""" Random Forest + CLV combined strategy. Walk-forward: train RF on prior seasons to predict away win probability, then bet only when BOTH model pA >= threshold AND CLV > 5%. """ import pandas as pd import numpy as np from sklearn.ensemble import RandomForestClassifier from sklearn.calibration import CalibratedClassifierCV import warnings warnings.filterwarnings('ignore') START = 1_000.0 CLV_THRESH = 0.05 PA_THRESH = 0.31 # model confidence threshold (from old pipeline) KELLY_MULTS = [0.10, 0.25, 0.50] seasons_map = { 2017: '1718', 2018: '1819', 2019: '1920', 2020: '2021', 2021: '2122', 2022: '2223', 2023: '2324', 2024: '2425', } OOS_API_SEASONS = [2021, 2022, 2023, 2024] # = 2122, 2223, 2324, 2425 league_names = { 'E0': 'EPL', 'E1': 'Championship', 'E2': 'League One', 'SP1': 'La Liga', 'SP2': 'Segunda', 'I1': 'Serie A', 'I2': 'Serie B', 'D1': 'Bundesliga', 'D2': '2.Bundesliga', 'F1': 'Ligue 1', 'N1': 'Eredivisie', 'B1': 'Belgian Pro League', 'T1': 'Super Lig', 'G1': 'Greek SL', 'SC0': 'Scottish Prem', } def remove_margin(h, d, a): over = 1/h + 1/d + 1/a return 1/h/over, 1/d/over, 1/a/over # ── Load features ──────────────────────────────────────────────────────────── print("Loading features...") feat = pd.read_csv('E:/soccer/raw_api/match_features.csv') feat['date'] = pd.to_datetime(feat['date']) # Pivot: one row per match with home_ and away_ features w5_cols = [c for c in feat.columns if '_w5' in c] w10_cols = [c for c in feat.columns if '_w10' in c] feature_cols = w5_cols + w10_cols home_feat = feat[feat['is_home'] == 1][['fixture_id', 'date', 'div', 'season', 'home_team', 'away_team', 'home_goals', 'away_goals'] + feature_cols].copy() home_feat.columns = ['fixture_id', 'date', 'div', 'season', 'home_team', 'away_team', 'home_goals', 'away_goals'] + [f'h_{c}' for c in feature_cols] away_feat = feat[feat['is_home'] == 0][['fixture_id'] + feature_cols].copy() away_feat.columns = ['fixture_id'] + [f'a_{c}' for c in feature_cols] matches = home_feat.merge(away_feat, on='fixture_id', how='inner') matches['away_win'] = (matches['away_goals'] > matches['home_goals']).astype(int) print(f" {len(matches):,} matches with features") # ── Load odds from league CSVs ────────────────────────────────────────────── print("Loading odds...") csv_seasons = ['1718','1819','1920','2021','2122','2223','2324','2425','2526'] all_odds = [] for code in league_names.keys(): for s in csv_seasons: path = f'E:/soccer/league_data/{code}_{s}.csv' try: d = pd.read_csv(path, encoding='utf-8-sig') if 'B365A' not in d.columns or 'PSCH' not in d.columns: continue d = d.dropna(subset=['FTR','B365A','B365H','B365D','PSCH','PSCD','PSCA']) d['Date'] = pd.to_datetime(d['Date'], dayfirst=True, errors='coerce') d = d.dropna(subset=['Date']) d['div'] = code all_odds.append(d) except: pass odds_df = pd.concat(all_odds, ignore_index=True) # Compute fair odds and CLV tH, tD, tA = zip(*odds_df.apply(lambda r: remove_margin(r['PSCH'],r['PSCD'],r['PSCA']), axis=1)) odds_df['p_A'] = pd.Series(tA, index=odds_df.index) odds_df['PSC_fair_A'] = 1 / odds_df['p_A'] odds_df['edge_A'] = (odds_df['B365A'] - odds_df['PSC_fair_A']) / odds_df['PSC_fair_A'] print(f" {len(odds_df):,} matches with odds") # ── Merge features with odds ──────────────────────────────────────────────── print("Merging features with odds...") # Match on date + home team + away team + league