Add backtest pipeline, betting_brain filters, score coherence + social v3
betting_brain.py: - HARD_MIN_SAMPLES=50 floor for calibrator bypass - ev_edge < 0 + >= 0.20 hard vetoes - BTTS muted (grid search found no profitable config) - Per-market optimal envelopes (MS, OU25) - Score coherence filter: main_pick must agree with score prediction - HTFT reversal cross-check for MS picks feature_builder.py / data_loader.py: - Real home/away_position from data (was hardcoded 10) - Cup detection wired into UpsetEngine - _estimate_league_position with 300-day season filter New scripts: - diagnostic_backtest.py: per-bet diagnostic backtest with loss patterns - optimize_filters.py: grid search per-market optimal thresholds - analyze_backtest_csv.py: root-cause hypothesis testing on CSV - compare_backtests.py: side-by-side validation with verdict - test_score_coherence.py: smoke test for coherence filter (20/20 pass) Reports: - diagnostic_backtest_20260525_024437 (50-match smoke) - diagnostic_backtest_20260525_035649 (1000-match in-sample) - filter_optimization_patch.json (grid search winners per market) Social poster v3: - satori + resvg HTML/CSS rendering pipeline - Twemoji football/basketball + flag SVGs - caption SEO: 12 curated hashtags per post - image SEO: descriptive filenames + .json metadata sidecar - /health, /preview-png, /run-now endpoints Docs: - mds/SESSION_HANDOFF.md: full session state for cross-machine continuity - mds/SOCIAL_POSTER_SETUP.md: API keys + test commands Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This commit is contained in:
@@ -0,0 +1,51 @@
|
||||
match_id,match_date,league_id,score_home,score_away,ht_score_home,ht_score_away,market,pick,odds,stake_units,playable,won,unit_profit,raw_confidence,calibrated_confidence,play_score,ev_edge,bet_grade,is_value_sniper,bb_score,bb_action,bb_vetoes,bb_issues,bb_positives,bb_model_prob,bb_implied_prob,bb_model_market_gap,bb_divergence,bb_trap_market,v27_consensus,data_quality_score,data_quality_flags,risk_level,odds_reliability,htft_reversal_prob,htft_top_pick,league_name,is_cup,model_version,decision_reason
|
||||
5iam9c9dw3ggz3y1ohr9uh53o,2026-05-24,8nbwkj392b0xzssqpw9jwmzdn,0,0,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.33,missing_full_ms_odds;lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
8c90p2ft4zxjdck8wlgq1a61g,2026-05-24,a9vrdkelbgif0gtu3wxsr75xo,2,2,0,0,OU25,Üst,1.41,0.2,True,True,0.082,59.5,65.4,58.4,-0.1557,B,True,26.4,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5955,0.7092,-0.1137,0.0339,True,DISAGREE,0.74,ai_features_inferred_from_history,MEDIUM,0.6618,,,,False,v28-pro-max,betting_brain_approved
|
||||
9ljz1grea3a8jajif4e9b7bpw,2026-05-24,2wolc27r8z03itcvwp43e38c5,1,1,1,1,BTTS,KG Var,1.62,0.2,True,True,0.124,53.7,69.9,48.4,-0.1476,B,True,1.6,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5371,0.6173,-0.0802,,True,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,HIGH,0.5592,,,,False,v28-pro-max,betting_brain_approved
|
||||
1r7iq0nhg2b674jpcm92ragpg,2026-05-24,1zp1du9n4rj36p1ss9zbxtqfb,4,0,1,0,BTTS,KG Var,1.89,0.2,True,False,-0.2,53.2,69.9,52.9,-0.0832,B,True,0.0,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5317,0.5291,0.0026,,True,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,HIGH,0.5353,,,,False,v28-pro-max,betting_brain_approved
|
||||
