Bulletin 2 The Case for Multiple Systems

The Traderclub Forum: Traders Club Bulletins: Bulletin 2 The Case for Multiple Systems

David Elden (Admin) on Monday, February 8, 1999 - 08:20 pm:

Chuck LeBeau's System Traders Club
BULLETIN Vol. 1 Number 2 March/April 98

Table of Contents:
The Case for Multiple Systems
Traders Toolkit:The Chandelier Exit

We were pleasantly surprised and nearly overwhelmed by the interest expressed in the traders club idea. Mailing the responses has kept us busy for several weeks and we have worn a path between our office and the post office. We are especially grateful for the many favorable comments about our "tell it like it is" marketing approach and the many kind words about our book. We hope you will be pleased to hear that our publisher tells us the book is considered a classic and has asked us to work on a second edition. Computer Analysis of the Futures Market (2nd edition) should be released late this year.

Thank you all for your enthusiastic support. We promise to do our best to live up to your expectations.

Our original philosophy about trading systems has changed considerably over the last year or two. We once believed that the best trading results were obtained by employing one system that traded all markets long and short. Then we spent years of research developing such a system. We produced an ADX based system that traded all markets long and short with identical parameters and showed outstanding test results. The system was then traded with millions of dollars and performed quite well for several years. Then it abruptly started producing losses rather than profits. Our research into the causes of the decline in performance led us to conclude that it was not realistic to expect one system to be able to perform well in all markets under all market conditions.

Like generations of traders before us we had somehow assumed that the answer to good trading could be found in creating that one great system. Looking back now we wonder why we ever made such an illogical assumption. No one system can be expected to perform well under all conditions. In addition, in order to perform well across a wide range of markets the system often has to compromise outstanding performance in particular markets. The consequence of this approach is to employ a system that barely works in many markets yet isn't outstanding in any. Typical "do all" systems average about 30% to 40% winning trades and produce long and severe drawdowns that eventually become unsustainable.

Consider now the merits of multiple systems: We would obviously start with a trend following system for the big trends. But we must also have a system that works well when there are no trends. We would trade systems where the profits are taken at target points and have systems that patiently let the profits run. We could design systems for long only and short only. We could have systems with tight stops and systems with wide stops. When we were trading we would have long term, intermediate term and short term systems all running at the same time. We could even use systems that day trade.

Why do we try to make one system do everything? What's the point? Why not design systems for very specific purposes and specific markets? Once a system's purpose is clearly and precisely defined it becomes much easier to create very sound logic for the system. Each part falls into place precisely without needing to compromise.

Ten years ago it might have been operationally impractical to attempt to operate multiple systems. But with today's sophisticated technology any trader with a computer can easily run as many systems as they have capital to support.

Knowledgeable traders have long been aware of the advantages of diversification among markets. Now a few innovative professionals are also reaping the rewards of diversification among systems. In fact a few very large and well known CTAs are currently trading more than a hundred systems at a time. The idea makes good sense.

However, we believe that for the multiple system approach to work as intended the systems need to be carefully matched to avoid duplication and overlapping trades. For example, when we designed the "25 X 25" and "Big Dipper" bond systems, we were careful to design them so they are unlikely to enter the market at the same time because they both trade on an intermediate time frame. The "Little Dipper" was designed to trade on a very short term time frame.Because the systems were designed for a very specific purpose, buying a strong bond market on dips, we were able to achieve very impressive historical results using simple logic. If we had to worry about trading the short side of the corn market with the same system, the problems in developing such a system would have compounded and we would not have been able to produce the same results.

Now many system traders are going to argue that single purpose systems are prone to optimization. This may well be true, but it could also be argued that single purpose systems can produce excellent results without the need for optimization, whereas multi-market multipurpose systems are so difficult to produce that the designers must resort to optimization to obtain acceptable historical results.

One of the problems we faced with our long term trend following system was low trading frequency which tended to make our performance erratic and prone to peaks and valleys. In order to take advantage of our statistical "edge" we might have to complete thirty or forty trades. Using along term trend following system it could easily take a year or two to generate this many trades even with multiple markets. However a combination of good systems should produce a more statistical meaningful number of trades over a much shorter time period. Therefore the trading results should be noticeably more consistent and dependable. Ask yourself: Would you rather be trading one multi purpose trend following system with a low percentage of winners or trading several high percentage multiple systems: Systems designed to follow trends, buy and sell corrections, trade within a range, profit from small and large trends, and trade long and short with equal accuracy. We believe that a combination of properly designed systems, each with a specific purpose, should substantially out perform conventional "do everything" systems.

We can envision the time when savvy traders will insist on operating several carefully designed systems for every market they trade. Where will all these systems come from? We hope they will come from the Traders Club.

My results when testing the 25 X 25 system are different than yours. Why is that?

We have had this question from several members and so far we have been able to help each of them match our test data. In most cases the difference was that the member was not testing using " day session only" data. In a few cases the member was not using data in a format that allowed for prices and ticks in 32nds. In one or two cases the starting dates and maxbarsback were slightly different. Once these differences were resolved the test results were nearly identical.

You mention that your testing was done using day session data only, so I presume the values of the variables were optimized for the day session only. If, later, I want to trade the Globex, can I safely use the same values for the variables, or should I re-optimize them?

We don't like trading the Globex or similar markets that sometimes suffer from a lack of liquidity and frequently show odd prices that can distort our indicators. If you were to look at the combined data including the Globex, the opening of Tuesday's market is really the Monday trading on Globex. This distorts the logic of the system and has nothing to do with optimization. We want to buy when the day session prices are strong and we don't care what the Globex does. We also try to design systems that are practical to trade in real time so we avoid night sessions, 24 hour markets, and foreign markets because we want to "have a life" as the saying goes.

