Inspiration
Our inspiration was our own enjoyment of playing poker. We wanted a bot that reads its opponent the way a good player does, learns that opponent's habits during the match, and picks the move worth the most chips against that particular player.
What it does
It reads the opponents hand, learns the opponents habits, including the likeliness of the opponent to fold, it manages it's own speed, and exploits bots that crash.
How we built it
We build an evaluator, particle filter, an expected value decision layer, and robustness checks.
Challenges we ran into
We had a problem with the time constraints and with our model being too cautious.
Accomplishments that we're proud of
It beats all our previous versions and all the baselines.
What we learned
A bot that adapts beats a fixed rule set.
What's next for Heads Up Bot
Making our own play more balanced and harder to read.