70dgok3yq76g076vemaps0178,2026-05-24,6lwpjhktjhl9g7x2w7njmzva6,2,1,1,0,BTTS,KG Var,1.74,0.2,True,True,0.148,51.9,68.4,60.4,-0.1255,B,True,15.7,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5192,0.5747,-0.0555,,True,AGREE,0.74,ai_features_inferred_from_history,MEDIUM,0.6164,,,,False,v28-pro-max,betting_brain_approved
|
||||
72q9d4uimmby6g6bor26taz9w,2026-05-24,6jgwiu2gq3dllmrwt45pfdn2z,2,0,2,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.555,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,LOW,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
67dd2t7043kv0yw1zj1buwdg4,2026-05-24,8n9w0n3i9kk05echhtmstn6o9,1,1,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.698,missing_full_ms_odds;lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
ejdwfph35q57phtfz7jr8st1w,2026-05-24,8ey0ww2zsosdmwr8ehsorh6t7,0,2,0,0,BTTS,KG Var,1.82,0.2,True,False,-0.2,53.2,69.9,63.5,-0.1097,B,True,12.1,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5317,0.5495,-0.0178,,True,DISAGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.51,,X/X,,False,v28-pro-max,betting_brain_approved
|
||||
68y8tlfnilw5trs1oqi4dhfkk,2026-05-24,8n9w0n3i9kk05echhtmstn6o9,2,2,1,1,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,DISAGREE,0.535,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
68fghiwdtwspk1m0ft5mspzx0,2026-05-24,8n9w0n3i9kk05echhtmstn6o9,1,2,1,2,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,DISAGREE,0.602,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
a1kqq0ggywfl6sl4srntxy2hg,2026-05-24,3ww12jab49q8q8mk9avdwjqgk,1,1,1,0,OU25,Üst,1.62,0.2,True,False,-0.2,61.4,65.4,60.6,-0.148,B,True,27.0,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6144,0.6173,-0.0029,0.0854,True,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5961,,,,False,v28-pro-max,betting_brain_approved
|
||||
a1vpp4i6t61v7qm3dfy6iuj2s,2026-05-24,3ww12jab49q8q8mk9avdwjqgk,2,1,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.74,missing_full_ms_odds;live_match_pre_match_features;ai_features_inferred_from_history,LOW,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
28rk26zqah60qj7h78qh5beac,2026-05-24,663a54fmymndjeev47qm7d3nf,2,2,2,1,OU25,Üst,1.39,0.2,True,True,0.078,61.1,65.4,57.5,-0.2039,B,True,13.3,BET,,low_reliability_league;inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6107,0.7194,-0.1087,0.0577,True,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,MEDIUM,0.3522,,,,False,v28-pro-max,betting_brain_approved
|
||||
a26stt54g5ju5ecodpcbcqxw4,2026-05-24,3ww12jab49q8q8mk9avdwjqgk,2,2,1,1,BTTS,KG Var,2.05,0.2,True,True,0.21,53.7,69.9,70.6,0.0354,B,True,33.5,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;usable_historical_sample,0.5371,0.4878,0.0493,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5961,,,,False,v28-pro-max,betting_brain_approved
|
||||
318rrnl9b1vlenjd9hwahpvro,2026-05-24,9z5643nd06afqu01ea2wt8y4g,1,0,0,0,DC,1X,1.12,0.0,False,True,0.0,73.5,73.5,72.9,-0.0553,PASS,True,41.8,WATCH_NO_VALUE,odds_below_minimum,base_model_not_playable;inferred_statistical_features,value_sniper_override;v25_v27_aligned;triple_value_confirmed;usable_historical_sample,0.7351,0.8929,-0.1578,0.031,False,AGREE,0.51,lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,0.8734,,,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
65tlfs3m6sc70261z37i90jys,2026-05-24,89ovpy1rarewwzqvi30bfdr8b,4,3,1,1,OU25,Üst,1.36,0.2,True,True,0.072,60.2,65.4,53.5,-0.2107,B,True,24.8,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6018,0.7353,-0.1335,0.0982,True,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.7068,,,,False,v28-pro-max,betting_brain_approved
|
||||
