Overview
Our bot is designed to play according to similar "mental math" principles used by human players. These strategies work by minimizing long-term losses and capitalizing on strong hands. It works primarily in two stages: pre-flop and post-flop.
Pre-Flop
Pre-flop, the bot judges potential hands that are worth playing. Examples of this are straight and flush draws, pairs, suited cards, or multiple high cards. If it doesn't encounter these, the bot folds to cut its losses. If it detects a particularly strong hand, it will raise to build the pot or bluff low to induce raises.
Post-Flop
Post-flop, our bot shifts to a different method of hand evaluation. The bot checks which cards it needs to complete one of the above hands and calculates the probability of receiving those cards. It then compares the potential strength of its hand to the amount required to stay in the game and plays accordingly.
Testing
We created a script to test different versions of our bot against each other. By creating multiple different bots with new strategies, we were able to eliminate weaknesses in our final bot. We ran each experiment multiple times with a varying number of rounds for more consistency.
Log in or sign up for Devpost to join the conversation.