We believe that systems are as likely to be robust as the logic is sound. We don't rely on nearly as much optimization as you might think. In fact I am sure you could optimize our systems and produce better historical than we have, but this would be unlikely to improve the real time trading.

The primary reason we run optimizations is to insure that our chosen parameters fall in the middle of a broad range of profitable values and to make sure that we haven't accidentally stumbled into a curve fit that has little hope of success in the future.

Sound logic is much more important to us than optimized test results. The best insurance of a profitable system is sound logic, not historical test results no matter how carefully they have been obtained.

Can the bond systems be operated using SuperCharts?

Yes. We are going to provide the code on a floppy disk so it can be input into SuperCharts. Let us know if you need a copy.

Will these bond systems also work on the short side?

The test results on the short side aren't particularly encouraging , however the logic of the system should be valid in either direction. As we all know, the bond market has had a sustained period of rising prices so the historical results of any two sided systems are not going to look very favorable from the short side. Because of the obvious data bias we know that any declines have turned out to be buying opportunities.

If we were to go back and create a short side system that did well over the test data it might not be the best system to use during a sustained period of declining prices. Because of the rising trend in our test data any short side system that tested well would have to take profits rather quickly. However a short side system that lets profits run might prove to be the best system if we get into a sustained downward trend. We will work on generating short side systems for bonds, and try to come up with one logical short term system and one logical intermediate term system.

Do the bond systems work in any other markets?

Yes, but not as well as in bonds. Looking at results in other markets is a
good method of out of sample testing. Because these systems are simple and the logic is sound, it is not surprising that they may produce acceptable results in some other markets.

To test the bond systems in markets other than bonds, use the same setup rules and in the case of the "25 X 25" system convert the 18 ticks entry trigger to a fraction of Average True Range. Then use that value as the entry trigger in the other markets. For example, if we assume that 18 ticks is about .6 of one ATR we can buy Soybeans at the previous close plus .6 of one Soybean ATR. (We usually suggest using 50 days of data to calculate the ATRs.) Fortunately the parameters of the systems are so robust that we have a very wide margin of error on exactly how many points we use for the entry trigger.

One of our members in Europe enthusiastically reported that he has obtained positive test results using the "25 X 25" system in world wide financial markets including Canadian, Italian and German Bonds, and the DAX and Hang Seng stock indexes. He is testing both long and short. To adapt the system to these markets he has simply taken the "25 X 25" system and changed the entry trigger to .75 of a ten day ATR and changed the $2500 stop to 2 or 3 ATRs depending on the market. He also reports that his version of the system has tested particularly well on U.S. T Note futures.

You might want to run a market by market optimization to make sure whatever parameters you have selected for the entry fall in the middle of a broad range of profitability.


This will be the first article of a series featuring various techniques we have found to be useful in developing trading systems.

THE CHANDELIER EXIT: We have often advocated the importance of good exits and this is one of our favorites. The exit stop is placed at a multiple of average true ranges from the highest high or highest close since the entry of the trade. As the highs get higher the stop moves up but it never moves downward. (For future reference please note that we almost always give our examples from the long side and if there are any difference on the short trades we will try to point them out.)

Example of Chandelier Exit:
Exit at the highest high since entry minus 3 ATRs on a stop.
Or exit at the highest close since entry minus 2.5 ATRs on a stop.

Application: We like the Chandelier Exit as one of our primary exits for trend following systems. (The name is derived from the fact that the exit is hung downward from the ceiling of a trade.) This exit is extremely effective at letting profits run in the direction of a trend while still offering some protection against any reversal in trend. In fact our research and that of our friend Dr. Van Tharp has shown that this exit is so effective that we can literally enter markets at random and if we use this exit the results over time are likely to be profitable. (If you don't believe us just test it yourself over a handful of markets.) In general the best values for the ATR in most markets ranges between 2.5 and 4.0.


As of the date of this bulletin the "25 X 25" and "Little Dipper" bond systems have been out of the market for a while because the trend does not meet the requirements for our long only trading. (The filters are working as intended and turn off the systems as the market heads down.)

The Big Dipper, which as the name implies, was designed to enter after bigger declines, gave an entry signal on March 23rd and is presently long the June contract at a price of 120.30.


We are looking for several knowledgeable and experienced brokers who specialize in systems trading and who own TradeStation or similar software and are willing to assist our members with their trading. These brokers must be willing to provide members with special services and accommodations that are not generally available to the public. Brokers who qualify will be placed on a list of recommended brokers that will be sent to those club members who contact us and specifically request broker referrals. (Member information will not be given to brokers. The club member must initiate any broker contact.)


Many of you are already members of Club 3000 which has been around for many years. If you haven't seen this publication you should look into it. The members write the articles and there are lively discussions on various trading topics. (Not very technical.) Fred Kastead has taken over from Bo Thunman as editor and is doing a good job of keeping this well loved publication going. For subscription info contact Fred at 616 623-8235 or email him at Club3000@juno.com (As I recall the cost to subscribe is less than $100 per year.)

If you are a serious computer buff and TradeStation user who wants to share information about software and coding you should try the Omega newsgroup. To subscribe just send a message to: Omega-digest-request@eskimo.com There is no charge but you will spend a lot of time sorting and deleting hundreds of junk messages (mostly Omega bashing) and savoring the occasional nuggets of useful information.


Many of you will be pleased to learn that we are working on a short term S&P system. We have also been working on a system for Crude Oil and a short side system for the T Bonds. If you have other requests please pass them along.