tictactoe AI

implemented with MCTS + lookup table of observed outcomes

This project is an effort towards creating a board game playing agent.

Store game histories and terminal states to disk
- exploring TFRecord and hdf5 for this. It would be nice to ensure that each board state is matched up to its result.
Load terminal states from disk at start
Replace evaluator with a function of board state, instead of random policy playout.

Name		Name	Last commit message	Last commit date
Latest commit History 21 Commits
.gitignore		.gitignore
README.md		README.md
board.py		board.py
player.py		player.py
pytac.py		pytac.py
referee.py		referee.py
tictac_mcts.py		tictac_mcts.py