PokerBroker

What it Does

PokerBroker is our poker bot that utilizes Monte Carlo Search Tree and heuristics to determine an action against another bot. The result is a competitive bot that utilizes its hand and analyzes the opponent’s tendencies to win in a modified version of Poker.

How We Built It

Python: Used vanilla python along with the given boilerplate framework to build our bot. We utilized Monte Carlo Tree Search by building a custom tree and handling updates of action history.

Challenges We Faced

Bugs with Bot Connection: We had issues with robot connections that involved properly calling the correct classes (such as RaiseAction()) and we were able to debug that and fix the issue

Accomplishments We’re Proud Of

Finally being able to win some games against the all win bot. Successfully implemented Monte Carlo Search Tree and also utilized our knowledge of Poker to build heuristics.

What We Learned

Learned how to utilize and manage Monte Carlo Search trees for poker gameplay. Learned how to implement probabilistic methods for critical decision-making.

What’s Next

In the future, we hope to make improvements to the way in which the bot analyzes its opponents and sets its risk strategy based on that information. For example, an LSTM could be used due to its versatility with less or no training.