Solving Poker

jimdmcevoy · #1 11-04-2004, 04:57 PM

I am curious, does anyone know if anyone has ever tried to solve a specific form of poker perfectly?

I am not talking about an algorithm that makes a good poker bot, I am talking about playing a poker game in such a way that your EV is positive or zero no matter how your opponent(s) play.

Rudbaeck · #2 11-04-2004, 05:31 PM

It's clearly not possible when there is a rake. I think it's a pretty useless excercise anyway, as it probably won't be the most profitable play in any conceivable game.

jimdmcevoy · #3 11-04-2004, 06:09 PM

The only way it won't be the most profitable strategy is if you are playing against other players who play non-optimally. But the problem is you have to know how they play non-optimally. If you observe that in a certain situation the raised 19 out of 24 times and optimal strategy calls for raising 13.6 out of 24 times in that certain situation, you still can't be sure he is playing non-optimally, he may actually be thinking "I should raise 13.6 times out of 24" but the fact he has to do it randomly could cause him to do it 19 out of 24 times. So if you adapt to what you percieve is his non-optimal play, you are in danger of making negative EV moves. I think Mike Caro wrote about this once. Besides, I reckon since no one plays perfectly whatever edge a perfect strategy has over other players would beat the rake.

Seems like a usefull excercise to me.

Rudbaeck · #4 11-04-2004, 06:58 PM

Knowing the nature of your opponents non-perfect play is where virtually all the profit comes from. Even though it's a hard nut to crack you have to come to a working hypothesis quickly when you face all new and unknown faces at the felt.

jimdmcevoy · #5 11-04-2004, 08:29 PM

Yah okay in the real world a pro would probably make more money than some one playing perfect strategy, even thought I still reckon perfect strategy would beat the rake. But I still think it would be cool if some one created an unbeatable algorithm, by unbeatable I mean you have about the same chance as beating craps.

danderso8 · #6 11-04-2004, 08:57 PM

Look up Phzon's post on guts from a few months ago...If I remember correctly, he had pretty much solved it. the original poster was rocketmanjames.

--dan

M.B.E. · #7 11-05-2004, 01:49 AM

[ QUOTE ]
It's clearly not possible when there is a rake.

[/ QUOTE ]
What does a rake have to do with it?

There are various theorems in game theory that will say whether the game is solvable or not. My understanding is that any reasonable form of poker with three or more players is not solvable, but heads-up poker is.

The group at University of Alberta came up with a "pseudo" solution for heads-up limit holdem. It is not exact (since that would have taken too much computing time) but they used some clever shortcuts to come up with a solution that appeared to be very close to optimal.

For that work, they won the Distinguished Paper Award at the Eighteenth International Joint Conference on Artifical Intelligence:

http://www.cs.ualberta.ca/~darse/Papers/IJCAI03.html

Rudbaeck · #8 11-05-2004, 03:53 AM

[ QUOTE ]
[ QUOTE ]
It's clearly not possible when there is a rake.

[/ QUOTE ]
What does a rake have to do with it?

There are various theorems in game theory that will say whether the game is solvable or not. My understanding is that any reasonable form of poker with three or more players is not solvable, but heads-up poker is.

[/ QUOTE ]

I didn't know the later part.

With a rake, if everyone plays 'perfect' only the house wins. So even perfect play will in some situations make you a loser, which defied the OPs intention.

laja · #9 11-05-2004, 04:24 AM

-Finding the game theoretic solution would absolutely crush the other players, I don't believe you would lose enough money on not exploiting them to make it not viable, they are going to be playing definately far enough away from the perfect way.

-who thinks in reality poker is an N person game where there are actually coalitions, like maybe there are 2 moves you can do that would have relatively the same +ev , but you choose 1 that helps another player more than the other?

jimdmcevoy · #10 11-05-2004, 06:57 PM

Yeah I would be very interested if some one solved three person limit poker one day. I think it would be possible, it would just be harder due to the cooperation added to the mix. It would very interesting if there was a very smooth grey scale in cooperation, from none to full, which would create all sorts of problems with rules and ethics in poker.