Spaces:
Runtime error
Runtime error
File size: 20,659 Bytes
246a547 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 | """
Advanced Feature Engineering Module
Generates 150+ features per match for improved prediction accuracy:
- Core statistics (shots, corners, cards)
- Form features with time decay
- Head-to-head history
- xG-based features
- Market/odds features
- Contextual features
"""
import pandas as pd
import numpy as np
from pathlib import Path
from typing import Dict, List, Optional, Tuple
from datetime import datetime, timedelta
import logging
logger = logging.getLogger(__name__)
# Base paths
DATA_DIR = Path(__file__).parent.parent.parent / "data"
class AdvancedFeatureEngine:
"""Generates 150+ features per match for ML prediction"""
def __init__(self, historical_data: Optional[pd.DataFrame] = None):
self.historical_data = historical_data
self.team_stats_cache = {}
self.h2h_cache = {}
if historical_data is not None:
self._build_caches()
def _build_caches(self) -> None:
"""Build team statistics and H2H caches from historical data"""
if self.historical_data is None or self.historical_data.empty:
return
df = self.historical_data
# Build team stats cache
for team in set(df.get('home_team', [])) | set(df.get('HomeTeam', [])):
if isinstance(team, str):
self.team_stats_cache[team.lower()] = self._calculate_team_stats(team)
logger.info(f"Built cache for {len(self.team_stats_cache)} teams")
def _calculate_team_stats(self, team: str) -> Dict:
"""Calculate historical statistics for a team"""
df = self.historical_data
team_lower = team.lower()
# Get home and away matches
home_col = 'home_team' if 'home_team' in df.columns else 'HomeTeam'
away_col = 'away_team' if 'away_team' in df.columns else 'AwayTeam'
home_matches = df[df[home_col].str.lower() == team_lower] if home_col in df.columns else pd.DataFrame()
away_matches = df[df[away_col].str.lower() == team_lower] if away_col in df.columns else pd.DataFrame()
stats = {
# Goals
'goals_scored_home': home_matches.get('home_goals', home_matches.get('FTHG', pd.Series())).mean() or 1.5,
'goals_conceded_home': home_matches.get('away_goals', home_matches.get('FTAG', pd.Series())).mean() or 1.2,
'goals_scored_away': away_matches.get('away_goals', away_matches.get('FTAG', pd.Series())).mean() or 1.1,
'goals_conceded_away': away_matches.get('home_goals', away_matches.get('FTHG', pd.Series())).mean() or 1.4,
# Shots
'shots_home': home_matches.get('home_shots', home_matches.get('HS', pd.Series())).mean() or 12,
'shots_away': away_matches.get('away_shots', away_matches.get('AS', pd.Series())).mean() or 10,
'shots_target_home': home_matches.get('home_shots_target', home_matches.get('HST', pd.Series())).mean() or 4,
'shots_target_away': away_matches.get('away_shots_target', away_matches.get('AST', pd.Series())).mean() or 3,
# Corners
'corners_home': home_matches.get('home_corners', home_matches.get('HC', pd.Series())).mean() or 5,
'corners_away': away_matches.get('away_corners', away_matches.get('AC', pd.Series())).mean() or 4,
# Cards
'yellows_home': home_matches.get('home_yellows', home_matches.get('HY', pd.Series())).mean() or 1.5,
'yellows_away': away_matches.get('away_yellows', away_matches.get('AY', pd.Series())).mean() or 1.7,
'reds_home': home_matches.get('home_reds', home_matches.get('HR', pd.Series())).mean() or 0.05,
'reds_away': away_matches.get('away_reds', away_matches.get('AR', pd.Series())).mean() or 0.05,
# Fouls
'fouls_home': home_matches.get('home_fouls', home_matches.get('HF', pd.Series())).mean() or 11,
'fouls_away': away_matches.get('away_fouls', away_matches.get('AF', pd.Series())).mean() or 12,
# Match counts
'home_matches': len(home_matches),
'away_matches': len(away_matches),
'total_matches': len(home_matches) + len(away_matches),
# Win rates
'home_win_rate': self._calculate_win_rate(home_matches, 'home'),
'away_win_rate': self._calculate_win_rate(away_matches, 'away'),
# xG (if available)
'xg_home': home_matches.get('home_xg', pd.Series()).mean() or 0,
'xg_away': away_matches.get('away_xg', pd.Series()).mean() or 0,
