Poker-AI.org

Right, so i decided to do some project at college which is to make a poker AI nlhu bot. It probably doesn't need to be very good at all. But i only have 2 months. What is the best methods to use to create one, and what is the best papers on them

Thankyou.

im currently reading:

http://mlanctot.info/files/papers/PhD_Thesis_MarcLanctot.pdf

What kind of experience do you have in coding, or game theory? I have a feeling that trying to implement a system similar to that in the paper will be extremely difficult in just 2 months. However, if you know you're stuff it's slightly more doable.

Are there any specifications for the project?

If it doesn't have to be great, you may find working on a simulation/MCTS paper easier. In my opinion they tend to be easier to understand and have more basic concepts. Although it's not going to produce a bot that's top notch, it will at least give you results.

a pokerai from zero in two months is impossible in my opinion. even a losing one.
if you have a buddy and working 24/7 than it could work, but its still a losing ai

What constraints are you working under?
- full / part time
- programming experience
- poker experience
- language experience
- must have academic content, eg must you use an AI approach

Writing a bot for poker-academy would cut out a lot of work. It wouldn't be too hard to write a little expert system that uses the following variables (and maybe a few others)
- strength and potential of hand (use a prebuilt evaluator for this)
- position
- pot
- amount to call
- round

The very earliest Alberta AIs had a lot of expert content in them

Your best bet might be using the CFRM code amax posted in the old forum and spend two months working on abstractions and LUTs.

Just do a basic AI will rules raising say 80% of hands on the button, OOP have it call 37% of hands or so, Post flop just play top pair plus, open draws and forget all bluffing. Use Timmys evaluator for hand strength, hand info which is simple although you will have to figure out what numbers mean what. For Allins or large bets just stack off with 10% of hands or something like that pre and anything >= 90% hand strength post flop.

That AI would be too tight to win but its easy and would work. This would only take a day or two to code for HU for a novice programmer imo.

thanks for replies

urm previous experience is that ive done 3 years of programming (never in AI techniques though). Im a decent player and a winner at 400nl on stars. I could create a rule based bot but i did want to learn something (like some technique) and somehow "contribute to the subject". Not sure though

i have a lot of time on my hands i guess, will obv have other module work to do but most can be spent on this. Ive also programmed the whole system GUI, rules, hand evaluator, its just the AI i needed to do