|
PureBytes Links
Trading Reference Links
|
> We want to do analysis of our trading systems. We are looking to
> analyze 10-50 variables used in each of 10-30 stock trading systems.
> The goal would be variable set selection.
The first thing that leaps out at me is that you are analyzing
FAR FAR too many variables. With that many degrees of freedom,
you would need many thousands of trades (with each system) to
have any confidence that your results are not curve-fit.
It sounds to me as though your models are extremely complex.
Complex models, in my experience, seldom perform well out-of-
sample. That's because no matter how much backtesting you
perform, you are almost always cherry-picking anomalous results
that are caused more by random chance than by actual repeatable
trading choices.
Gary
|