Thanks everyone

ironhak wrote:

Thank you! So if my collection has 300 strategies and the generator is still running, if it finds a strategy with a Sharpe better than the worst of the 300 I have, it'll replace the worst one with the new found one?

Basically yes, but it might not be a direct replacement, it will not take the last place unless it is the smallest Sharpe of the 300.

Yes sure, the worst would be dropped to make room for the new one (which could also be better than the first one)

Thank you! So if my collection has 300 strategies and the generator is still running, if it finds a strategy with a Sharpe better than the worst of the 300 I have, it'll replace the worst one with the new found one?

Basically yes, but it might not be a direct replacement, it will not take the last place unless it is the smallest Sharpe of the 300.

ironhak wrote:

Hello

Even if I run the generator without any filter/acceptance criteria, not all generated strategies are moved to collection, but only a fraction. I see this comes from the "search best" parameter.

But how this actually works? Assuming I have no filter and I set "search best Sharpe ratio", then what is the filter? All strategies with Sharpe > 0 will get added to collection or what? What if I used "search best R-Squared", only those with R^2>x will get added? If so, what x level?

Thanks

If it searches for best Sharpe ratio, like in your example, then it adds strategies, which are current strat's Sharpe > collection's worst Sharpe. The same applies for R-Squared of whatever you have chosen. Addition to collection happens when a newly found strat outperforms the worst one in the collection, in other words a new better one gets added to the collection and by doing so it pushes out the worst performer.

Thank you! So if my collection has 300 strategies and the generator is still running, if it finds a strategy with a Sharpe better than the worst of the 300 I have, it'll replace the worst one with the new found one?

Also I wanted to ask, with a generator config. like this:

Using "Out-of-sample" will apply some sort of additional filter? I.e. if a strategy did well IS, will it be added to collection regardless of what happened in the OOS period?

Thanks

No, what would be the point of that? OOS performance is taken into account, if it fails, it's discarded.

Hello

Even if I run the generator without any filter/acceptance criteria, not all generated strategies are moved to collection, but only a fraction. I see this comes from the "search best" parameter.

But how this actually works? Assuming I have no filter and I set "search best Sharpe ratio", then what is the filter? All strategies with Sharpe > 0 will get added to collection or what? What if I used "search best R-Squared", only those with R^2>x will get added? If so, what x level?

Thanks

If it searches for best Sharpe ratio, like in your example, then it adds strategies, which are current strat's Sharpe > collection's worst Sharpe. The same applies for R-Squared of whatever you have chosen. Addition to collection happens when a newly found strat outperforms the worst one in the collection, in other words a new better one gets added to the collection and by doing so it pushes out the worst performer.
EDIT: the last sentence applies for when collection is full, at its maximum count, then better ones push out worst ones.

Using "Out-of-sample" will apply some sort of additional filter? I.e. if a strategy did well IS, will it be added to collection regardless of what happened in the OOS period?

Thanks

Even if I run the generator without any filter/acceptance criteria, not all generated strategies are moved to collection, but only a fraction. I see this comes from the "search best" parameter.

But how this actually works? Assuming I have no filter and I set "search best Sharpe ratio", then what is the filter? All strategies with Sharpe > 0 will get added to collection or what? What if I used "search best R-Squared", only those with R^2>x will get added? If so, what x level?

Thanks