The Sunny Side of the Street    
Your browser does not support Java, needed to view this page.

1999

BREAKING NEWS
Headlines
Press Releases
Weather

HOT DATA
Top 10
IPOs
Whisper Nbrs
Economy

CALENDAR
Users Groups
Conferences
Industry Events
Economic Rpts

MEDIA
 Television
Zack's
Radio
Newspapers

PERFORMANCE
Tables

DISCLAIMER

 

 


Charts courtesy of BigCharts.com

TESTING, TESTING, 1, 2, 3... 

There is only one reason to trade the markets: To make money. Trading for a thrill is not a rational choice, and is usually short-lived because losing money generates just as much of an adrenaline rush as making it. Trading without a business plan, without a clear-cut system, can be fun, but it's dangerous and costly.

As a mathematician and programmer, my view of the world is based in symbolic logic and if ... then statements. I'm a strong believer in cause and effect relationships. Thus, as a trader, I view the markets in systematic fashion, with mathematical definition. Decision trees (if this happens, then something will follow) rooted in mathematics are the basis for my trading.

Some say they trade on intuition, and that is their skill. My premise is intuition is based on an unconscious set of rules discovered over a long period of time and those rules can be translated into a computer program. Those who trade with a set of clearly defined rules can immediately translate them to a program.

There are many traders who test their rules by hand, with paper, pencil and sometimes a calculator. There's nothing wrong with that. However, these methods are time consuming and most traders have a tendency to give up after testing only a small subset of data.

Further, the hand testing method can be prone to two types of errors: Calculation and wishful thinking. A calculation error is obvious. It occurs when a mathematical mistake is made. But, a wishful thinking error typically occurs when a trader says, "I would have taken that trade even though this system doesn't exist." It's akin to outsmarting the system, but is subtler.

Another commonly used method of testing a system is through the use of a spreadsheet program, such as Lotus 1-2-3 or Excel. In the beginning of my trading experience, that's what I did. In fact, I still use spreadsheets to verify the accuracy of test results from other computer programs. Yet, the limitations with this method include the amount of data a spreadsheet can handle and the convenience and clarity of the output.

Computer testing of systems is like running a scientific experiment. Hypotheses are generated, assumptions are made and the experiment is set up so as not to prejudice the results with the bias of the tester. A trading system, tested over the largest set of data available, is valid only after rigorous testing and verification.

There is much discussion in the trading community about statistics. The most commonly asked question seems to be "how many trades does it take to be statistically valid?" The answer I have heard most often is 30, but I couldn't disagree more.

Testing a system over a set of data that produces only 30 trades would generate a nice model, which would work well in the past. However, as the markets fluctuate and the mood changes from bull to bear to sideways and back again, the model with 30 trades probably would prove to be accurate only over that particular subset of data.

If a system passes historical testing criteria and produces favorable results, then it must be traded without variation. If the trader's testing is accurate, then the key to success is to hang in there. A trader should not give up during periods of drawdowns. Drawdowns are an unavoidable part of trading. Computer testing allows the most complete and accurate analysis of market data.

Computer trading is similar to flying a plane. First you must test the plane's instruments and test them again. Then you must rely on those instruments and follow your flight plan. The same is true in computerized trading. Establish your system, test it, retest it and then trust your instruments. Trading then becomes as easy as one, two, three.

ARTICLE ARCHIVE

SUNNY'S ARTICLES ON OTHER WEBSITES

COMMENTARY ARCHIVE

2001