News for hackers. Stuff that matters.
Re:A $10k poker win changed how I think by RivieraKid on Monday September 06, 2021 @04:58PM (#28438038)
Is anyone here familiar with current state of poker AIs / bots? For a long time, I've been thinking about making a multiplayer bot that would be able to play at a somewhat decent level but it seems like an extremely hard problem.

I've read the Pluribus paper but it assumes fixed stack sizes and more importantly, would be very hard to implement. There has to be a simpler approach.

Re:A $10k poker win changed how I think by onekorg on Tuesday September 07, 2021 @04:26AM (#28442372)
I read the 2019 paper[0] where they beat 5 pros in a 6-max table. It did not assume fixed stack sizes, it uses a technique they call "action abstraction" where they train on some number of stack sizes. During a game they use these pre-computed values as starting points for real time searching of the game tree.

Their papers are very well written. With enough determination someone could build a multi player bot.

I was also thinking about trying to build something simpler that works decently well. They published a paper a while back about a bot that only did real time search with no training. It performed better than older poker bots that used to require offline computation. I don't think you'd be able to beat pros with it, but it should do well vs amateurs.

[0] https://www.science.org/doi/10.1126/science.aay2400

Re:A $10k poker win changed how I think by RivieraKid on Tuesday September 07, 2021 @07:37AM (#28443464)
Are you really sure the starting stack sizes are variable? From what I remember, they aren't.
Re:A $10k poker win changed how I think by onekorg on Tuesday September 07, 2021 @08:15PM (#28451403)
My bad, for some reason assumed you were talking about bet sizes.

Starting stack sizes were static, but I don't think it's something important. They varied during the course of play. If you're playing cash instead of tournament you'd just need to set each player's stack sizes when they join the table.

Re:A $10k poker win changed how I think by btilly on Monday September 06, 2021 @05:19PM (#28438218)
According to poker experts that I know, bots have long beat humans in multiplayer because they are better able to track what is happening with all players at the table.

After https://www.deepstack.ai/ beat no-limit heads up, poker bots simply became better than humans.

Re:A $10k poker win changed how I think by spurgu on Tuesday September 07, 2021 @07:20AM (#28443379)
Humans have been using tracking software (that collects data from hand histories and overlays stats on top of the tables) since around 2005 at least...
Re:A $10k poker win changed how I think by awb on Monday September 06, 2021 @07:25PM (#28439224)
Better than humans at 1v1 poker. But to my knowledge no bots are beating a traditional full table of poker pros.
Re:A $10k poker win changed how I think by Forlien on Monday September 06, 2021 @08:01PM (#28439422)
The best I can find is https://en.wikipedia.org/wiki/Pluribus_(poker_bot), but it sounds like it was limited to a few demonstrations.
Re:A $10k poker win changed how I think by jeffreyrogers on Tuesday September 07, 2021 @12:50PM (#28446970)
You are correct that 1v1 is easier than multi-player.
Re:A $10k poker win changed how I think by jeffreyrogers on Monday September 06, 2021 @06:39PM (#28438908)
I have an acquaintance who wrote one about a decade ago. He said he was profitable in small stakes play, but the average player was pretty bad then. I've played recently and the average player seems quite a bit better now.

I would think that a pretty naive bot would do fine at lower stakes since most people there are not that disciplined, so just sticking to a basic strategy and not making obvious mistakes would probably have positive EV.

Seems unlikely that anyone is going to publish a system that currently works though.

Re:A $10k poker win changed how I think by awb on Monday September 06, 2021 @07:27PM (#28439233)
Plus it’s illegal on every major online poker site to use a bot.
Re:A $10k poker win changed how I think by hsbauauvhabzb on Monday September 06, 2021 @08:01PM (#28439424)
No, it’s against EULA, there’s a very distinct difference.
Re:A $10k poker win changed how I think by TameAntelope on Tuesday September 07, 2021 @12:58AM (#28441083)
I would not be surprised if there were federal laws against the use of Poker bots online.
Re:A $10k poker win changed how I think by hsbauauvhabzb on Tuesday September 07, 2021 @02:04AM (#28441439)
Good luck with that. I’m sure Eastern Europeans really care about your federal laws.
Re:A $10k poker win changed how I think by TameAntelope on Tuesday September 07, 2021 @02:07AM (#28441459)
Good luck with what?
Re:A $10k poker win changed how I think by hsbauauvhabzb on Tuesday September 07, 2021 @02:49AM (#28441731)
Any hope of jurisdiction bound federal law preventing something on the internet which is not politically or socially charged.

Hell, there’s a widespread issue of ransomware currently - certain governments overlook cyber gangs in all but the most (financially) extreme circumstances. There’s not a chance any political force would bother policing internet poker.

Re:A $10k poker win changed how I think by TameAntelope on Tuesday September 07, 2021 @09:20AM (#28444454)
Internet poker is heavily regulated in the US.
Re:A $10k poker win changed how I think by hsbauauvhabzb on Tuesday September 07, 2021 @11:52PM (#28452823)
Sure, but that’s the government controlling a handful of well established entities, not a bunch people running bots which are basically invisible to the poker sites, even if the government cared enough to outlaw bots they couldn’t
Re:A $10k poker win changed how I think by doix on Tuesday September 07, 2021 @04:59AM (#28442551)
I once met someone that (allegedly?) ran a poker bot ring in the mid-late 2000's. The basic idea was that you get as many bots per table as you can and share information between them. Suddenly you have much more information about the cards in play and it significantly increases your chances of winning. No machine learning back then, just pure stats and a set of rules to follow on how to play vs chance of winning. You didn't care which bot won since you could equalize stack sizes as soon as the human player folded.

I think back in those days, the poker companies didn't care too much since they were happy collecting fees from bots passing chips back and forth. I'm assuming now days it's different. I also don't know how true it was, since it was a random drunk guy in a bar, but it passed my sniff-test and he was buying beers so I didn't care too much.

Re:A $10k poker win changed how I think by spurgu on Tuesday September 07, 2021 @07:19AM (#28443370)
I mean I played everything from 4 to 12 tables back in the day and you of course do collect more information on other players like that whether you're a human or a bot.

Oftentimes, especially when down to fewer tables, I would have a couple of other tables open where I was on the waiting list and thus I gathered some information from hands that went to showdown there as well.

Re:A $10k poker win changed how I think by currymj on Tuesday September 07, 2021 @08:43AM (#28444012)
DeepMind’s OpenSpiel repository on Github might be a good place to start, they provide infrastructure for a lot of games including poker, and example implementations of algorithms like CFR.