Optimization Observations

[brooksrimes]

In optimizing, I find that using Return on Account for fitness (vs. Net Profit) during in sample testing, results in parameters that work much better in out of sample testing. In fact, I find that optimizing on Net Profit, often leads to marginal or negative out of sample results.

Would be interested in hearing others experience in this area ...

Brooks

 

[Adamus]

How about Return on Account divided by maximum drawdown?

 

[brooksrimes]

That's a good fitness criteria also. Is there a common name for that, something like "recovery ratio" ...?

 

[Adamus]

I haven't heard of a name for it. It's not available as a option in NinjaTrader's system optimizer. "Recovery ratio" sounds good but sounds like one of those stats you use to measure drawdown per individual trade. I guess I could have just said "profit over drawdown" and you would have understood.