The Guts of Walk Forward Optimization

Looking at the WFO report from the previous post, below:

we have some very valuable, in fact invaluable, information. These Out Of sample (OOS) numbers are the best estimate we can have of how an optimized algo will perform live. The one deficiency is that the impact of slippage is missing. Slippage will possibly happen on non limit orders. ooking at my algo, it's the losing trades that may have slippage. The winners enter on limit otders.

So to allow for this possible slippage, I must adjust the size of my average trade to account for this. If I have a 65% win rate then there can be slippage on 35% of my trades. I need to look at how the algo trades and estimate what percentage of that 35% will have slippage. Worst case is 100% of those will have slippage.

In the above example, there are 793 trades. Of these 35% or 278 trades can have slippage. If I say that all 278 will have a tick slippage then my $83,250 profit needs to be reduces by 278 times $12.5 ($3475) making the OOS adjusted profit $79,775.

There are lots of other available metrics that gives me more information  to avaluate the robustness of my algo.