game_solver/lib.rs
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373
//! `game_solver` is a library for solving games.
//!
//! If you want to read how to properly use this library,
//! [the book](https://leodog896.github.io/game-solver/book) is
//! a great place to start.
pub mod disjoint_game;
pub mod game;
pub mod player;
pub mod stats;
pub mod loopy;
// TODO: reinforcement
// #[cfg(feature = "reinforcement")]
// pub mod reinforcement;
pub mod transposition;
use core::panic;
#[cfg(feature = "rayon")]
use std::hash::BuildHasher;
use std::sync::atomic::Ordering;
use game::{upper_bound, GameState};
use player::{ImpartialPlayer, TwoPlayer};
use stats::Stats;
use crate::game::Game;
use crate::transposition::{Score, TranspositionTable};
use std::hash::Hash;
use thiserror::Error;
#[derive(Error, Debug)]
pub enum GameSolveError<T: Game> {
#[error("could not make a move")]
MoveError(T::MoveError),
}
/// Runs the two-player minimax variant on a zero-sum game.
/// Since it uses alpha-beta pruning, you can specify an alpha beta window.
fn negamax<T: Game<Player = impl TwoPlayer + 'static> + Eq + Hash>(
game: &T,
transposition_table: &mut dyn TranspositionTable<T>,
mut alpha: isize,
mut beta: isize,
stats: Option<&Stats<T::Player>>,
) -> Result<isize, GameSolveError<T>> {
if let Some(stats) = stats {
stats.states_explored.fetch_add(1, Ordering::Relaxed);
}
// TODO: debug-based depth counting
// if let Some(stats) = stats {
// stats.max_depth.fetch_max(depth, Ordering::Relaxed);
// }
// TODO(perf): if find_immediately_resolvable_game satisfies its contract,
// we can ignore this at larger depths.
match game.state() {
GameState::Playable => (),
GameState::Tie => {
if let Some(stats) = stats {
stats.terminal_ends.tie.fetch_add(1, Ordering::Relaxed);
}
return Ok(0);
}
GameState::Win(winning_player) => {
// TODO: can we not duplicate this
if let Some(stats) = stats {
if let Ok(player) = castaway::cast!(winning_player, ImpartialPlayer) {
if ImpartialPlayer::from_move_count(
stats.original_move_count,
game.move_count(),
) == player
{
stats.terminal_ends.winning.fetch_add(1, Ordering::Relaxed);
} else {
stats.terminal_ends.losing.fetch_add(1, Ordering::Relaxed);
}
} else if stats.original_player == winning_player {
stats.terminal_ends.winning.fetch_add(1, Ordering::Relaxed);
} else {
stats.terminal_ends.losing.fetch_add(1, Ordering::Relaxed);
}
}
// if the next player is the winning player,
// the score should be positive.
if game.player() == winning_player {
// we add one to make sure games that use up every move
// aren't represented by ties.
//
// take the 2 heap game where each heap has one object in Nim, for example
// player 2 will always win since 2 moves will always be used,
// but since the upper bound is 2, 2 - 2 = 0,
// but we reserve 0 for ties.
return Ok(upper_bound(game) - game.move_count() as isize + 1);
} else {
return Ok(-(upper_bound(game) - game.move_count() as isize + 1));
}
}
};
// check if this is a winning configuration
if let Ok(Some(board)) = game.find_immediately_resolvable_game() {
match board.state() {
GameState::Playable => panic!("A resolvable game should not be playable."),
GameState::Tie => {
if let Some(stats) = stats {
stats.terminal_ends.tie.fetch_add(1, Ordering::Relaxed);
}
return Ok(0);
}
GameState::Win(winning_player) => {
if let Some(stats) = stats {
if let Ok(player) = castaway::cast!(winning_player, ImpartialPlayer) {
if ImpartialPlayer::from_move_count(
stats.original_move_count,
game.move_count(),
) == player
{
stats.terminal_ends.winning.fetch_add(1, Ordering::Relaxed);
} else {
stats.terminal_ends.losing.fetch_add(1, Ordering::Relaxed);
}
} else if stats.original_player == winning_player {
stats.terminal_ends.winning.fetch_add(1, Ordering::Relaxed);
} else {
stats.terminal_ends.losing.fetch_add(1, Ordering::Relaxed);
}
}
if game.player().turn() == winning_player {
return Ok(upper_bound(&board) - board.move_count() as isize + 1);
} else {
return Ok(-(upper_bound(&board) - board.move_count() as isize + 1));
}
}
}
}
// fetch values from the transposition table
{
let score = transposition_table
.get(game)
.unwrap_or_else(|| Score::UpperBound(upper_bound(game)));
match score {
Score::UpperBound(max) => {
if beta > max {
beta = max;
if alpha >= beta {
if let Some(stats) = stats {
stats.cache_hits.fetch_add(1, Ordering::Relaxed);
}
return Ok(beta);
}
}
}
Score::LowerBound(min) => {
if alpha < min {
alpha = min;
if alpha >= beta {
if let Some(stats) = stats {
stats.cache_hits.fetch_add(1, Ordering::Relaxed);
}
return Ok(alpha);
}
}
}
};
}
// for [principal variation search](https://www.chessprogramming.org/Principal_Variation_Search)
let mut first_child = true;
for m in &mut game.possible_moves() {
let mut board = game.clone();
board
.make_move(&m)
.map_err(|err| GameSolveError::MoveError::<T>(err))?;
let score = if first_child {
-negamax(
&board,
transposition_table,
-beta,
-alpha,
stats
)?
