I've read a little bit about walk forward efficiency, but I am still not quite sure how to calculate WFE correctly.
Assume we have N windows (or sessions) in a walk forward optimization.
In each window, the in sample profit per bar or per day is PIn (n=1,2,...N), the out of sample profit per bar or per day is POn.
So which formula is WFE?
1, sum(PO1, PO2, ..., POn) / sum(PI1, PI2, ..., PIn)
or 2, sum(PO1 / PI1, PO2 / PI2, ..., POn / PIn) / N
I'm currently using formula 1 but not sure if it's correct.
More questions on walk forward,
1, The profit should be in money or in pips, or either is OK?
2, Among the optimization windows, there may be some windows, of which both the in-sample and out-of-sample profit are negative. A simple division of them should result a positive ratio, which is obviously wrong. How to deal with that? Currently I just turn the ratio to negative. And better solution? Of course this is not a matter in the formula 1, but beside WFE, I also count the max and min ratio on the windows.
Thanks
Assume we have N windows (or sessions) in a walk forward optimization.
In each window, the in sample profit per bar or per day is PIn (n=1,2,...N), the out of sample profit per bar or per day is POn.
So which formula is WFE?
1, sum(PO1, PO2, ..., POn) / sum(PI1, PI2, ..., PIn)
or 2, sum(PO1 / PI1, PO2 / PI2, ..., POn / PIn) / N
I'm currently using formula 1 but not sure if it's correct.
More questions on walk forward,
1, The profit should be in money or in pips, or either is OK?
2, Among the optimization windows, there may be some windows, of which both the in-sample and out-of-sample profit are negative. A simple division of them should result a positive ratio, which is obviously wrong. How to deal with that? Currently I just turn the ratio to negative. And better solution? Of course this is not a matter in the formula 1, but beside WFE, I also count the max and min ratio on the windows.
Thanks