World Series of Poker Robots

[Arnfinn Madsen]

My theory after watching bots improving and then finally crushing humans in chess and backgammon I think the same will happen in poker, since bots' skill level increases more each year than humans.

BUT,
it is a very long way to go. Understanding human psychology is after all computer's main shortcoming. At some point in time a bot will manage to calculate bluff %-chance based on millions of hands versus thousands of players. However humans change gears, so the bot will also have to have enough experience and clever programmers to spot signs of changing gears.

My guesstimate is that it will crush .25/.50 in 2 years time, since players there do not wary enough (after all many players here can design a system to crush those limits). I expect, however up to 10 years before it is able to crush 2/4. In the end, it will crush 100/200 too.

Just stay ahead [img]/images/graemlins/cool.gif[/img].

[Gregg777]

[ QUOTE ]
Why would the blinds eat them away? the blinds are incremently as safe on the lower limits as the higher limits... its the same game just more money.

[/ QUOTE ]

There are plenty of reasons, but just look at Party 25NL vs 200NL. If you sit at 25NL for 5 rotations waiting for a decent hand in position, then raise, you almost always get callers. Do that on a 200NL table and they will simply fold. Not every time, but enough for the blinds to really come into play. Throw in the fact of fewer tables, more familiarity, opponents playing back at you more, making fewer mistakes, etc. And your bot simply would not survive.

[AaronBrown]

Maybe so, maybe not. I think the problems are harder than you suggest. Chess and backgammon are complete information games in which the main skill is memory (keeping a lot of potential future positions in short-term memory at once to evaluate patterns). Poker is mostly attention.

However, even if you are right about the higher limits, it will be much longer before we get a good no-limit computer poker player.