Struct Mcts 

pub struct Mcts { /* private fields */ }

UCT Monte Carlo tree search for sequential games, with chance nodes.

Assumes exactly one active player per decision node (like Minimax) and two-player zero-sum outcomes. All node values are kept from a fixed root player’s perspective: the root maximizes, the opponent minimizes (a sign flip in selection), and chance nodes average their sampled children, which is exactly the expectiminimax behavior. Chance nodes are descended by sampling Game::chance_outcomes, never by UCT.

Rollouts are uniform-random. Randomness (rollouts and chance sampling) comes from an internal seeded generator, so a run is reproducible.

Implementations

impl Mcts

pub const fn new(iterations: u32, seed: u64) -> Self

Creates a search running iterations simulations per move, seeded from seed. Uses the standard UCT exploration constant (sqrt 2).

pub const fn with_exploration(self, exploration: f64) -> Self

Sets the UCT exploration constant (higher explores more).

pub fn evaluate<G>( &mut self, game: &G, state: &G::State, player: PlayerId, ) -> f64

where G: Game, G::State: Clone, G::Action: Clone,

Estimates the value of state for player as the mean rollout reward, searching from player’s perspective. Works even when the root is a chance node (no decision to make), unlike Bot::choose.

Trait Implementations

impl<G> Bot<G> for Mcts

where G: Game, G::State: Clone, G::Action: Clone,

fn choose( &mut self, game: &G, state: &G::State, player: PlayerId, ) -> Option<G::Action>

Returns the action to play for player in state, or None if there is nothing to do (no legal actions, e.g. a terminal state).

impl<G> RankedBot<G> for Mcts

where G: Game, G::State: Clone, G::Action: Clone,

fn rank( &mut self, game: &G, state: &G::State, player: PlayerId, ) -> Vec<(G::Action, f64)>

Ranks root actions by visit share (the MCTS-recommended policy). Scores are the fraction of simulations spent on each move and sum to 1.