File size: 9,688 Bytes
e3e5e3d |
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 |
"""
Advanced Move Ordering Techniques
Critical for alpha-beta pruning efficiency
Research References:
- Stockfish move ordering (2023) - History heuristic + killer moves
- Fruit 2.1 - MVV-LVA and late move reductions
- Crafty - History tables and counter moves
Good move ordering can reduce search tree by 10-100x!
"""
import chess
from typing import List, Optional, Dict, Tuple
import numpy as np
class MoveOrderer:
"""
Advanced move ordering system
Combines multiple heuristics for optimal move ordering
"""
# Piece values for MVV-LVA (victim values)
PIECE_VALUES = {
chess.PAWN: 100,
chess.KNIGHT: 320,
chess.BISHOP: 330,
chess.ROOK: 500,
chess.QUEEN: 900,
chess.KING: 20000
}
def __init__(self):
"""Initialize move ordering data structures"""
# Killer moves (quiet moves that caused beta cutoff)
# killer_moves[depth] = [move1, move2]
self.killer_moves: Dict[int, List[Optional[chess.Move]]] = {}
self.max_killers = 2 # Store top 2 killer moves per depth
# History heuristic (move success rates)
# history[from_square][to_square] = score
self.history = np.zeros((64, 64), dtype=np.int32)
# Counter moves (refutation table)
# counter_moves[prev_move] = best_response
self.counter_moves: Dict[chess.Move, Optional[chess.Move]] = {}
# Statistics
self.history_hits = 0
self.killer_hits = 0
def order_moves(
self,
board: chess.Board,
moves: List[chess.Move],
depth: int,
tt_move: Optional[chess.Move] = None,
previous_move: Optional[chess.Move] = None
) -> List[chess.Move]:
"""
Order moves for optimal alpha-beta pruning
Priority order (research-based):
1. TT move (from transposition table)
2. Winning captures (MVV-LVA)
3. Killer moves
4. Counter moves
5. History heuristic
6. Losing captures
7. Quiet moves
Args:
board: Current position
moves: Legal moves to order
depth: Current search depth
tt_move: Best move from transposition table
previous_move: Opponent's last move
Returns:
Ordered list of moves
"""
scored_moves = []
for move in moves:
score = 0
# ========== PRIORITY 1: TT Move ==========
if tt_move and move == tt_move:
score += 1000000
# ========== PRIORITY 2: Captures (MVV-LVA) ==========
elif board.is_capture(move):
score += self._score_capture(board, move)
# ========== PRIORITY 3-7: Quiet Moves ==========
else:
# Check if it's a killer move
if self._is_killer_move(move, depth):
score += 9000
self.killer_hits += 1
# Counter move bonus
if previous_move and move == self.counter_moves.get(previous_move):
score += 8000
# History heuristic
history_score = self.history[move.from_square, move.to_square]
score += min(history_score, 7000)
# Promotions
if move.promotion == chess.QUEEN:
score += 10000
elif move.promotion in [chess.KNIGHT, chess.ROOK, chess.BISHOP]:
score += 5000
# Checks
board.push(move)
if board.is_check():
score += 6000
board.pop()
# Castling
if board.is_castling(move):
score += 3000
# Positional bonuses
score += self._score_positional(board, move)
scored_moves.append((score, move))
# Sort by score (descending)
scored_moves.sort(key=lambda x: x[0], reverse=True)
return [move for _, move in scored_moves]
def _score_capture(self, board: chess.Board, move: chess.Move) -> int:
"""
Score capture using MVV-LVA (Most Valuable Victim - Least Valuable Attacker)
Enhanced with SEE (Static Exchange Evaluation)
"""
captured_piece = board.piece_at(move.to_square)
moving_piece = board.piece_at(move.from_square)
if not captured_piece or not moving_piece:
return 0
victim_value = self.PIECE_VALUES.get(captured_piece.piece_type, 0)
attacker_value = self.PIECE_VALUES.get(moving_piece.piece_type, 1)
# MVV-LVA base score
mvv_lva_score = (victim_value * 10 - attacker_value) * 100
# Bonus for en passant
if board.is_en_passant(move):
mvv_lva_score += 10500
# Penalty for hanging pieces (simplified SEE)
# Check if capture square is defended
if board.is_attacked_by(not board.turn, move.to_square):
# Losing capture if victim < attacker
if victim_value < attacker_value:
mvv_lva_score -= 5000 # Deprioritize losing captures
return mvv_lva_score
def _score_positional(self, board: chess.Board, move: chess.Move) -> int:
"""
Positional scoring for quiet moves
Based on Stockfish evaluation terms
"""
score = 0
piece = board.piece_at(move.from_square)
if not piece:
return 0
# Center control (e4, d4, e5, d5)
center_squares = [chess.E4, chess.D4, chess.E5, chess.D5]
if move.to_square in center_squares:
score += 50
# Extended center
extended_center = [
chess.C3, chess.D3, chess.E3, chess.F3,
chess.C4, chess.F4,
chess.C5, chess.F5,
chess.C6, chess.D6, chess.E6, chess.F6
]
if move.to_square in extended_center:
score += 20
# Piece development (from back rank)
if piece.piece_type in [chess.KNIGHT, chess.BISHOP]:
from_rank = move.from_square // 8
if from_rank in [0, 7]: # Back rank
score += 30
# Knight outposts (protected squares in enemy territory)
if piece.piece_type == chess.KNIGHT:
to_rank = move.to_square // 8
if (board.turn == chess.WHITE and to_rank >= 4) or \
(board.turn == chess.BLACK and to_rank <= 3):
# Check if protected by pawn
board.push(move)
if board.is_attacked_by(board.turn, move.to_square):
score += 40
board.pop()
return score
def _is_killer_move(self, move: chess.Move, depth: int) -> bool:
"""Check if move is a killer move at this depth"""
killers = self.killer_moves.get(depth, [])
return move in killers
def update_killer_move(self, move: chess.Move, depth: int):
"""
Update killer moves for this depth
Killer moves are quiet moves that caused beta cutoff
"""
if depth not in self.killer_moves:
self.killer_moves[depth] = []
killers = self.killer_moves[depth]
# Add move if not already in list
if move not in killers:
killers.insert(0, move)
# Keep only top N killers
self.killer_moves[depth] = killers[:self.max_killers]
def update_history(self, move: chess.Move, depth: int, success: bool):
"""
Update history heuristic
Args:
move: Move that was tried
depth: Search depth
success: True if move caused beta cutoff
"""
if success:
# Increase score (depth squared bonus for deeper searches)
bonus = depth * depth
self.history[move.from_square, move.to_square] += bonus
self.history_hits += 1
else:
# Slight penalty for moves that failed low
self.history[move.from_square, move.to_square] -= 1
# Cap values to prevent overflow
self.history = np.clip(self.history, -10000, 10000)
def update_counter_move(self, previous_move: chess.Move, refutation: chess.Move):
"""
Update counter move table
Counter move = best response to opponent's last move
"""
self.counter_moves[previous_move] = refutation
def clear_history(self):
"""Clear history table (called at new search)"""
self.history.fill(0)
self.killer_moves.clear()
self.history_hits = 0
self.killer_hits = 0
def age_history(self, factor: float = 0.9):
"""
Age history table (reduce all values)
Prevents old data from dominating
"""
self.history = (self.history * factor).astype(np.int32)
def get_stats(self) -> Dict:
"""Get move ordering statistics"""
return {
'killer_hits': self.killer_hits,
'history_hits': self.history_hits,
'history_max': int(np.max(self.history)),
'killer_depths': len(self.killer_moves),
'counter_moves': len(self.counter_moves)
} |