}
return stats
def _calculate_win_rate(self, matches: pd.DataFrame, team_type: str) -> float:
"""Calculate win rate from matches"""
if matches.empty:
return 0.33
result_col = 'result' if 'result' in matches.columns else 'FTR'
if result_col not in matches.columns:
return 0.33
if team_type == 'home':
wins = (matches[result_col] == 'H').sum()
else:
wins = (matches[result_col] == 'A').sum()
return wins / len(matches) if len(matches) > 0 else 0.33
def rolling_form(self, team: str, n_matches: int = 5,
decay: float = 0.9) -> Dict[str, float]:
"""Calculate rolling form with exponential time decay"""
if self.historical_data is None:
return self._default_form()
df = self.historical_data
team_lower = team.lower()
# Find team matches
home_col = 'home_team' if 'home_team' in df.columns else 'HomeTeam'
away_col = 'away_team' if 'away_team' in df.columns else 'AwayTeam'
result_col = 'result' if 'result' in df.columns else 'FTR'
# Get recent matches
home_mask = df[home_col].str.lower() == team_lower if home_col in df.columns else pd.Series([False] * len(df))
away_mask = df[away_col].str.lower() == team_lower if away_col in df.columns else pd.Series([False] * len(df))
team_matches = df[home_mask | away_mask].head(n_matches)
if team_matches.empty:
return self._default_form()
# Calculate weighted form
points = []
goals_for = []
goals_against = []
for i, (_, match) in enumerate(team_matches.iterrows()):
weight = decay ** i
is_home = str(match.get(home_col, '')).lower() == team_lower
result = match.get(result_col, 'D')
# Points
if (is_home and result == 'H') or (not is_home and result == 'A'):
points.append(3 * weight)
elif result == 'D':
points.append(1 * weight)
else:
points.append(0)
# Goals
home_goals = match.get('home_goals', match.get('FTHG', 0)) or 0
away_goals = match.get('away_goals', match.get('FTAG', 0)) or 0
if is_home:
goals_for.append(home_goals * weight)
goals_against.append(away_goals * weight)
else:
goals_for.append(away_goals * weight)
goals_against.append(home_goals * weight)
total_weight = sum(decay ** i for i in range(len(team_matches)))
return {
'form_points': sum(points) / total_weight if total_weight > 0 else 1.0,
'form_goals_scored': sum(goals_for) / total_weight if total_weight > 0 else 1.0,
'form_goals_conceded': sum(goals_against) / total_weight if total_weight > 0 else 1.0,
'form_goal_diff': (sum(goals_for) - sum(goals_against)) / total_weight if total_weight > 0 else 0,
'form_matches': len(team_matches),
'form_wins': sum(1 for p in points if p > 2),
'form_draws': sum(1 for p in points if 0 < p <= 1),
'form_losses': sum(1 for p in points if p == 0),
}
def _default_form(self) -> Dict[str, float]:
return {
'form_points': 1.5, 'form_goals_scored': 1.2, 'form_goals_conceded': 1.2,
'form_goal_diff': 0, 'form_matches': 0, 'form_wins': 0,
'form_draws': 0, 'form_losses': 0
}
def head_to_head(self, home_team: str, away_team: str,
n_matches: int = 5) -> Dict[str, float]:
"""Get head-to-head statistics"""
if self.historical_data is None:
return self._default_h2h()
cache_key = f"{home_team.lower()}_{away_team.lower()}"
if cache_key in self.h2h_cache:
return self.h2h_cache[cache_key]
df = self.historical_data
home_col = 'home_team' if 'home_team' in df.columns else 'HomeTeam'
away_col = 'away_team' if 'away_team' in df.columns else 'AwayTeam'
result_col = 'result' if 'result' in df.columns else 'FTR'
# Find H2H matches (either home or away)
mask1 = (df[home_col].str.lower() == home_team.lower()) & (df[away_col].str.lower() == away_team.lower())
mask2 = (df[home_col].str.lower() == away_team.lower()) & (df[away_col].str.lower() == home_team.lower())
h2h_matches = df[mask1 | mask2].head(n_matches)
if h2h_matches.empty:
return self._default_h2h()
# Calculate H2H stats
home_wins = 0
away_wins = 0
draws = 0
home_goals = 0
away_goals = 0
for _, match in h2h_matches.iterrows():
is_home_in_this_match = str(match.get(home_col, '')).lower() == home_team.lower()
result = match.get(result_col, 'D')
hg = match.get('home_goals', match.get('FTHG', 0)) or 0