matches['date_str'] = matches['date'].dt.strftime('%Y-%m-%d') odds_df['date_str'] = odds_df['Date'].dt.strftime('%Y-%m-%d') # Normalize team names for matching TEAM_MAP = { # API-Football name -> football-data name 'manchester city': 'man city', 'manchester united': 'man united', 'nottingham forest': "nott'm forest", 'tottenham hotspur': 'tottenham', 'newcastle united': 'newcastle', 'wolverhampton wanderers': 'wolves', 'west ham united': 'west ham', 'leicester city': 'leicester', 'leeds united': 'leeds', 'brighton & hove albion': 'brighton', 'brighton hove albion': 'brighton', 'afc bournemouth': 'bournemouth', 'sheffield united': 'sheffield united', 'burnley fc': 'burnley', 'crystal palace': 'crystal palace', 'atlético madrid': 'ath madrid', 'atletico madrid': 'ath madrid', 'athletic club': 'ath bilbao', 'athletic bilbao': 'ath bilbao', 'real betis': 'betis', 'real sociedad': 'sociedad', 'real valladolid': 'valladolid', 'celta vigo': 'celta', 'deportivo alavés': 'alaves', 'deportivo alaves': 'alaves', 'rayo vallecano': 'vallecano', 'cádiz': 'cadiz', 'almería': 'almeria', 'las palmas': 'las palmas', 'fc barcelona': 'barcelona', 'real madrid': 'real madrid', 'paris saint germain': 'paris sg', 'as monaco': 'monaco', 'olympique lyonnais': 'lyon', 'olympique de marseille': 'marseille', 'as saint-étienne': 'st etienne', 'rc strasbourg alsace': 'strasbourg', 'stade rennais fc': 'rennes', 'stade de reims': 'reims', 'rc lens': 'lens', 'losc lille': 'lille', 'fc nantes': 'nantes', 'girondins de bordeaux': 'bordeaux', 'angers sco': 'angers', 'clermont foot': 'clermont foot', 'estac troyes': 'troyes', 'toulouse fc': 'toulouse', 'fc lorient': 'lorient', 'ac ajaccio': 'ajaccio', 'stade brestois 29': 'brest', 'montpellier hsc': 'montpellier', 'ogc nice': 'nice', 'le havre ac': 'le havre', 'borussia dortmund': 'dortmund', 'borussia mönchengladbach': "m'gladbach", 'borussia monchengladbach': "m'gladbach", 'bayern munich': 'bayern munich', 'bayer 04 leverkusen': 'leverkusen', 'bayer leverkusen': 'leverkusen', 'rb leipzig': 'rb leipzig', 'eintracht frankfurt': 'ein frankfurt', 'vfb stuttgart': 'stuttgart', 'vfl wolfsburg': 'wolfsburg', 'sc freiburg': 'freiburg', '1. fc union berlin': 'union berlin', '1. fc köln': 'fc koln', '1. fc koln': 'fc koln', '1. fc k\x94ln': 'fc koln', 'tsg hoffenheim': 'hoffenheim', '1899 hoffenheim': 'hoffenheim', 'fc augsburg': 'augsburg', 'sv werder bremen': 'werder bremen', 'werder bremen': 'werder bremen', '1. fc heidenheim': 'heidenheim', '1. fc heidenheim 1846': 'heidenheim', 'hertha berlin': 'hertha', 'hertha bsc': 'hertha', 'arminia bielefeld': 'bielefeld', 'vfl bochum': 'bochum', 'fc schalke 04': 'schalke 04', 'greuther fürth': 'greuther furth', 'greuther furth': 'greuther furth', 'fortuna düsseldorf': 'fortuna dusseldorf', 'fortuna dusseldorf': 'fortuna dusseldorf', 'darmstadt 98': 'darmstadt', 'sv darmstadt 98': 'darmstadt', 'holstein kiel': 'holstein kiel', 'fc st. pauli': 'st pauli', 'inter': 'inter', 'internazionale': 'inter', 'ac milan': 'milan', 'as roma': 'roma', 'ssc napoli': 'napoli', 'ss lazio': 'lazio', 'juventus': 'juventus', 'atalanta': 'atalanta', 'us sassuolo calcio': 'sassuolo', 'hellas verona': 'verona', 'bologna fc 1909': 'bologna', 'bologna': 'bologna', 'udinese': 'udinese', 'uc sampdoria': 'sampdoria', 'genoa': 'genoa', 'us salernitana 1919': 'salernitana', 'torino': 'torino', 'acf fiorentina': 'fiorentina', 'us lecce': 'lecce', 'empoli': 