Poker Bot piece

[Pirc Defense]

[ QUOTE ]

That being said it doesn't and won't exist in our lifetimes.

[/ QUOTE ]

I can't help but think this statement is short-sighted. As noted, it flies in the face of nearly every (applicable) event in history. Someone, sooner than later, will figure out a way to program a bot to play winning poker in a ring game. This thread is going look awfully dumb in a few years, but not for the reason that Jimbo thinks.


[ QUOTE ]
Pirc keeps thinking poker is like chess, it is not and is not comparable.

[/ QUOTE ]

They are different games, to be sure, but whatever difference there is does not prove your contention that never will there exist, in your lifetime, a computer that can "win" in, I guess, a full ring game.

[ QUOTE ]
In fact an exploitable flaw in an otherwise poker bot would be quickly discovered by the majority of it's oponnents and exploited.

[/ QUOTE ]

Come on, Jimbo, this doesn't happen all that often against human players with exploitable flaws, so why should it happen with a computer?

[ QUOTE ]
This flaw may be so deeply ingrained that is is not only impossible to fix but to identify in the first place.

[/ QUOTE ]

If you were the programmer of the program, you think you wouldn't either notice the flaw, or at understand the flaw when pointed out to you (or after several head-scratching losing sessions)? What flaw-type are we talking about? Let's assume a flaw is that it undervalues mid-pair/decent kicker against two opponents. This would show up in hand/play reviews, no? This type of playing error wouldn't be too hard to figure out as well: find the bit of programming that assigns values to hand holdings and tweak.

Jimbo, I'm almost willing to put some money down on this. If we could agree on terms (which will be difficult), would you be willing to bet $100, in support of your belief? Not to be construed as an offer, but I'd say that within five years there will be a program that can beat a 10-person table in limit poker, comprised of players of a skill level that you would expect to find at a $20/$40 game at either Paradise or Party poker. By "beat" I mean earn at least one BB/hr over enough hands to be x% certain that the bb/hour metric is correct. Questions of how the bot would read it's card and such I'm not really addressing. Something like this sound reasonable?