JEsISEsDOCAKTT
02-05-2025 10:54,
I have been programming a couple of EAs in the last 6 months which all work well in backtests but neglect in live scenario. The result is a devastating loss, either a huge gain, looking too good to be true. In no cases there's a relation between the backtest results along with the results.
So I pursued some efforts to reduce overfitting the curve by choosing as best parameters not those that give the biggest returns, but those that seems sensible.
This wasn't adequate and I am starting to think there's a magic number in the amount on parameters. If there are too many, the EA correspond to a very particular situation which is not likely to reproduce in the not too distant future.
So, for instance, the robot I am working on includes a couple of heaps of parameters, on account of the amount of indiors used. I focus on optimizing only 8 of them, hoping it will create a more generic egy.
What's your choosen variety of parameters? Anyone optimizing only three or 2 and having great results? Or is it better to have as many parameters as our CPU could eat?
So I pursued some efforts to reduce overfitting the curve by choosing as best parameters not those that give the biggest returns, but those that seems sensible.
This wasn't adequate and I am starting to think there's a magic number in the amount on parameters. If there are too many, the EA correspond to a very particular situation which is not likely to reproduce in the not too distant future.
So, for instance, the robot I am working on includes a couple of heaps of parameters, on account of the amount of indiors used. I focus on optimizing only 8 of them, hoping it will create a more generic egy.
What's your choosen variety of parameters? Anyone optimizing only three or 2 and having great results? Or is it better to have as many parameters as our CPU could eat?