'empoli', 'spezia': 'spezia', 'us cremonese': 'cremonese', 'cagliari': 'cagliari', 'venezia': 'venezia', 'frosinone': 'frosinone', 'monza': 'monza', 'como': 'como', 'parma': 'parma', 'ajax': 'ajax', 'feyenoord': 'feyenoord', 'psv eindhoven': 'psv', 'az alkmaar': 'az alkmaar', 'az': 'az alkmaar', 'fc twente': 'twente', 'sc heerenveen': 'heerenveen', 'fc utrecht': 'fc utrecht', 'fc groningen': 'groningen', 'vitesse': 'vitesse', 'sparta rotterdam': 'sparta rotterdam', 'rkc waalwijk': 'waalwijk', 'nec nijmegen': 'nec', 'sc cambuur': 'cambuur', 'go ahead eagles': 'go ahead eagles', 'fc volendam': 'volendam', 'excelsior': 'excelsior', 'almere city fc': 'almere city', 'fortuna sittard': 'fort sittard', 'pec zwolle': 'pec zwolle', 'heracles almelo': 'heracles', 'willem ii': 'willem ii', 'nac breda': 'nac breda', 'club brugge kv': 'club brugge', 'royal antwerp': 'antwerp', 'anderlecht': 'anderlecht', 'krc genk': 'genk', 'standard liège': 'standard', 'standard liege': 'standard', 'kaa gent': 'gent', 'oh leuven': 'oh leuven', 'union saint-gilloise': 'st gilloise', 'union st. gilloise': 'st gilloise', 'cercle brugge': 'cercle brugge', 'kv mechelen': 'mechelen', 'charleroi': 'charleroi', 'kv kortrijk': 'kortrijk', 'sint-truidense vv': 'st truiden', 'rwdm': 'rwdm', 'as eupen': 'eupen', 'beerschot va': 'beerschot', 'seraing': 'seraing', 'waasland-beveren': 'waasland', 'westerlo': 'westerlo', 'fc dender': 'dender', 'celtic': 'celtic', 'rangers': 'rangers', 'aberdeen': 'aberdeen', 'heart of midlothian': 'hearts', 'hearts': 'hearts', 'hibernian': 'hibernian', 'motherwell': 'motherwell', 'st. johnstone': 'st johnstone', 'dundee': 'dundee', 'livingston': 'livingston', 'ross county': 'ross county', 'kilmarnock': 'kilmarnock', 'st. mirren': 'st mirren', 'dundee united': 'dundee utd', 'galatasaray': 'galatasaray', 'fenerbahce': 'fenerbahce', 'fenerbahçe': 'fenerbahce', 'besiktas': 'besiktas', 'beşiktaş': 'besiktas', 'trabzonspor': 'trabzonspor', 'istanbul basaksehir': 'basaksehir', 'istanbul başakşehir': 'basaksehir', 'adana demirspor': 'adana demirspor', 'antalyaspor': 'antalyaspor', 'sivasspor': 'sivasspor', 'kayserispor': 'kayserispor', 'alanyaspor': 'alanyaspor', 'konyaspor': 'konyaspor', 'kasimpasa': 'kasimpasa', 'kasımpaşa': 'kasimpasa', 'gaziantep fk': 'gaziantep', 'ankaragücü': 'ankaragucu', 'ankaragucu': 'ankaragucu', 'hatayspor': 'hatayspor', 'giresunspor': 'giresunspor', 'ümraniyespor': 'umraniye', 'istanbulspor': 'istanbulspor', 'pendikspor': 'pendikspor', 'rizespor': 'rizespor', 'samsunspor': 'samsunspor', 'bodrum fk': 'bodrum fk', 'eyüpspor': 'eyupspor', 'eyupspor': 'eyupspor', 'olympiacos piraeus': 'olympiakos', 'panathinaikos': 'panathinaikos', 'paok thessaloniki fc': 'paok', 'aris thessalonikis': 'aris', 'aek athens fc': 'aek', 'volos nfc': 'volos', 'asteras tripolis': 'asteras tripolis', 'ionikos': 'ionikos', 'ofi': 'ofi', 'atromitos': 'atromitos', 'pas giannina': 'giannina', 'lamia': 'lamia', 'levadiakos': 'levadiakos', 'panserraikos': 'panserraikos', 'kallithea': 'kallithea', # Bundesliga extras 'fsv mainz 05': 'mainz', 'mainz 05': 'mainz', '1. fsv mainz 05': 'mainz', 'sc paderborn 07': 'paderborn', 'fc nurnberg': 'nurnberg', '1. fc nürnberg': 'nurnberg', 'hannover 96': 'hannover', 'bayern münchen': 'bayern munich', 'bayern m\x81nchen': 'bayern munich', 'dynamo dresden': 'dresden', 'jahn regensburg': 'regensburg', 'fc kaiserslautern': 'kaiserslautern', '1. fc