|
|
So this talk probably already happened a long time ago, and even if it didn't, you're probably not gonna change your whole approach anyway, but I had a thought about how to talk about poker in a comprehensive mathematical sense (though it would be theoretical to the degree of retardation).
You could hypothetically breakdown all poker players into 5 different types who have very certain VPIP/PFR numbers. There are LAggs (50/25), TAggs (25/20), stations (40/10), live nits (15/5) and online nits (15/12).
Even though in this hypothetical there are only 5 types of players in the entire universe and they have very predictable preflop numbers, their postflop play isn't a pure extrapolation from their preflop play. In other words, even though TAggs and LAggs have similar "aggression stats" preflop, a TAgg is going to have more balanced betting ranges postflop, and they're going to choose different types of hands to bluff and value bet (for example TAggs are going to bluff a lot of draws on flops and turns and bluff a lot of blockers/nut lows on the river, whereas LAggs are more likely to bet their best hands, check/fold their worst hands and check/call things in-between).
When you set up such stringent premises (that aren't true, obviously), then every decision in poker can just be a combination of Bayes Theorem and exploitation theory. So you can play out a hand against someone that you only have a 10-hand sample size against then you can use Bayes Theorem to calculate the likelihood that they are each given player type, and then maximize your exploitation of a player who is, eg, 40% likely to be a LAgg, 30% likely to be a TAgg, 24% likely to be a station, 4% likely to be an online nit and 2% likely to be a live nit. You can play a hand out to the river in a situation where each of those players would have very very different ranges (say, a draw-heavy board where the opponent raised once: the passive players would have really strong ranges, the LAgg would be all over the place and the TAgg would have a balanced range of draws and value hands). The calculations would get kinda sorta on the complicated side, and you would get out of it a certain, optimal strategy on how to play every single hand in your range, which I think would be a good balance of daunting yet cool.
Then, you flip the switch, and instead of the player having volunteered money in 4/10 pots, raising 2 of them, now we say that he's played 3 of them, raising 2 of them, and we see how such a small change can make the player makeup ever so slightly more likely to be an okay player (TAgg or online nit), which changes the balance of how to play the situation. Then, you can make the sample size 1,000 hands larger and show how now you're dealing with player makeups that are more like 99.something% likely to be one type, and now you have much more specific exploitative strategies instead of just taking a mean approach.
I mean, since player types don't work this way, no one uses Bayes Theorem like this in the real world, but I think it's a cool way to simplify the game just enough to be able to talk about math and game theory in a very certain way.
Again, this is probably the dumbest bump ever because no one's going to use it either for poker or for poker talks, but I just thought I'd share.
|
Poker talk
Reply With Quote
Originally Posted by d0zer
