University Thesis- Poker A.I.

[Flint{$N}]

Hey all,

Haven't made too many posts, but heres another.

I'm looking into creating Artificial Intellegance for limit hold em. I have just started to think about it as a thesis topic for school recently. I'm actually meeting with a professor this week to disscuss.

I'm pretty sure A.I. has been attempted before. If so, do you know who created it and if it was successful? I'm going to read up on the wilson software and see what sort of things they have done.

So a thesis on strucutred hold em AI, what do you think?

Flint

[Lurker4]

hey, I've always been interested in this kind of thing...this website might help you out: U of Alberta Computer Poker Research Group (makers of Poki AI for Poker Academy software)

[Flint{$N}]

Cool, thanks for the link.

Didn't know there was a dedicated research group in alberta.

Thanks

[EliteNinja]

If you do complete your thesis, do not publish it.
You should use the bot to your own advantage.
That would make you more money than your PhD ever will.

[youngin20]

Mmm...too hard. already tried it with machine learning, the state space is absurd. Try using naieve bayes maybe, and then training it on itself. Could look in to TD-Gammon, and stuff, but its really hard. Oh yeah, its already been done, Poki's Poker school or something, at U-Alberta has a program, its pretty processor intensive.

[DcifrThs]

[ QUOTE ]
Hey all,

Haven't made too many posts, but heres another.

I'm looking into creating Artificial Intellegance for limit hold em. I have just started to think about it as a thesis topic for school recently. I'm actually meeting with a professor this week to disscuss.

I'm pretty sure A.I. has been attempted before. If so, do you know who created it and if it was successful? I'm going to read up on the wilson software and see what sort of things they have done.

So a thesis on strucutred hold em AI, what do you think?

Flint

[/ QUOTE ]

Dr. Brian Alpsach, University of Alberta. gaming studies.

he has the best right now, called pokibot 2.something, or maybe its at 3.something now...

this bot's win rate/hr of playing vs. opponents increases linearly!! and it DOES NOT PLATEAU until a seriously large # of hours (hands.)

-Barron

[Flint{$N}]

When you say it increases linearly, is this becuase of historical data that characterizes the style of play by players?

[DcifrThs]

[ QUOTE ]
When you say it increases linearly, is this becuase of historical data that characterizes the style of play by players?

[/ QUOTE ]

yes. and it also applies that information (historical data) incredibly well.

not only style of play, but actual Pr(future street play | past street play, past hand tendencies, # players, position, etc..) ---> Dist. of hands & Pr(dist. of hands) <,> hero's current hand

-Barron