search

EricSteinberger / PokerRL

Framework for Multi-Agent Deep Reinforcement Learning in Poker
MIT License
415 stars 88 forks source link

BR Texas Holdem #7

Closed SavvaI closed 5 years ago

SavvaI commented 5 years ago

Re-open from #6

Thank you again for response) I just proposed a modification of Deep CFR algorithm, which works really fine with small leduc poker game, but as you know it is extremely toy game, and to achieve any academical results I should measure exploitability in Texas game. Maybe, if you have access to proprietary tools for exploitability measurement, you could test my bot, and in the case of success and therefore publication, it could be a collaboration. Another question, you previously mentioned that local best response is supported for large games, but as I know it is modified version of best response, so it is quite strange that vanilla BR is supported only for toy games).

Originally posted by @SavvaI in https://github.com/TinkeringCode/PokerRL/issues/6#issuecomment-489189654

SavvaI commented 5 years ago

Sorry, I am not sure, if I have correctly reopened issue :)

EricSteinberger commented 5 years ago

Hi

Cool! I love that people are using PokerRL and the other repos for research. If you want to, we can discuss your approach via email - my e-mail is eric@steinberger-ai.com - don't worry, I won't steal the idea. I hope you haven't accidentally rediscovered Single Deep CFR https://arxiv.org/pdf/1901.07621.pdf though?

To answer your question: I don't have a tool for exploitability in NLH or similarly sized games as I did not need that until now. I think it is fine if you measure exploitability in a small game and use LBR in a big one. Exploitability is infeasible to measure on large game trees anyway.

LBR =/= BR. LBR is an approximation from samples that is totally useless for anything but evaluation. Code for BR could be reused to write bots for exact Turn and River solving in No Limit Hold'em games, which I don't want to be out there.

I will leave this open in case you have further questions. If this is fine, please close the issue. :)