Struct Minimax 

pub struct Minimax { /* private fields */ }

Depth-limited alpha-beta minimax.

Assumes exactly one active player per node (sequential play) and two-player zero-sum outcomes. Leaves are evaluated with Game::reward, so give the search enough depth to reach terminals (or a game whose reward doubles as a heuristic at non-terminal nodes).

Two search paths are offered and always agree: Minimax::best_action clones state per node (needs State: Clone), and Minimax::best_action_unmake uses a game’s Reversible make/unmake.

Implementations

impl Minimax

pub const fn new(max_depth: u32) -> Self

Creates a search that looks up to max_depth plies ahead.

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

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

Returns the value-maximizing action for player using cloned search.

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

where G: Reversible, G::Action: Clone,

Returns the value-maximizing action for player using make/unmake.

Mutates state during the search but restores it exactly, so state is unchanged on return.

Trait Implementations

impl<G> Bot<G> for Minimax

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 Minimax

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

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

Scores each root action with a full alpha-beta window (no cross-sibling pruning), so every score is the move’s exact minimax value rather than a bound. Costlier than Minimax::best_action; use it for analysis, not hot self-play.