Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Heads Up (two player) Limit (the raise size is fixed, raising is a binary decision) Hold 'Em is weakly solved. A mathematically determined strategy is arbitrarily close to optimal. There are likely other equally optimal strategies, but there can't be any which would actually win over time. http://poker.srv.ualberta.ca/about

Heads Up No limit is not solved, but machines are so good at it now that the world's top humans lost money to a machine when this was last attempted. https://en.wikipedia.org/wiki/Pluribus_(poker_bot)

In multi-seat games, like the ones Floptimal is designed for, an AI has the problem that rationally the humans should all work together to knock it out, simply because it's obviously stronger than they are. The humans needn't cheat, they just choose to take a different line against the bot than they'd take against other humans, so that the bot accumulates money much more slowly than the dominant human player.



Given the success of AlphaGo, and Pluribus’ ability to dominate humans 1-on-1, I’d be surprised if Pluribus couldn’t eventually be generalized to ring games. I’d also be surprised if today’s batch of online poker bots weren’t actively informed by ML or AI. Nor am I confident poker can be played in a fully online setting in a way that is impervious to cheating with bots.

Perhaps augmented reality experiences could help professionals play more hands of live poker, balancing volume with integrity of the games.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: