|
|
|
|
|
|
|
The individual investor will stop trading at the point that he loses a certain amount of money from his account. If the individual can use a system that matches or exceeds my criteria above, he will continue trading for a long time. |
|
|
|
|
|
|
|
|
Neal: Let me ask the statistical guru of the group what statistical tools he uses to trade. |
|
|
|
|
|
|
|
|
Committee: When I first started developing trading systems, I began with regression analysis. This analysis was developed back in the latter part of the 19th century. The purpose of regression analysis is to develop a multivariable predictive equation. I used Stepwise Regression. This analyzed all the variables I supplied, chose the best one, and then went on to choose the second best, and so on. This analysis would choose only the worthwhile variables and leave out the "garbage" variables. I began with 60 variables, and the resulting formulas usually contained anywhere from two to ten variables. After this, I graduated to Genetic Algorithms to test different variables. I usually tested several hundred variables. I was continuously looking for good variables to use in market prediction. |
|
|
|
|
|
|
|
|
Neal: How did your early efforts turn out? |
|
|
|
|
|
|
|
|
Committee: The results were mediocre. There were two reasons for this. First, with all the variables I tested, both the regression analysis and the genetic algorithm assumed a linear relationship. By linear, I mean "cause and effect." For example, I assumed that if a variable rose 20%, then the commodity's price would rise by 10%. However, this did not happen. I didn't find a meaningful cause-and-effect relationship. I am not denying the existence of a cause-and-effect relationship in the markets. I am only saying that in my early days, I did not find a good predictive variable or combination of variables. The second reason involves a danger that can affect all of us today, "curve fitting." Every writer of systems has to address curve fitting because it can easily occur. My definition of |
|
|
|
|
|