Hi

I am developing a strategy and with WFO=13 and 5 years backtest, I got MI 100$ and Capital 300$. Amazing right? Not really. The equity curve is mostly flat and I just got a huge earnings in a small period of time which in averague provides such a monthly income. If I change the backtest period, then it is a disaster.

My wonder is if it makes sense to perform the same backtest of a strategy in 1,2,3,4,5 years time or in 1 year time but in diferent years and compare all of them. If the PF is similar in all the periods it means that the strategy is robust and it is not overfitted or unprofitable at all.

Such analysis could explain the losses in Z12. I mean, the backtest from 2012 is ok. What about similar backtest from 2014? I know that WFO use diferent periods of time for testing, but it is not the same to begin the test of a strategy from 2014 with 2000$ than to begin the same strategy with 2000$ capital from 2012. A strategy that needs 2000$ in a backtest from 2012 may need 4000$ in a backtest from 2014.

By doing such analysis automatically, one could get a better insight of the profitability of a strategy.

Thanks for the suggestion. It can indeed make sense to start a WFO test several times with different starting points that are shifted over a WFO test period. This multiple test would catch such problems as you've described. I'll do some experiments with this method.