Skip to Content
Our misssion: to make the life easier for the researcher of free ebooks.

Automatic Discovery of Trading Systems: The Next Step in Mechanical Trading

The process of developing mechanical trading systems often leads to frustration and to a financial disaster if the systems do not perform as expected. The majority of those who attempt to develop mechanical trading systems quit sooner or later and if they decide to continue trading adopt an alternative style. As it turns out, developing a winning trading system is an art and a science and requires both quantitative skills and experience with the operation of the markets. To succeed, it is not enough purchasing a fast computer, getting a high speed Internet connection, reading a few good books on technical analysis and attending a few seminars. It is also required that traders invest personal time and go through the learning process of system development. Very few make it to the end of the road and, of course, the end is not the “Holy Grail” but a trading system with an acceptable performance.

An alternative to developing a system is getting a black box that advertises huge monthly returns. However, most traders do not trust black-box systems because they have a poor record of meeting expectations and prefer to develop their own using one of those sophisticated software programs for trading system development and analysis. There is at least a dozen or more such program available in the marketplace at a moderate price. Some are even offered at no extra cost to those that open an account with a broker. These programs have a high-level programming language for implementing and testing trading system and advanced capabilities for statistical analysis of the results. There are at least two problems with the use of such programs. The first problem is that implementing even a simple trading system requires knowledge of programming. The presence of high-level language eases but does not eliminate completely this requirement. The second problem is more fundamental and it has to do with the fact that these “back testing” programs do not offer a clue of how to come up with a winning trading system in the first place -- all you can do with them is to test it after you find it. The burden of finding a system is on the user and this must be done before it is implemented in these programs.

Just a small fraction of traders have the skills and the experience to develop a system suitable for mechanical trading and some of them are very successful indeed and end up making a lots of money trading the markets. Having a winning trading system and the discipline to follow its signals is like a license to print your own money. This is the main reason traders are so persistent in their search for winning systems. For those that do not succeed in finding a good system using the conventional methodology discussed above there is an alternative method: develop a program that automatically discovers trading systems.

In principle, the idea of automatic discovery of trading system is simple. You let the computer do the work based on a set of rules of what constitutes a winning trading system. The computer will search for profitable trading systems that fulfill the user criteria and requirements. The trader is not required to do any work, just inspect the results and see whether the computer did a good job. Since computers do not take coffee break, do not plan for vacation, take no maternity leave and can work three shifts a day, etc., the efficiency of such process can be extremely high. The trader can concentrate on other productive tasks and let the computer search for winning trading systems. This is at least theoretically sound and also an appealing concept but it requires an essential ingredient to be practical: a program that can do the job. The problem is then developing such a program.

Automatic Discovery of Trading Systems: The Next Step in Mechanical Trading