Crate ai_tictactoe 

ai-tictactoe

A tic-tac-toe rules and search library with perfect play.

Use it

Add the crate to your project:

[dependencies]
ai-tictactoe = "0.1"

Build the tablebase once, then use it for each position:

use ai_tictactoe::{Algorithm, Position, Tablebase, search};

let position = Position::default();
let tablebase = Tablebase::build();
let result = search(position, Algorithm::Tablebase, &tablebase);

assert!(result.best_move.is_some());

The crate also includes negamax, memoized search, symmetry reduction, and a small line-oriented engine.

cargo test
cargo run

Licensed under the MIT License.

Structs

Candidate

DecisionReport

Engine

One protocol session. This is shared by the native binary and WASM.

Move

A tic-tac-toe move is exactly one square.

Position

The whole game state: one nine-bit board per side and the side to move.

SearchReport

Complete exact-search result for a position.

SearchStats

SearchTree

A bounded tree annotated with exact tablebase values.

Square

One of the nine squares, numbered rank-first from a1 to c3.

Tablebase

An exact table for every position reachable from the empty board.

TreeEdge

Enums

Algorithm

Exact solving method used by search.

DecisionReason

GameResult

MoveError

Outcome

Exact game-theoretic result from the current side’s perspective.

PlayStrategy

A move-selection policy used for the playable opponent.

Side

Functions

build_tree

Builds an exact, tablebase-annotated tree up to depth plies.

choose_move

Chooses one move with strategy.

perft

Counts leaf positions reached at exactly depth plies or at game end.

search

Solves position exactly with the selected algorithm.