Optimizazion results
2 replies
massidm
6 years ago #117889
Hi there,
i was making a run for optimize a strategy which initial data were 13k usd net profit, winning rate 53%, profit factor 2.8, return/drawdown 9, stability 0.6.
I set to rank results for weighted fitness, including the parameters i just mentioned (net prof, pf, win, r/dd, stability).
At the start of the optimization (genetic, so 15.000 runs), in the databank i saw a lot of strategies that surpassed the original strategy for at least 4 out of 5 parameters. Then they slowly were replaced, and at the end i had only strategies that were well below the performance of original strategy. And i included 5.000 results in the databank, so 1/3 of the total runs.
Why this is happening?
thanks for help
tomas262
6 years ago #145143
Hello,
it could be caused by the fact that you optimize for the best fitness function value. So even if the strategy has lower Net Profit for example it can still have higher fitness function value because of other ‘better’ performance metrics so SQ will prioritize it. If it was a bug you would be able to see it when you optimize for a single criteria like Profit Factor for example. In this case the optimization process should always prioritize and store strategies with higher PF
massidm
6 years ago #145146
Hi Tomas,
yes, this is the case. Even optimizing for another parameter (i.e. ret/DD), it replaces lower fitness values with higher ones. How to solve it?
thanks for your help
Viewing 2 replies - 1 through 2 (of 2 total)