How It Works
Built with Deep Counterfactual Regret Minimization (Deep CFR), a reinforcement learning technique that approximates Nash Equilibrium by iteratively playing against itself and training deep neural networks to minimize regret over past decisions.
Note: To keep the AI's processing manageable and strategy optimal, starting chips are reset to 100 every hand.
Nash Equilibrium
A strategy no opponent can exploit —
provably optimal in a game-theoretic
sense.
Self-Play
Zero human data. The AI discovers
strategy entirely by playing millions of
games against itself.
12 Neural Networks
8 advantage nets learn regret per
betting round. 4 strategy nets
accumulate the final equilibrium policy.
C++ + PyTorch
Parallel C++ game tree traversal runs
concurrently with PyTorch gradient
updates every iteration.
0
Your Wins
vs
0
AI Wins
P1
You
Waiting…
Stack 100
Bet 0
Preflop
POT 0
Community cards will appear here
AI
Deep CFR AI
Nash Equilibrium Strategy
Stack 100
Bet 0
Decision Breakdown
AI decision breakdown will appear here after the first move.
You'll see the exact probability the AI assigned to each action and the raw neural network outputs.
AI's Last Decision —
Nash Strategy Distribution Mixed equilibrium — probability
assigned to each action
Neural Network Output Brain Advantage per action in chips —
positive means "do this more"