Strategy Tester Optimization In A Nutshell?

New comment
 

I was hoping someone could help explain to this newb (in a nutshell) how you optimize a new EA with the strategy tester?

I've figured out how to test different variables, and the results are generated in the Optimization Results tab...however, I am not sure how to pull the best sets of data from the pile of data that is produced after 500 passes. How do you select which variables to keep for the live version based on the optimization results?

 
Calpurnia wrote >>

I was hoping someone could help explain to this newb (in a nutshell) how you optimize a new EA with the strategy tester?

I've figured out how to test different variables, and the results are generated in the Optimization Results tab...however, I am not sure how to pull the best sets of data from the pile of data that is produced after 500 passes. How do you select which variables to keep for the live version based on the optimization results?

I use Copy All and then paste into excel. then run sorts, for example highest profit, lowest drawdown.

the other thing I like to do is run mode to get the most frequent top 30 value for a particular parameter.

for Example select the first 30 or so cells in a column, at the top enter =mode(a2:a32) and you will get the parameter which occurs most often.

In MHO it is best to have the parameter which occurs most often rather than an individual result that is pure curve fitting.

HTH and I'd be interested in other's thoughts and opinions.

Keith

 
Thanks! I'll give it a shot :)
 

In the Optimization Results tab, you can click on the column header directly to sort the column you are interested in.


I normally sort on Profit, Profit, Drawdown$ (in no particular order). Something with high Profit, but also high Drawdown is not necessarily the best. Then again, something with low Drawdown may also have very low Profit, so there is not much point.


Something else to watch out for is big differences between neighbouring steps in parameter settings. For eg, you may test a parameter setting from 1 to 5. For 1,2,4,5 you may get very low Profit. For 3, you get a very high Profit. This is perhaps not as good as 1,2 being low, 3 being medium and 4,5 being high Profit. Don't ask me for a solid mathematical proof for why one may be better than the other, because I don't have one. :)

New comment