Solution accuracy for only flop/turn
Hello,
When you run solutions in say pio where you will only look at flop/turn subsets, am i correct in assuming then you do not need to run the sims as accurate? Say if you genereally run the solution down to 0.5%, if you will not look at any river solutions will 2% be fine?
3 Replies
2% sounds horrible if you have multiple sizings on the nodes you're looking at, it will look like a wash. For 1 size will look fine probably in a general sense but combo by combo will likely miss nuance
2% sounds horrible if you have multiple sizings on the nodes you're looking at, it will look like a wash. For 1 size will look fine probably in a general sense but combo by combo will likely miss nuance
I ran some btn v bb experiments on 2 flop, 2 turn, 3 river sizes. I think the difference is not that drastic between 0.5 and 2%, the evs for flop and turn seem to have converged to basically the same values combo by combo. The biggest difference occur in flop frequency where all actions for most hands have same ev, or on river where it makes more errors in terms of mixing calls with -ev hands for example.
I ran some btn v bb experiments on 2 flop, 2 turn, 3 river sizes. I think the difference is not that drastic between 0.5 and 2%, the evs for flop and turn seem to have converged to basically the same values combo by combo. The biggest difference occur in flop frequency where all actions for most hands have same ev, or on river where it makes more errors in terms of mixing calls with -ev hands for example.
you're comparing apples to oranges. One is your EV against a decently close to GTO strategy, the other is your EV against another suboptimal strategy, the fact that they're similar numbers doesn't mean it's as good of a strategy, but more so that it's as good of a strategy vs whatever their facing respectively . Differences in frequencies are what you should be comparing.
Having said so, you can run your sims to whatever accuracy you want, no sim is actual GTO and everything is an approximation. But depending on what you want to study you need to be at a certain distance from GTO or it becomes totally useless.
If you want to analyse aggregate data, you can probably get away with less accuracy, the closer you get to studying how you should play a specific hand the more useless that sim will become. "Everything is close in EV" is a bit of a lazy copout, not a good way to study
