File size: 8,431 Bytes
94d271b |
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 |
"""
Transposition Table with Zobrist Hashing
Research: Stockfish uses 2GB TT, we use 400MB for Colab constraints
References:
- Zobrist (1970) - Hash functions for chess positions
- Stockfish TT - Replacement strategies
- AlphaBeta enhancements - Exact/Lower/Upper bounds
"""
import chess
import numpy as np
from typing import Optional, Dict, Tuple
from enum import Enum
class NodeType(Enum):
"""Type of transposition table entry"""
EXACT = 0 # PV-node (exact score)
LOWER_BOUND = 1 # Cut-node (beta cutoff)
UPPER_BOUND = 2 # All-node (failed low)
class TTEntry:
"""Single transposition table entry"""
__slots__ = ['zobrist_key', 'depth', 'score', 'node_type', 'best_move', 'age']
def __init__(
self,
zobrist_key: int,
depth: int,
score: float,
node_type: NodeType,
best_move: Optional[chess.Move],
age: int
):
self.zobrist_key = zobrist_key
self.depth = depth
self.score = score
self.node_type = node_type
self.best_move = best_move
self.age = age
class TranspositionTable:
"""
Zobrist-hashed transposition table
Replacement strategy: Always replace if deeper or newer
"""
def __init__(self, size_mb: int = 256):
"""
Initialize transposition table
Args:
size_mb: Table size in megabytes (default 256MB)
"""
# Calculate number of entries (each entry ~64 bytes)
bytes_per_entry = 64
self.max_entries = (size_mb * 1024 * 1024) // bytes_per_entry
# Hash table (dict for simplicity, could use array for speed)
self.table: Dict[int, TTEntry] = {}
# Statistics
self.hits = 0
self.misses = 0
self.collisions = 0
self.current_age = 0
# Zobrist keys for hashing (initialized once)
self._init_zobrist_keys()
def _init_zobrist_keys(self):
"""
Initialize Zobrist random keys
One key per (piece_type, color, square) combination
"""
np.random.seed(42) # Reproducible keys
self.zobrist_pieces = np.random.randint(
0, 2**63, size=(12, 64), dtype=np.int64
)
# Additional keys for game state
self.zobrist_turn = np.random.randint(0, 2**63, dtype=np.int64)
self.zobrist_castling = np.random.randint(0, 2**63, size=4, dtype=np.int64)
self.zobrist_ep = np.random.randint(0, 2**63, size=8, dtype=np.int64)
def compute_zobrist_key(self, board: chess.Board) -> int:
"""
Compute Zobrist hash for position
Args:
board: chess.Board
Returns:
64-bit Zobrist key
"""
key = 0
# Piece positions
piece_to_index = {
(chess.PAWN, chess.WHITE): 0,
(chess.KNIGHT, chess.WHITE): 1,
(chess.BISHOP, chess.WHITE): 2,
(chess.ROOK, chess.WHITE): 3,
(chess.QUEEN, chess.WHITE): 4,
(chess.KING, chess.WHITE): 5,
(chess.PAWN, chess.BLACK): 6,
(chess.KNIGHT, chess.BLACK): 7,
(chess.BISHOP, chess.BLACK): 8,
(chess.ROOK, chess.BLACK): 9,
(chess.QUEEN, chess.BLACK): 10,
(chess.KING, chess.BLACK): 11,
}
for square, piece in board.piece_map().items():
piece_idx = piece_to_index[(piece.piece_type, piece.color)]
key ^= self.zobrist_pieces[piece_idx, square]
# Turn
if board.turn == chess.BLACK:
key ^= self.zobrist_turn
# Castling rights
if board.has_kingside_castling_rights(chess.WHITE):
key ^= self.zobrist_castling[0]
if board.has_queenside_castling_rights(chess.WHITE):
key ^= self.zobrist_castling[1]
if board.has_kingside_castling_rights(chess.BLACK):
key ^= self.zobrist_castling[2]
if board.has_queenside_castling_rights(chess.BLACK):
key ^= self.zobrist_castling[3]
# En passant
if board.ep_square is not None:
ep_file = board.ep_square % 8
key ^= self.zobrist_ep[ep_file]
return key
def probe(
self,
zobrist_key: int,
depth: int,
alpha: float,
beta: float
) -> Optional[Tuple[float, Optional[chess.Move]]]:
"""
Probe transposition table
Args:
zobrist_key: Zobrist hash of position
depth: Current search depth
alpha: Alpha value
beta: Beta value
Returns:
(score, best_move) if usable entry found, else None
"""
entry = self.table.get(zobrist_key)
if entry is None:
self.misses += 1
return None
# Zobrist collision check
if entry.zobrist_key != zobrist_key:
self.collisions += 1
return None
# Depth check: only use if searched deeper
if entry.depth < depth:
self.misses += 1
return None
self.hits += 1
# Check if score is usable based on node type
score = entry.score
if entry.node_type == NodeType.EXACT:
return (score, entry.best_move)
elif entry.node_type == NodeType.LOWER_BOUND:
if score >= beta:
return (score, entry.best_move)
elif entry.node_type == NodeType.UPPER_BOUND:
if score <= alpha:
return (score, entry.best_move)
# Entry exists but not usable for cutoff
# Still return best_move for move ordering
return (None, entry.best_move)
def store(
self,
zobrist_key: int,
depth: int,
score: float,
node_type: NodeType,
best_move: Optional[chess.Move]
):
"""
Store entry in transposition table
Args:
zobrist_key: Zobrist hash
depth: Search depth
score: Position score
node_type: Type of node (exact/lower/upper)
best_move: Best move found
"""
# Check if we should replace existing entry
existing = self.table.get(zobrist_key)
if existing is not None:
# Always replace if:
# 1. New search is deeper
# 2. Same depth but newer (generational replacement)
if depth < existing.depth and existing.age == self.current_age:
return # Keep existing deeper entry
# Store new entry
self.table[zobrist_key] = TTEntry(
zobrist_key=zobrist_key,
depth=depth,
score=score,
node_type=node_type,
best_move=best_move,
age=self.current_age
)
# Cleanup if table too large (simple strategy)
if len(self.table) > self.max_entries:
self._cleanup_old_entries()
def _cleanup_old_entries(self):
"""Remove oldest 10% of entries"""
entries_to_remove = self.max_entries // 10
# Remove oldest entries (by age)
old_keys = sorted(
self.table.keys(),
key=lambda k: self.table[k].age
)[:entries_to_remove]
for key in old_keys:
del self.table[key]
def increment_age(self):
"""Increment generation counter (call at search start)"""
self.current_age += 1
def clear(self):
"""Clear all entries"""
self.table.clear()
self.hits = 0
self.misses = 0
self.collisions = 0
def get_stats(self) -> Dict:
"""Get table statistics"""
total_probes = self.hits + self.misses
hit_rate = (self.hits / total_probes * 100) if total_probes > 0 else 0
return {
'entries': len(self.table),
'max_entries': self.max_entries,
'usage_percent': len(self.table) / self.max_entries * 100,
'hits': self.hits,
'misses': self.misses,
'hit_rate': hit_rate,
'collisions': self.collisions
} |