ag = match.get('away_goals', match.get('FTAG', 0)) or 0
if is_home_in_this_match:
home_goals += hg
away_goals += ag
if result == 'H':
home_wins += 1
elif result == 'A':
away_wins += 1
else:
draws += 1
else:
home_goals += ag
away_goals += hg
if result == 'A':
home_wins += 1
elif result == 'H':
away_wins += 1
else:
draws += 1
n = len(h2h_matches)
h2h_stats = {
'h2h_matches': n,
'h2h_home_wins': home_wins,
'h2h_away_wins': away_wins,
'h2h_draws': draws,
'h2h_home_win_rate': home_wins / n if n > 0 else 0.33,
'h2h_away_win_rate': away_wins / n if n > 0 else 0.33,
'h2h_draw_rate': draws / n if n > 0 else 0.33,
'h2h_home_goals_avg': home_goals / n if n > 0 else 1.2,
'h2h_away_goals_avg': away_goals / n if n > 0 else 1.0,
'h2h_total_goals_avg': (home_goals + away_goals) / n if n > 0 else 2.2,
}
self.h2h_cache[cache_key] = h2h_stats
return h2h_stats
def _default_h2h(self) -> Dict[str, float]:
return {
'h2h_matches': 0, 'h2h_home_wins': 0, 'h2h_away_wins': 0, 'h2h_draws': 0,
'h2h_home_win_rate': 0.4, 'h2h_away_win_rate': 0.3, 'h2h_draw_rate': 0.3,
'h2h_home_goals_avg': 1.2, 'h2h_away_goals_avg': 1.0, 'h2h_total_goals_avg': 2.2
}
def odds_features(self, home_odds: float = 2.0, draw_odds: float = 3.3,
away_odds: float = 3.5) -> Dict[str, float]:
"""Extract features from betting odds"""
# Convert odds to implied probabilities
total_prob = (1/home_odds + 1/draw_odds + 1/away_odds)
home_prob = (1/home_odds) / total_prob
draw_prob = (1/draw_odds) / total_prob
away_prob = (1/away_odds) / total_prob
# Overround (bookmaker margin)
overround = total_prob - 1
return {
'odds_home': home_odds,
'odds_draw': draw_odds,
'odds_away': away_odds,
'implied_home_prob': home_prob,
'implied_draw_prob': draw_prob,
'implied_away_prob': away_prob,
'odds_overround': overround,
'odds_favorite_margin': max(home_prob, away_prob) - min(home_prob, away_prob),
'odds_is_home_favorite': 1 if home_prob > away_prob else 0,
'odds_is_away_favorite': 1 if away_prob > home_prob else 0,
'odds_home_value': home_odds * home_prob, # EV indicator
'odds_away_value': away_odds * away_prob,
}
def contextual_features(self, home_team: str, away_team: str,
match_date: Optional[datetime] = None,
is_cup: bool = False,
is_derby: bool = False) -> Dict[str, float]:
"""Extract contextual features about the match"""
if match_date is None:
match_date = datetime.now()
# Time-based features
day_of_week = match_date.weekday()
month = match_date.month
# Season position (rough estimate)
if month >= 8:
season_progress = (month - 8) / 10 # Aug to May
else:
season_progress = (month + 4) / 10
return {
'ctx_day_of_week': day_of_week,
'ctx_is_weekend': 1 if day_of_week >= 5 else 0,
'ctx_month': month,
'ctx_season_progress': min(1.0, max(0.0, season_progress)),
'ctx_is_cup': 1 if is_cup else 0,
'ctx_is_derby': 1 if is_derby else 0,
'ctx_end_of_season': 1 if month in [4, 5] else 0,
'ctx_start_of_season': 1 if month in [8, 9] else 0,
}
def extract_all_features(self, home_team: str, away_team: str,
home_odds: float = 2.0, draw_odds: float = 3.3,
away_odds: float = 3.5,
match_date: Optional[datetime] = None) -> np.ndarray:
"""Extract all 150+ features for a match"""
features = {}
# 1. Get team statistics (40+ features)
home_stats = self.team_stats_cache.get(home_team.lower(), self._calculate_team_stats(home_team))
away_stats = self.team_stats_cache.get(away_team.lower(), self._calculate_team_stats(away_team))
for key, value in home_stats.items():
features[f'home_{key}'] = value if isinstance(value, (int, float)) else 0
for key, value in away_stats.items():
features[f'away_{key}'] = value if isinstance(value, (int, float)) else 0
# 2. Rolling form (16 features: 8 per team)
home_form = self.rolling_form(home_team)
away_form = self.rolling_form(away_team)
for key, value in home_form.items():
features[f'home_{key}'] = value
for key, value in away_form.items():
features[f'away_{key}'] = value
# 3. Head-to-head (10 features)
h2h = self.head_to_head(home_team, away_team)
features.update(h2h)
# 4. Odds features (12 features)
odds_feats = self.odds_features(home_odds, draw_odds, away_odds)
