Inspiration
I'm inspired by gambling
What it does
It helps me gamble in texas hold'em style, and I want to keep gambling
How We Built It
Our B4G Hold'em poker bot employs a hybrid approach combining reinforcement learning with poker-specific domain knowledge:
Q-Learning Core
- Uses feature-based state representation (hand strength, pot odds, position)
- Linear function approximation with tunable weights
- Learns optimal policies through experience and reward signals
Hand Evaluation System
- Monte Carlo simulation for accurate hand strength estimation
- Specialized for B4G format (3 hole cards, 4 community cards)
- Detects premium hands (royal flush, quads, etc.) for special handling
Strategic Framework
- Expert system for premium hands guarantees optimal play
- Q-learning for everyday situations ensures adaptability
- Position-aware strategy adjusts for acting first or last
- Street-specific logic for preflop vs postflop decisions
Optimizations
- Hand strength caching saves computational resources
- Time management prevents tournament timeouts
- Dynamic simulation depth based on decision importance
Challenges we ran into
Getting significant wins between iterations and against baseline bots
Accomplishments that we're proud of
Gambling all my money away
What we learned
How to gamble as a computer scientist
What's next for Bobfartsnow - PokerBot
Go to Vegas
Log in or sign up for Devpost to join the conversation.