Help me understand what this implies: "How accurate are the ranges?" on Monkersolver FAQ
How accurate are the ranges?
This is difficult to measure for a number of reasons. We don't have the resources to calculate the exploitability of the full strategy. Even if we did, most of the exploitability comes from the suboptimal postflop play due to the postflop abstraction. How much of this trickles down to the preflop ranges is unclear. The question how exploitable the preflop ranges are given optimal postflop play is impossible for us to answer.
Comparing our preflop limit ranges, which use the same postflop abstraction as our other ranges, to the proven nash equilibrium strategies of Cepheus indicates that our ranges are very accurate.
I find this interesting because the Monkersolver results are generally considered un-exploitable, but here it seems like they themselves state that this is not proven for non-limit variants.
What do they mean by "postflop abstraction" ?
1 Reply
What do they mean by "postflop abstraction" ?
Preflop solvers use techniques like
to group similar hands togther during the postflop solving process. You already select these buckets in the parameters of Monker. For example, instead of AsKd on AcTc4h, it might be represented with a more general "top pair good kicker" class.
Why do they do this? Because preflop->river is an enormous gamespace. Without postflop abstractions, it would not be possible to solve preflop on a consumer-level PC. You'd need thousands of GB of RAM and weeks of solving time to solve a single preflop solution.
Poker Scientist attempted a no card abstraction solve a few years ago -- their preflop ranges ended up looking pretty much the same as everyone else's solutions.
I find this interesting because the Monkersolver results are generally considered un-exploitable, but here it seems like they themselves state that this is not proven for non-limit variants.
The whole idea of measuring exploitability to judge the quality of a strategy breaks down multiway. It doesn't even really make sense conceptually because the equilibrium EV isn't some guaranteed minimum like it is in heads up zero sum games. So in practice we use different metrics to judge the quality of a preflop solution.
