My PyTorch implementation of Carsten Friedrich's "teaching-a-computer-to-play-tic-tac-toe"
TODO list:
- add Experience Replay to NN-Player
- add Duelling Q Network to NN-Player
| Player | P1 Win | P2 Win | Draw |
|---|---|---|---|
| Random - Random | 58.4% | 29.2% | 12.4% |
| MinMax - Random | 99.4% | 0.0% | 0.6% |
| Random - MinMax | 0.0% | 80.1% | 19.9% |
| Random - MinMax | 0.0% | 80.4% | 19.6% |
| MinMax - MinMax | 0.0% | 0.0% | 100.0% |
| TQ - Random | 95.1% | 1.5% | 3.4% |
| TQ - MinMax | 0.0% | 5.7% | 94.3% |