TicTacToeGo_Zero

TicTacToe Monte-Carlo-Tree-Search and Reinforcement Learning

The project contains two seperate AIs for playing TicTacToe:

1. Monte Carlo tree search
2. Neural Network trained by reinforcement learning (Not working yet!)

To launch the game, simply start the play.py script with the following parameters:

1. Size of game (e.g. 3 for standard TicTacToe)
2. Winning condition (number of consecutive pieces)
3. Start player (0 for Human, 1 for computer)
4. AI mode (tree, network(only for 3x3 game) or random)(optional, default=network)
5. Number of Monte Carlo simulations (optional, default=1000)

Examples:

./play.py 3 3 0
./play.py 5 4 1 tree 3000
./play.py 6 4 0 random

Dependencies:

- numpy
- pytorch (for network)