Another question from Hanns K. about the Strategy Cruncher and adding filters to your strategy to make more profit with fewer trades.
Hanns K.
I’ve read that you shouldn’t optimize with a too high “resolution” in order to avoid overfitting and not to create too high expectations of the strategy’s performance, since it probably will never reach this historically optimal performance in live trading. I have the feeling the Cruncher is doing that to some extent and that the brute force method probably is a bit more conservative by doing larger step sizes during optimization.
Dave:
Good question, Hanns.
I see this as a variation of the round number fallacy, where traders choose round numbers for the values for their rules even though it’s not optimal as a way of somehow avoiding curve fitting.
The strategy and the market don’t care that you’ve chosen a round number that appears more “pleasing” compared with the optimal number.
Also, let’s take the precision argument to its logical conclusion.
If you think using less precision will be better in the long run, then maybe you should use a LOT less precision – that would be even better according to this theory.
So let’s say the Cruncher suggests a value of 5.6 for gap percent for improving your strategy. That is, if you were to apply the rule and only take trades where the gap percent is greater than 5.6, you would get the same profit with fewer trades.
In an effort to supposedly avoid curve fitting, you choose 6 for the value (rounding up from 5.6).
If that’s somehow an improvement, why not choose 10? That’s an even more pleasing number!
You can see where that leads.
The best number to use is the one that has shown itself to be optimal over a long historical period.
Thanks for the question, Hanns K.
-Dave
P.S. Are you still fooling with Excel to try to create your trading strategy? The Strategy Cruncher lets you generate more profit from fewer trades in your backtest in seconds, not days or weeks.