features.update(odds_feats)
# 5. Contextual features (8 features)
ctx_feats = self.contextual_features(home_team, away_team, match_date)
features.update(ctx_feats)
# 6. Derived features (20+ features)
features['diff_goals_scored'] = home_stats.get('goals_scored_home', 1.5) - away_stats.get('goals_scored_away', 1.1)
features['diff_goals_conceded'] = away_stats.get('goals_conceded_away', 1.4) - home_stats.get('goals_conceded_home', 1.2)
features['diff_shots'] = home_stats.get('shots_home', 12) - away_stats.get('shots_away', 10)
features['diff_shots_target'] = home_stats.get('shots_target_home', 4) - away_stats.get('shots_target_away', 3)
features['diff_corners'] = home_stats.get('corners_home', 5) - away_stats.get('corners_away', 4)
features['diff_form_points'] = home_form.get('form_points', 1.5) - away_form.get('form_points', 1.5)
features['diff_win_rate'] = home_stats.get('home_win_rate', 0.4) - away_stats.get('away_win_rate', 0.3)
# Expected total goals
features['expected_total_goals'] = (
home_stats.get('goals_scored_home', 1.5) +
away_stats.get('goals_scored_away', 1.1)
)
# BTTS indicator
features['btts_indicator'] = min(
1 - (1 - home_stats.get('goals_scored_home', 1.5) / 3),
1 - (1 - away_stats.get('goals_scored_away', 1.1) / 3)
)
# Convert to numpy array
feature_values = [float(v) if isinstance(v, (int, float)) and not np.isnan(v) else 0.0
for v in features.values()]
return np.array(feature_values)
def get_feature_names(self) -> List[str]:
"""Get list of all feature names"""
# Generate a dummy extraction to get feature names
features = {}
# Add all feature groups
for prefix in ['home_', 'away_']:
for stat in ['goals_scored_home', 'goals_conceded_home', 'goals_scored_away',
'goals_conceded_away', 'shots_home', 'shots_away', 'shots_target_home',
'shots_target_away', 'corners_home', 'corners_away', 'yellows_home',
'yellows_away', 'reds_home', 'reds_away', 'fouls_home', 'fouls_away',
'home_matches', 'away_matches', 'total_matches', 'home_win_rate',
'away_win_rate', 'xg_home', 'xg_away']:
features[f'{prefix}{stat}'] = 0
for form in ['form_points', 'form_goals_scored', 'form_goals_conceded',
'form_goal_diff', 'form_matches', 'form_wins', 'form_draws', 'form_losses']:
features[f'{prefix}{form}'] = 0
# H2H
for h2h in ['h2h_matches', 'h2h_home_wins', 'h2h_away_wins', 'h2h_draws',
'h2h_home_win_rate', 'h2h_away_win_rate', 'h2h_draw_rate',
'h2h_home_goals_avg', 'h2h_away_goals_avg', 'h2h_total_goals_avg']:
features[h2h] = 0
# Odds
for odds in ['odds_home', 'odds_draw', 'odds_away', 'implied_home_prob',
'implied_draw_prob', 'implied_away_prob', 'odds_overround',
'odds_favorite_margin', 'odds_is_home_favorite', 'odds_is_away_favorite',
'odds_home_value', 'odds_away_value']:
features[odds] = 0
# Context
for ctx in ['ctx_day_of_week', 'ctx_is_weekend', 'ctx_month', 'ctx_season_progress',
'ctx_is_cup', 'ctx_is_derby', 'ctx_end_of_season', 'ctx_start_of_season']:
features[ctx] = 0
# Derived
for diff in ['diff_goals_scored', 'diff_goals_conceded', 'diff_shots',
'diff_shots_target', 'diff_corners', 'diff_form_points', 'diff_win_rate',
'expected_total_goals', 'btts_indicator']:
features[diff] = 0
return list(features.keys())
def create_feature_engine(data_path: Optional[Path] = None) -> AdvancedFeatureEngine:
"""Create a feature engine with historical data"""
if data_path is None:
data_path = DATA_DIR / "processed" / "master_training_data.csv"
if data_path.exists():
df = pd.read_csv(data_path)
return AdvancedFeatureEngine(df)
# Try existing training data
existing = DATA_DIR / "comprehensive_training_data.csv"
if existing.exists():
df = pd.read_csv(existing)
return AdvancedFeatureEngine(df)
return AdvancedFeatureEngine()
if __name__ == "__main__":
logging.basicConfig(level=logging.INFO)
# Test feature extraction
engine = create_feature_engine()
features = engine.extract_all_features(
home_team="Arsenal",
away_team="Chelsea",
home_odds=2.1,
draw_odds=3.4,
away_odds=3.2
)
print(f"Generated {len(features)} features")
print(f"Feature names: {len(engine.get_feature_names())}")
|