I’m not sure how poker bot AI is usually programmed. I assume they sacrifice some accuracy in order to obtain reasonable performance. What I’m after here is a total understanding of my odds in any situation, rather than merely a set of good heuristics.

I’m working on the program now, in C. I was going to use python, but I really need the extra performance. Right now I’m just simulating with a single opponent, and there’s about 2 billion permutations to go through for the pre-flop (50 choose 5 for the table, times 45 choose 2 for the opponent). I’ve only got it checking for a straight flush and then high card right now, and it takes 6 and a half minutes to finish, even with full compiler optimizations. I’m wondering if there’s some kind of time/memory trade-off that I could use to speed it up.

Let me know if you get into working on your poker thing, or if you want to see my code.

Long time no see, but I’m an avid reader of your blog(s).

I love playing Texas Hold’em, too, and thought about programming it all out as a web app in JS. I’d have a Cards class a Game class and a Player interface. It would be cool then, to put it up and have other people extend the player interface and we can have a bunch of AI players versing each other to see who can come up with the better AI.

Everything above would be pretty straight forward except for actually creating an “intelligent” player. Like you, the staggering number of permutations overwhelm me and hence I haven’t executed on that idea as of yet.

I think it’s a fun project to work on when I have a free chance, though, maybe you can help with those calculations. ;)