Quote:
How much ram need to calculate preflop 6 max at 100bb stacks for big blind with calculate precision 0.05bb /100?
We currently don't ship faster/more precise solver which uses more RAM so once you are able to build the tree the RAM requirements end there.
Quote:
big blind with calculate precision 0.05bb /100?
It won't converge to it for several reasons:
-we save a lot of memory postflop and therefore postflop play is not very precise (still provides good enough estimations of EVs though which is all that is needed for precise preflop solutoi)
-we use a trick which is based on calculating all-in preflop on 1755 flops (instead of a subset) this is good for solutoins but it requires MES/EVs to be approximated a bit, at 0.05bb you are already within the range of that error (this is also the reason the solver won't work well on very small subsets like 2 flops)
-our algorithm in the first place rarely goes to 0 accuracy, again the reasons are: memory usage and behavior on bigger trees with many bet sizes; it's easy to make it converge to 0.01 or w/e, it's harder to make its behavior decent on the kind of trees most of our users want to solve in reasonable time to reasonable accuracy and on reasonable hardware.
Quote:
What is the precision in the calculation(bb/100) preflop with 60 flop subset, 176 flop subset?
We calculate max exploit on the subset (obviously we can't do it on all flops as we don't have solution there in the first place) so the subset size is not relevant (maybe very small subsets are problamatic).
The goal of the preflop solver is to get preflop ranges and maybe see some overall frequencies/behavior postflop without going into too much detail. If you want more precise postflop play - use the postflop one. Fortunately adding bet sizes postflop is pointless when it comes to getting good preflop ranges as is going to very low accuracy overall (as most of it is in postflop play and preflop play is already good enough).