kaiserslautern': 'kaiserslautern', 'hamburger sv': 'hamburger sv', 'karlsruher sc': 'karlsruhe', 'fc magdeburg': 'magdeburg', '1. fc magdeburg': 'magdeburg', 'eintracht braunschweig': 'braunschweig', 'fc erzgebirge aue': 'erzgebirge aue', 'ssv ulm 1846': 'ulm', 'preussen munster': 'preussen munster', 'fc elversberg': 'elversberg', # Greek extras 'apollon smirnis': 'apollon', 'olympiakos piraeus': 'olympiakos', # Dutch extras 'roda': 'roda', 'de graafschap': 'de graafschap', 'fc eindhoven': 'eindhoven', 'ado den haag': 'den haag', # Italian extras 'chievo': 'chievo', # Scottish extras 'hamilton academical': 'hamilton', # Spanish extras 'deportivo la coruna': 'la coruna', 'espanyol': 'espanol', 'real sociedad ii': 'sociedad b', 'villarreal ii': 'villarreal b', 'barcelona b': 'barcelona b', 'rayo majadahonda': 'majadahonda', # French extras 'saint etienne': 'st etienne', 'guingamp': 'guingamp', 'caen': 'caen', # Championship / League One / League Two additions 'west bromwich albion': 'west brom', 'queens park rangers': 'qpr', 'sheffield wednesday': 'sheffield weds', 'swansea city': 'swansea', 'stoke city': 'stoke', 'hull city': 'hull', 'cardiff city': 'cardiff', 'coventry city': 'coventry', 'norwich city': 'norwich', 'watford': 'watford', 'blackburn rovers': 'blackburn', 'bristol city': 'bristol city', 'millwall': 'millwall', 'middlesbrough': 'middlesbrough', 'sunderland': 'sunderland', 'preston north end': 'preston', 'luton town': 'luton', 'plymouth argyle': 'plymouth', 'ipswich town': 'ipswich', 'rotherham united': 'rotherham', 'huddersfield town': 'huddersfield', 'birmingham city': 'birmingham', 'reading': 'reading', 'blackpool': 'blackpool', 'wigan athletic': 'wigan', 'barnsley': 'barnsley', 'derby county': 'derby', 'peterborough united': 'peterborough', 'portsmouth': 'portsmouth', 'bolton wanderers': 'bolton', 'oxford united': 'oxford utd', 'charlton athletic': 'charlton', 'exeter city': 'exeter', 'port vale': 'port vale', 'shrewsbury town': 'shrewsbury', 'fleetwood town': 'fleetwood town', 'cambridge united': 'cambridge utd', 'cheltenham town': 'cheltenham', 'burton albion': 'burton', 'wycombe wanderers': 'wycombe', 'morecambe': 'morecambe', 'accrington st': 'accrington', 'accrington stanley': 'accrington', 'lincoln city': 'lincoln', 'mk dons': 'mk dons', 'milton keynes dons': 'mk dons', # Segunda 'racing santander': 'racing sant', 'real zaragoza': 'zaragoza', 'real oviedo': 'oviedo', 'sd huesca': 'huesca', 'racing ferrol': 'ferrol', 'fc cartagena': 'cartagena', 'cd mirandés': 'mirandes', 'cd mirandes': 'mirandes', 'sd eibar': 'eibar', 'cd tenerife': 'tenerife', 'cd leganés': 'leganes', 'cd leganes': 'leganes', 'elche': 'elche', 'málaga': 'malaga', 'malaga': 'malaga', 'albacete': 'albacete', 'burgos cf': 'burgos', 'ponferradina': 'ponferradina', 'lugo': 'lugo', 'ibiza': 'ibiza', 'villarreal b': 'villarreal b', 'ud las palmas': 'las palmas', 'sporting gijón': 'sp gijon', 'sporting gijon': 'sp gijon', 'granada cf': 'granada', 'levante': 'levante', 'real sporting de gijón': 'sp gijon', # Serie B 'reggina': 'reggina', 'perugia': 'perugia', 'como 1907': 'como', 'spal': 'spal', 'benevento': 'benevento', 'brescia': 'brescia', 'ascoli': 'ascoli', 'ternana': 'ternana', 'cosenza': 'cosenza', 'cittadella': 'cittadella', 'pisa': 'pisa', 'bari': 'bari', 'südtirol': 'sudtirol', 'sudtirol': 'sudtirol', 'palermo': 'palermo', 'catanzaro': 'catanzaro', 'modena': 'modena', 'sampdoria': 'sampdoria', 'reggiana': 'reggiana', 'juve stabia': 'juve stabia', 'mantova': 'mantova', 'cesena': 'cesena', 'carrarese': 'carrarese', } def norm(s): import unicodedata, re s = str(s).lower().strip() # Remove accents s = unicodedata.normalize('NFKD', s).encode('ascii', 'ignore').decode('ascii') if s in TEAM_MAP: return TEAM_MAP[s] # Try without FC/AFC/SC prefix/suffix s2 = re.sub(r'\b(fc|afc|sc|cf|cd|sd|ss|us|uc|rc|ac|as|og|rk|sv|vf[blr]|tsg|ssv|1\.