Interpreting R2 vs other metrics

One of the steps I go through in building a bot is determining how many trades a particular logic should take. I start by Testing 99, and then check the Performance Report to see the maxtrades that were ever taken. From that figure, I work my way backwards to 1, to determine which combination is best.

I noticed that in some cases as I approach maxtrades 1 (which also limits the total # of trades), that the APR and PF will increase while R2 decreases. I should also note that I am using MonteCarlo 2000 simulations @ 95% confidence.

So I'm curious if it's more important to maximize R2 or to maximize the APR/PF figures? For example I could choose the level that yields the highest R2, but it would not be the highest APR/PF at that level.

Thank you

The R2 is more subject to random fluctuations as it depends on the order of trades. I believe AR and PF are more stable criteria.