} else {
let score = -negamax(
&board,
transposition_table,
-alpha - 1,
-alpha,
stats
)?;
if score > alpha {
-negamax(
&board,
transposition_table,
-beta,
-alpha,
stats
)?
} else {
score
}
};
// alpha-beta pruning - we can return early
if score >= beta {
if let Some(stats) = stats {
stats.pruning_cutoffs.fetch_add(1, Ordering::Relaxed);
}
transposition_table.insert(game.clone(), Score::LowerBound(score));
return Ok(beta);
}
if score > alpha {
alpha = score;
}
first_child = false;
}
transposition_table.insert(game.clone(), Score::UpperBound(alpha));
Ok(alpha)
}
/// Solves a game, returning the evaluated score.
///
/// The score of a position is defined by the best possible end result for the player whose turn it is.
/// In 2 player games, if a score > 0, then the player whose turn it is has a winning strategy.
/// If a score < 0, then the player whose turn it is has a losing strategy.
/// Else, the game is a draw (score = 0).
pub fn solve<T: Game<Player = impl TwoPlayer + 'static> + Eq + Hash>(
game: &T,
transposition_table: &mut dyn TranspositionTable<T>,
stats: Option<&Stats<T::Player>>
) -> Result<isize, GameSolveError<T>> {
let mut alpha = -upper_bound(game);
let mut beta = upper_bound(game) + 1;
// we're trying to guess the score of the board via null windows
while alpha < beta {
let med = alpha + (beta - alpha) / 2;
// do a [null window search](https://www.chessprogramming.org/Null_Window)
let evaluation = negamax(
game,
transposition_table,
med,
med + 1,
stats
)?;
if evaluation <= med {
beta = evaluation;
} else {
alpha = evaluation;
}
}
Ok(alpha)
}
/// Utility function to get a list of the move scores of a certain game.
/// Since its evaluating the same game, you can use the same transposition table.
///
/// If you want to evaluate the score of a board as a whole, use the `solve` function.
///
/// # Returns
///
/// An iterator of tuples of the form `(move, score)`.
pub fn move_scores<'a, T: Game<Player = impl TwoPlayer + 'static> + Eq + Hash>(
game: &'a T,
transposition_table: &'a mut dyn TranspositionTable<T>,
stats: Option<&'a Stats<T::Player>>
) -> impl Iterator<Item = Result<(T::Move, isize), GameSolveError<T>>> + 'a {
game.possible_moves().map(move |m| {
let mut board = game.clone();
board
.make_move(&m)
.map_err(|err| GameSolveError::MoveError(err))?;
// We flip the sign of the score because we want the score from the
// perspective of the player playing the move, not the player whose turn it is.
Ok((
m,
-solve(&board, transposition_table, stats)?,
))
})
}
pub type CollectedMoves<T> = Vec<Result<(<T as Game>::Move, isize), GameSolveError<T>>>;
/// Parallelized version of `move_scores`. (faster by a large margin)
/// This requires the `rayon` feature to be enabled.
/// It uses rayon's parallel iterators to evaluate the scores of each move in parallel.
///
/// This also allows you to pass in your own hasher, for transposition table optimization.
///
/// # Returns
///
/// A vector of tuples of the form `(move, score)`.
#[cfg(feature = "rayon")]
pub fn par_move_scores_with_hasher<
T: Game<Player = impl TwoPlayer + Sync + 'static> + Eq + Hash + Sync + Send + 'static,
S,
>(
game: &T,
stats: Option<&Stats<T::Player>>
) -> CollectedMoves<T>
where
T::Move: Sync + Send,
T::MoveError: Sync + Send,
S: BuildHasher + Default + Sync + Send + Clone + 'static,
{
use crate::transposition::TranspositionCache;
use rayon::prelude::*;
use std::sync::Arc;
// we need to collect it first as we cant parallelize an already non-parallel iterator
let all_moves = game.possible_moves().collect::<Vec<_>>();
let hashmap = Arc::new(TranspositionCache::<T, S>::new());
all_moves
.par_iter()
.map(move |m| {
let mut board = game.clone();
board
.make_move(m)
.map_err(|err| GameSolveError::MoveError::<T>(err))?;
// We flip the sign of the score because we want the score from the
// perspective of the player pla`ying the move, not the player whose turn it is.
let mut map = Arc::clone(&hashmap);
Ok((
(*m).clone(),
-solve(&board, &mut map, stats)?,
))
})
.collect::<Vec<_>>()
}
/// Parallelized version of `move_scores`. (faster by a large margin)
/// This requires the `rayon` feature to be enabled.
/// It uses rayon's parallel iterators to evaluate the scores of each move in parallel.
///
/// By default, this uses the cryptograpphically unsecure `XxHash64` hasher.
/// If you want to use your own hasher, use [`par_move_scores_with_hasher`].
///
/// # Returns
///
/// A vector of tuples of the form `(move, score)`.
#[cfg(feature = "rayon")]
pub fn par_move_scores<
T: Game<Player = impl TwoPlayer + Sync + 'static> + Eq + Hash + Sync + Send + 'static,
>(
game: &T,
stats: Option<&Stats<T::Player>>
) -> CollectedMoves<T>
where
T::Move: Sync + Send,
T::MoveError: Sync + Send,
{
if cfg!(feature = "xxhash") {
use twox_hash::RandomXxHashBuilder64;
par_move_scores_with_hasher::<T, RandomXxHashBuilder64>(game, stats)
} else {
use std::collections::hash_map::RandomState;
par_move_scores_with_hasher::<T, RandomState>(game, stats)
}
}