|ssc)\b', '', s).strip() s2 = re.sub(r'\s+', ' ', s2).strip() if s2 in TEAM_MAP: return TEAM_MAP[s2] return TEAM_MAP.get(s, s) matches['h_norm'] = matches['home_team'].apply(norm) matches['a_norm'] = matches['away_team'].apply(norm) odds_df['h_norm'] = odds_df['HomeTeam'].apply(norm) odds_df['a_norm'] = odds_df['AwayTeam'].apply(norm) # Merge on date + league + normalized team names merged = matches.merge(odds_df[['date_str','div','h_norm','a_norm','B365A','B365H','B365D', 'p_A','PSC_fair_A','edge_A','FTR','FTHG','FTAG','HTHG','HTAG']], on=['date_str','div','h_norm','a_norm'], how='inner') # Report unmatched for debugging if len(merged) < len(matches) * 0.8: unmatched = matches[~matches['fixture_id'].isin(merged['fixture_id'])] um_teams = unmatched.groupby('div').apply( lambda g: list(set(g['h_norm'].unique()) - set(odds_df[odds_df['div']==g.name]['h_norm'].unique()))[:5] ) print(f" WARNING: Only {len(merged)/len(matches)*100:.0f}% matched. Unmatched team samples:") for div, teams in um_teams.items(): if teams: safe = [t.encode('ascii', 'replace').decode() for t in teams] print(f" {div}: {safe}") print(f" {len(merged):,} matched (features + odds)") # ── Build model features ──────────────────────────────────────────────────── h_features = [f'h_{c}' for c in feature_cols] a_features = [f'a_{c}' for c in feature_cols] # Difference features: home - away diff_features = [] for hc, ac in zip(h_features, a_features): diff_col = 'diff_' + hc[2:] # strip h_ prefix merged[diff_col] = merged[hc] - merged[ac] diff_features.append(diff_col) # Use both raw and diff features X_cols = h_features + a_features + diff_features print(f" {len(X_cols)} model features") # ── Walk-forward ───────────────────────────────────────────────────────────── print("\nWalk-forward training...") all_bets = [] for test_season in OOS_API_SEASONS: train = merged[merged['season'] < test_season].dropna(subset=X_cols) test = merged[merged['season'] == test_season].dropna(subset=X_cols) if len(train) < 500 or len(test) < 50: print(f" Season {test_season}: skipping (train={len(train)}, test={len(test)})") continue X_train = train[X_cols].values y_train = train['away_win'].values X_test = test[X_cols].values # Train calibrated RF rf = RandomForestClassifier(n_estimators=300, max_depth=12, min_samples_leaf=20, random_state=42, n_jobs=-1) cal_rf = CalibratedClassifierCV(rf, cv=5, method='isotonic') cal_rf.fit(X_train, y_train) # Predict pA = cal_rf.predict_proba(X_test)[:, 1] test = test.copy() test['model_pA'] = pA season_str = seasons_map[test_season] n_total = len(test) n_clv = (test['edge_A'] > CLV_THRESH).sum() n_model = (test['model_pA'] >= PA_THRESH).sum() # Combined filter: model pA >= threshold AND CLV > 5% combined = test[(test['model_pA'] >= PA_THRESH) & (test['edge_A'] > CLV_THRESH)] # Also track CLV-only and model-only for comparison clv_only = test[test['edge_A'] > CLV_THRESH] model_only = test[test['model_pA'] >= PA_THRESH] n_combined = len(combined) for _, r in combined.iterrows(): b = r['B365A'] - 1 p = r['p_A'] kelly_f = max((p * b - (1 - p)) / b, 0) if b > 0 else 0 if kelly_f <= 0: continue clv_pct = (r['B365A'] - r['PSC_fair_A']) / r['PSC_fair_A'] * 100 all_bets.append({ 'date': r['date_str'], 'season': season_str, 'league': league_names.get(r['div'], r['div']), 'league_code': r['div'], 'home_team': r['home_team'], 'away_team': r['away_team'], 'actual_result': r['FTR'], 'home_goals': int(r['FTHG']) if pd.notna(r.get('FTHG')) else '', 'away_goals': int(r['FTAG']) if pd.notna(r.get('FTAG')) else '', 'ht_home': int(r['HTHG']) if pd.notna(r.get('HTHG')) else '', 'ht_away': int(r['HTAG']) if pd.notna(r.get('HTAG')) else '', 'won': int(r['FTR'] == 'A'), 'odds': round(r['B365A'], 2), 'pinnacle_fair': round(r['PSC_fair_A'], 3), 'clv_pct': round(clv_pct, 2), 'model_pA': round(r['model_pA'], 4), 'kelly_fraction': round(kelly_f, 4), 'p_win_pinnacle': round(p, 4), }) # Season summary def flat_roi(df): if len(df) == 0: return 0, 0, 0 pls = [(r['B365A']-1) if r['FTR']=='A' else -1 for _, r in df.iterrows()] return len(pls), sum(pls)/len(pls)*100, df['away_win'].mean()*100 n_c, roi_c, wr_c = flat_roi(clv_only) n_m, roi_m, wr_m = flat_roi(model_only) n_b, roi_b, wr_b = flat_roi(combined) print(f"\n Season {season_str} ({n_total} matches):") print(f" CLV only (>5%): n={n_c:>4} WR={wr_c:5.1f}% ROI={roi_c:+6.1f}%") print(f" Model only (pA>={PA_THRESH}): n={n_m:>4} WR={wr_m:5.1f}% ROI={roi_m:+6.1f}%") print(f" Combined (CLV + model): n={n_b:>4} WR={wr_b:5.1f}% ROI={roi_b:+6.1f}%") bets = pd.DataFrame(all_bets).sort_values('date').reset_index(drop=True) print(f"\n{'='*70}") print(f"COMBINED STRATEGY: CLV >5% + RF model pA >= {PA_THRESH}") print(f"{'='*70}") print(f" Total OOS bets: {len(bets):,}") print(f" Date range: {bets['date'].min()} to {bets['date'].max()}") print(f" Win rate: {bets['won'].mean()*100:.1f}%") flat_pl = (bets['odds'] - 1).where(bets['won'] == 1, -1) print(f" Flat ROI: {flat_pl.mean()*100:+.2f}%") print(f" Total flat P&L: {flat_pl.sum():+.1f} units") print(f"\n By season:") for s, grp in bets.groupby('season'): fp = (grp['odds'] - 1).where(grp['won'] == 1, -1) print(f" {s} n={len(grp):>4} WR={grp['won'].mean()*100:.1f}% ROI={fp.mean()*100:+.1f}%") print(f"\n By league (min 10 bets):") lg = bets.groupby('league').agg(n=('won','count'), wr=('won','mean')).reset_index() lg['roi'] = bets.groupby('league').apply( lambda g: ((g['odds']-1).where(g['won']==1, -1)).mean()*100).values lg = lg[lg['n'] >= 10].sort_values('roi', ascending=False) for _, r in lg.iterrows(): print(f" {r['league']:<22} n={int(r['n']):>4} WR={r['wr']*100:.1f}% ROI={r['roi']:+.1f}%") # ── Kelly compounding ──────────────────────────────────────────────────────── print(f"\n Kelly compounding (from {START:,.0f}):") for km in KELLY_MULTS: bk = START; peak = START; max_dd = 0.0 for _, row in bets.iterrows(): f = min(row['kelly_fraction'] * km, 0.25) stake = bk * f pl = stake * (row['odds'] - 1) if row['won'] else -stake bk = max(bk + pl, 0) peak = max(peak, bk) max_dd = max(max_dd, (peak - bk) / peak) print(f" {km:.2f}x Kelly: {bk:>12,.2f} ({bk/START:.1f}x) MaxDD={max_dd*100:.1f}%") # ── Save bets ──────────────────────────────────────────────────────────────── out = r'E:\soccer\bets_rf_clv.csv' bets.to_csv(out, index=False) print(f"\nSaved to {out}") print("Done.")