory, one of the most successful branches of economics and finance. I can't think of a single field in economics with a better predictive record.
4. Along those lines, there needs to be a portfolio theory for futures trading. The closest thing I've seen is Schwager's "Managed Futures" and some of the writings of Ralph Vince.
It isn't clear to me, for example, why one should be better off managing a portfolio of *market-system combinations* than managing a portfolio of raw futures and options. It can be shown that futures and options, along with the underlying commodity or financial instrument and risk-free borrowing and lending, form a complete set of building blocks from which one can construct any feasible return distribution.
5. While the evidence for nonlinearity in financial markets is strong, there is still little evidence for that very specific type of nonlinear behavior known as chaos. People who claim to be using "chaos theory" in their trading systems are probably not exploiting an edge over the markets but the credibility of their customers.
M. Edward Borasky (Received via email)
Ed, I did visit your web site and I encourage members to visit it as well. You obviously put a lot of work into it and I intend to put a link to your site on our links page. I hope you will be among the first to help us launch the Forum section on our new web site by giving us more of your thoughts and stimulating some discussion.
Thanks to all for their contributions to this Bulletin. As you can see, its getting bigger and better as the Club grows.
I've been busy working on the web site and some new trading systems but I promise to have another Toolbox article in the next issue.
Good luck & good trading