Poker-AI.org Poker AI and Botting Discussion Forum 2013-05-11T19:04:14+00:00 http://poker-ai.org/phpbb/feed.php?f=24&t=2485 2013-05-11T19:04:14+00:00 2013-05-11T19:04:14+00:00 http://poker-ai.org/phpbb/viewtopic.php?t=2485&p=4130#p4130 <![CDATA[Re: Hand Histories for hand-range selection in decision syst]]>
You should read these forums some more, the old archive and the poker papers sections are great resources too for beginners, which you seem to be.

If you have specific questions go ahead and ask but I'm not sure if people will (want to) explain the different methods when your question is so broad.

Statistics: Posted by Coffee4tw — Sat May 11, 2013 7:04 pm

]]> 2013-05-06T12:31:33+00:00 2013-05-06T12:31:33+00:00 http://poker-ai.org/phpbb/viewtopic.php?t=2485&p=4110#p4110 <![CDATA[Hand Histories for hand-range selection in decision systems]]>
I am new to the forums here and been developing a online poker bot for some months now.
Did do a Masters in AI and a fairly bad poker player. Got all my interface and scraping, deciding model etc build and already got 4 accounts banned on Pokerstars. However on 0.02/0.04 FL games I made quite a stable profit. Now re-arranged my screenscraping to start playing on another site.

Now I use not real AI techniques yet but a very simple method. I use simple monte-carlo sampling and the VPIP values from handhistories to adjust the random hand selection range of my opponents if they folded (tried on betting and calling etc but that resulted in wrong selections). Basicly with VPIP of 20 I select the top 20% of cards if they call or raise and the rest if they fold. I play like a million games (if everyone checked till showdown) and see how many times I win and against whom I would lose resulting in some percentage. And then the AI just does if > 85% then raise else call. if <50% then fold/check.

This very simple method could make huge differences in winning chances based on folds/calls. I am not saying this is the greatest method but it was a start and proved to be pretty useful.

Now this was a first try and I thought it worked remarkably well on those games but I am wondering how others incorporate handhistories in their AI's and if/how they use hand range selection on their opponents.

Hope to get some good discussion on this! Greets

Statistics: Posted by ovanwijk — Mon May 06, 2013 12:31 pm

]]>