Rust implementation of Monte Carlo Tree Search for board game AIs.
Based on the existing implementation from jbradberry/mcts.
Pick the game to test by choosing which state to initialize:
// let mut state = NimState::new(10);
let mut state = AgricolaState::new(2);
Execute the game:
cd example-games/play-game
cargo run --release
In example-games/play-game/src/main.rs
, adjusting the iterations
number will increase the number of games played by the AI before making a decision.
iterations = 10001;
best_action = UCT(arena, state.clone(), iterations);
The following is a table of iterations to time per selection:
1000 - 1 second
10000 - 20 seconds
100000 - 260 seconds