ei6nilo5074tnb17kvv37oy6s,2026-05-24,3j81qr7yc4gdnakfwnxf95ovh,1,1,1,1,MS,2,1.11,0.0,False,False,0.0,75.5,90.9,66.9,-0.0497,PASS,True,0.0,WATCH_NO_VALUE,odds_below_minimum,base_model_not_playable;inferred_statistical_features;v25_v27_soft_disagreement;triple_value_not_confirmed,value_sniper_override;strong_historical_sample,0.7554,0.9009,-0.1455,0.204,False,AGREE,0.51,lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,0.8771,0.0,,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
cx8nb7w3nmell38mta1umh2qc,2026-05-24,54c65mhi143utomzvvv3q2avh,0,2,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.726,missing_full_ms_odds;lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
53348iqniod61z7xurb4tx250,2026-05-24,477yyajzheg2z8u7uick0e13e,2,0,0,0,OU25,Üst,1.93,0.2,True,False,-0.2,59.5,65.4,71.3,0.0414,B,True,45.8,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5955,0.5181,0.0774,0.1084,False,AGREE,0.74,lineup_probable_not_confirmed;ai_features_inferred_from_history,MEDIUM,0.4706,,,,False,v28-pro-max,betting_brain_approved
|
||||
2hw717s7fxi2v2w53kdsplhqs,2026-05-24,8najqkluatpaxvqws78b9s17c,1,1,0,0,DC,12,1.24,0.0,False,False,0.0,70.0,70.0,61.2,-0.12,PASS,True,0.0,WATCH_NO_VALUE,odds_below_minimum,base_model_not_playable;inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,value_sniper_override;v25_v27_aligned;strong_historical_sample,0.7001,0.8065,-0.1064,0.0075,True,AGREE,0.51,lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,HIGH,0.5082,,,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
1q1s55dy4d4z4gs34qk6vx9n8,2026-05-24,cegl2ivkc25blcatxp4jmk1ec,1,0,1,0,OU25,Üst,2.03,0.2,True,False,-0.2,50.0,57.7,54.3,0.0546,B,True,35.7,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5,0.4926,0.0074,0.0555,False,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,HIGH,0.5993,,,,False,v28-pro-max,betting_brain_approved
|
||||
648u8zd49cwcxpspmvinlmexg,2026-05-24,1eruend45vd20g9hbrpiggs5u,1,0,0,0,MS,1,1.32,0.2,True,True,0.064,63.1,65.5,55.7,-0.0828,B,True,39.4,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6313,0.7576,-0.1263,0.0683,False,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,MEDIUM,0.8083,0.0,,,False,v28-pro-max,betting_brain_approved
|
||||
1psnufak57w8dfs9e5cvbmgwk,2026-05-24,cegl2ivkc25blcatxp4jmk1ec,3,0,2,0,BTTS,KG Var,1.94,0.2,True,False,-0.2,53.7,69.9,53.1,-0.0765,B,True,5.3,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5371,0.5155,0.0216,,True,AGREE,0.74,lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,HIGH,0.5993,,,,False,v28-pro-max,betting_brain_approved
|
||||
921kqaviappxt0w1kfmq1ek2c,2026-05-24,byu00jvt1j6csyv4y1lkt2fm2,1,0,1,0,DC,X2,1.18,0.0,False,False,0.0,75.8,75.8,74.9,-0.1267,PASS,True,0.0,WATCH_NO_VALUE,odds_below_minimum,inferred_statistical_features;v25_v27_soft_disagreement;trap_market_market_overpriced;triple_value_not_confirmed;engine_consensus_disagree,base_model_playable;value_sniper_override;strong_historical_sample,0.7584,0.8475,-0.0891,0.1783,True,DISAGREE,0.74,ai_features_inferred_from_history,MEDIUM,0.5359,,,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
3m11hvh2fzailt3ykd0uhzz84,2026-05-24,54c65mhi143utomzvvv3q2avh,0,0,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.591,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
7kvvf6blnps2xk15100ccdedw,2026-05-24,4zwgbb66rif2spcoeeol2motx,5,0,3,0,BTTS,KG Var,1.33,0.2,True,False,-0.2,57.1,69.9,62.3,-0.1512,B,True,29.1,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5707,0.7519,-0.1812,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.6995,,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
