- Independently test all variables. Make sure they can all stand on their own, and then write down only the "profit cluster" for instance, if you run 2-50, and in your optimization report you see that only 15 -30 have consistent profit, then only feed those to the WFO.
- Try Different Criteria in the WFO, I find that optimizing for Pessimistic Return on Capital makes everything better, not just passing the test.
- if you use an indicator, don't optimize the length and the overbought oversold levels, only do either or. Optimizing both will guarantee you are curve fitting.