I noticed a bug regarding the APR column in the optimization results that gave me a bad WFO result. I was able to get the correct results by using the "best metric: profit" instead of the default "best metric: APR".
I noticed that for one In-Sample period there were no positive APRs and WL erroneously reported some of the worst results as "0.00" APR, which was then selected as the "best metric" for that period. Here is a screenshot of that result:
Here is a screenshot of the same period but this time I optimized for "profit" and it selected the best result this time. You can see the profitable OOS backtest results at the bottom as well. Basically, if I hadn't have caught this problem, I probably would have thrown this strategy out because of how bad it looked with the first result.
I noticed that for one In-Sample period there were no positive APRs and WL erroneously reported some of the worst results as "0.00" APR, which was then selected as the "best metric" for that period. Here is a screenshot of that result:
Here is a screenshot of the same period but this time I optimized for "profit" and it selected the best result this time. You can see the profitable OOS backtest results at the bottom as well. Basically, if I hadn't have caught this problem, I probably would have thrown this strategy out because of how bad it looked with the first result.
Rename
Thanks for reporting this!
Your Response
Post
Edit Post
Login is required