Is walk forward matrix curvefitting?

When using the walk forward matrix you look for the best number of runs and % out of sample on the total amount of available data. I think this is curve fitting or not?

 

Regards,

 

Dennis

 

 

Hello Dennis,

 

you should look not for the best but for the most robust. Robust means that the “best” field is surrounded by as many good results in adjacent cells as possible.

So you should choose 3×3 cells where all or big majority of them passed,and then use the middle one.

 

This “ensures” that the strategy was not curve fitted to one particular combination of run and % OOS, but that it works also for slightly changed combinations of these.