An interactive Connect Four command-line game, where two players attempt to be the first to connect four markings of their color.
This implementation provides an Ai opponent, which utilizes Monte Carlo tree search to search and learn optimal moves in self-play at runtime.
Monte Carlo search trees provide the foundation for more sophisticated reinforcement learning algorithms, such as AlphaZero, where node quality is evaluated by a neural network.
Rollout is currently performed by random simulation. For a simple game such as connect four, this already yields an agent of significant strength.
cargo run --release