Picking The One from a backtest optimization run
Cesar walks through how he takes the results from an AmiBroker optimization run with hundreds of variations and determines which variation he will trade or do additional analysis on.
Must be a member of The Crew to view this video. For information on joining, click here.
Thanks Cesar, some interesting points there. I liked Steven’s comment that, in the end, even with pure quant strategies, we use some discretion in selecting which versions to trade. Can’t get away from what makes us human beings…..
A question: do you test on all parameters in one session? You showed 6 in this example – what if you have more, and it would take many hours or even a day or two to complete? Do you then select the most important ones, and then repeat with the lesser ones?
P.S. Nice strategy – looks very robust.
When creating the strategy, I try different filters seeing what helps and what does not. Once I get to set of filters, I like then I will run an small optimization on that. From that I will see what values work better and rerun an optimization around those values.
It would be super helpful if you could briefly explain how you create this spreadsheet.
A short video would be even better of course.
I just can’t get it to work.
The spreadsheets are always great to get an overview of the system.
There is nothing really special about the spreadsheet. I take the output from the optimization run and paste into Excel. Then I hide the columns that I did not care about. Then turn on Filters in Excel.
I was an idiot, now everything works as expected.
Even if it means more effort, it would be good if you trade several versions of a system with completely different parameters right?
Then the live trading result should also be more stable (closer to the mean of the strategies).
Yes, that would be ideal. But the problem is that you need to have enough capital such that commissions don’t start becoming a drag on the portfolio