|
Post by Dubber on Nov 12, 2007 17:26:20 GMT -5
Okay, folks... I'm poker (and as we all know, strategy and AI) illiterate, but this Boing Boing post suggests that Poker bots are improving enough to win significant percentages of online poker matches -- enough that the online poker companies are getting involved. What do the poker players on the boards think about these bots?
|
|
|
Post by ExDeath on Nov 12, 2007 19:29:57 GMT -5
Eh, I don't know much about the really really advanced AIs like the ones being designed by large companies, but I've played against the lesser ones and I can say they're still easily exploitable. I think there will never be a bot that can win more often than a good professional player, but it is certianly possible to design one that will win money for you while you're off sipping martinis or watching porn or whatever. The sites do a pretty good job of keeping these banned.
|
|
Derfel
Ronin Warlock
Did I Do That?
Troublemaker
Posts: 283
|
Post by Derfel on Nov 12, 2007 20:02:55 GMT -5
There were rumours of some Russian folks who had designed a "poker bot" that was winning 65-70% of hands games on online poker sites. Not sure if that is true, but if so that's a pretty decent margin to be winning at, and could make a solid dollar for an individual or "gang"...
|
|
|
Post by ExDeath on Nov 12, 2007 20:16:42 GMT -5
Um, winning 65% of all hands dealt? I don't see how that's possible, even heads up (1 on 1). You shouldn't even be playing 20% of all hands in a full ring game, let alone winning 65%. It's not as if a bot gets better cards than a human or something. And usually winning money correlates to winning fewer hands, not more. I really hate when people who don't know how poker works present silly facts like this.
|
|
|
Post by Rycchus on Nov 12, 2007 20:30:41 GMT -5
Perhaps he means winning 65% of hands played? Would that make more sense? Well obviously it makes -more- sense. The question is: does it make enough sense to be plausible? I wouldn't have a clue
|
|
|
Post by ExDeath on Nov 13, 2007 0:05:53 GMT -5
The way to make money in poker is by getting more money in while ahead and less money in while behind, it has nothing to do with your actual win %. Everybody has the same win % over the long run. So the goal in creating successful AI is for it to judge where it's at in a hand at any given time.
|
|
|
Post by BioLogIn on Nov 13, 2007 0:49:44 GMT -5
I remember reading a news article regarding that Polaris a while ago. A quick summary follows:
--
In the end of this Jule in Vancouver a skilled human poker player Phil Laak and his less know teammate Ali Eslami were playing 2v2 mirrored poker match vs new Polaris poker bot, developed in Canadian Alberta university. During the first day two session were played, the first concluded into a draw, and the second was won by Polaris. However, after additional analysis of Polaris strategies humans came back in the second day winning both sets.
Phil and Ali commented that this match was one of the hardest in their entire life. Summing up all 4 games their advantage was minimal, so Polaris developers (including Jonathan Schaeffer, who created AI that mastered checkers and beat human checkers champions) are looking towards rematch. However, the poker version version played in this show match had maximum bet limitation, which makes life much easier for a bot.
Simplified version of Polaris can be found here: poker-academy.com/man-machine/play.php But Mr. Shaffer assured, that Polaris will not be used in online poker casino with real money bets.
--
|
|
Derfel
Ronin Warlock
Did I Do That?
Troublemaker
Posts: 283
|
Post by Derfel on Mar 16, 2008 13:12:24 GMT -5
I believe it was 65% of tournament-style matches. IE, after many hands dealt and played, the bots had cleared the table of opponents in 65% of experiments.
|
|