Just when the poker craze was hitting its peak
in early of 2005, I decided to take it upon myself to develop an
AI capable of competing and winning at the most popular game of
poker at the time: Texas Hold’Em. A daunting task - though,
those are the ones which are most fun. :)
After doing a fair amount research, I stumbled across the University
of Alberta Gaming Group. This group had created two Hold’Em-playing
bots which they referred to as Poki & PsOpti. The development
of PsOpti really interested me, as it was based on the principles
of Game Theory, whereas Poki was based on what seemed like intelligent
I have since created an advanced game-theoretic algorithm capable
of beating PsOpti in 2-Player Heads-Up Hold’Em. The general
approach used to create my player can be described simply as a combination
of appropriate abstraction techniques and player training using
the principles of Fictitious Play.
My work for this project has been accepted to the 2006 Conference
on Artificial Intelligence where I will present the paper which
is linked to at the bottom of this page.
If you have any questions, please feel free to send
me an email.
Project started in January
2005. Current version
is 4.0 written in VB.NET & C++.
Published in the proceedings
of the 2006 Conference on Artificial Intelligence:
Download the paper in Adobe PDF format:
Using Fictitious Play to Find
Pseudo-Optimal Solutions for Full-Scale Poker