Poker AI

[PairTheBoard]

Great list of links M.B.E. Thanks. It looks like they're working on it but have quite a ways to go yet. The Acespade software looks interesting. I wonder how it would do at a Party Poker 2-4 game.

[chesspain]

[ QUOTE ]
Think of it like chess, most chess players even ones who are pretty good probably can't even beat a storebought chess program set to max difficulty. If the poker equivalent of Deep Blue (or whatever the top chess program) were to be developed, I would be quite worried.

[/ QUOTE ]

But it is really quite different from chess, since most of the advantage that computers derives in chess comes from:

1) Their ability to store in memory tens of thousands of previously played or analyzed opening moves.

2) Their awesome processing power, which allows them to analyze a tremendous number of ply to the nth degree, which allows them to see with perfect accuracy potential positions which might occur.

The above two advantages are even more pronounced with programs that utilize a "brute force" over a selective search paradigm, with the latter being more analogous to human thought processing. The one advantage good human players in poker would have against a computer is that there is not an insurmountable degree of memory or computation required. Whereas a computer could probably make a fair amount of money playing limit poker against loose computation by just by maximizing value betting, I don't see how a computer would necessarily beat the Monday night 2+2 table.

[brianmarc]

OH: Thanks for your comments. You are probably correct with most of your comments.

[brianmarc]

Thanks. I was familiar with most, but there were some interesting new ones.

[M.B.E.]

Brianmarc -- put me in the camp of believing that it's possible (given the present state of computer science) to design a bot that would beat most ring games, even at high limits. In other words I think a well-programmed computer would play poker as well as the best ring-game players in the world (although it would play differently). The only way to beat it would be through cheating (i.e. collusion).

You just have to consider the PsOpti program designed by Darse Billings et al., which beat The Count at headsup limit holdem. The Count is the winningest shorthanded limit holdem player online. And PsOpti did not use opponent modeling. Think of how good a program would be if it did.

As for no-limit and pot-limit, I tend to think that a computer could play these well too, although it might require more power.

The one part of your post I didn't understand was this:

[ QUOTE ]
Three: Since your win rate requirement can be much lower than when you are playing in person, the bot can play much more patiently than any human player ever could. Since all current poker theory is based on squeezing the maximum out of every situation (i.e., to be as efficient as possible), an approach not driven by this consideration, in effect, allows for the creation of a hitherto unknown style of expert player.

[/ QUOTE ]

Why would you not want to have the highest possible win rate? I can think of a few reasons, e.g. create deception, reduce variance, not-get-caught, but in general it seems to me they wouldn't be worth it.

[clovenhoof]

That's funny, I was thinking exactly the opposite. Where the topic of discussion is "where are we going with this in the future", I don't see that a reply of "we aren't there now and other people will be somewhere else" adds anything of value to the conversation.

The statement that those who could make a winning bot could make more money designing other kinds of program (which is how I interpret the spinning straw into gold comments) in my view overlooks the essence of the spirit of invention. Society has advanced because of people responding to their own muse, not that of other people or what other people think would be more profitable. Few Nobel laureates started down their path to the prize by saying, "Yeah, this is it. This is what's going to make me rich." Broadway is full of some of the greatest actors in the world, who could have made more money in film. Medical progress has been the result of scientists who chose not to make more money by developing a better diet pill, or a pill that helps people quit smoking.

While a poker bot certainly isn't something that will make the world a better place, it is a fascinating problem of the highest order. If we're focusing on a bot that uses artificial intelligence (as opposed to just a really well written program) then in my view poker is the perfect platform -- it strikes a very nice balance between game theory and the inherent flaws in human behaviour.

My view is that given the problem posed -- a bot that can beat the crap out of the online games that currently exist -- the current processor speeds, storage capacities, and memory size are more than sufficient that such a thing is possible, particularly given how much time the sites give each player to make a decision before timing him out.

'hoof

[brianmarc]

Translated in into strategy, the bot could afford to pass up opportunities that the regular player could not. one simple example: require a premium on pot odds to go for a draw.

[Wake up CALL]

Brian,

Here is a Yahoo discussion group you might like to join which will be better able to answer your questions.

PokiPoker Group

[M.B.E.]

[ QUOTE ]
Translated in into strategy, the bot could afford to pass up opportunities that the regular player could not. one simple example: require a premium on pot odds to go for a draw.

[/ QUOTE ]
But why? Just to reduce variance?

[Bozeman]

to simplify programming or avoid possible counterstrategies (assuming not headsup)