7liir8zj32o7m2udr7cknb8d0,2026-05-24,4zwgbb66rif2spcoeeol2motx,3,0,2,0,OU25,Üst,1.33,0.2,True,True,0.066,61.4,65.4,58.3,-0.1437,B,True,27.0,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6144,0.7519,-0.1375,0.0279,True,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.6995,,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
7l74ilyz7olljclexvn8tbjtg,2026-05-24,4zwgbb66rif2spcoeeol2motx,5,1,4,0,BTTS,KG Var,1.55,0.2,True,True,0.11,57.1,69.9,68.2,-0.0873,B,True,32.7,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5707,0.6452,-0.0745,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.6995,,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
8f6gex4eh119d2hh9y2zb5clw,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,2,0,2,0,OU25,Üst,1.49,0.2,True,False,-0.2,66.5,65.4,77.9,0.0081,B,True,50.1,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6651,0.6711,-0.006,0.1144,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,LOW,0.5033,,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
8ee7ipt4u6kyk6baueedsdafo,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,0,2,0,2,BTTS,KG Var,1.69,0.2,True,False,-0.2,54.3,69.9,64.7,-0.0792,B,True,27.6,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5433,0.5917,-0.0484,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,,2/2,,False,v28-pro-max,betting_brain_approved
|
||||
8fydg367drpc25r1bobxqj3f8,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,3,1,2,1,OU25,Üst,1.61,0.2,True,True,0.122,50.1,57.7,50.7,-0.1072,B,True,35.0,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5009,0.6211,-0.1202,0.0553,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,,,,False,v28-pro-max,betting_brain_approved
|
||||
8fkdhce1peguwgnsunwoln3f8,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,0,2,0,2,OU25,Üst,1.24,0.0,False,False,0.0,61.4,65.4,61.1,-0.1548,PASS,True,16.8,WATCH_NO_VALUE,odds_below_minimum,inferred_statistical_features;triple_value_not_confirmed;historical_sample_too_low,base_model_playable;value_sniper_override;v25_v27_aligned,0.6144,0.8065,-0.1921,0.0942,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,LOW,0.5033,,,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
9g5hqtjja6ceqhkpghwmoy6ms,2026-05-24,2y8bntiif3a9y6gtmauv30gt,2,0,1,0,OU25,Üst,1.71,0.2,True,False,-0.2,50.1,57.7,52.6,-0.0794,B,True,36.5,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5009,0.5848,-0.0839,0.0481,False,DISAGREE,0.74,ai_features_inferred_from_history,MEDIUM,0.4782,,,,False,v28-pro-max,betting_brain_approved
|
||||
8h6429zr5ijqcxc8gjxygjtw4,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,3,0,1,0,MS,1,1.33,0.2,True,True,0.066,66.1,65.5,66.7,-0.1155,B,True,0.0,BET,,inferred_statistical_features;v25_v27_soft_disagreement;trap_market_market_overpriced;triple_value_not_confirmed;htft_reversal_prob_minor=0.11,base_model_playable;value_sniper_override;strong_historical_sample,0.6614,0.7519,-0.0905,0.2583,True,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,0.0814,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
77knm2ibdtb7akzrbltwz7axg,2026-05-24,bly7ema5au6j40i0grhl0pnub,1,1,1,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.726,missing_full_ms_odds;lineup_probable_not_confirmed;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
8es4680yd87gtmomg2jk3isyc,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,0,1,0,0,OU25,Üst,1.53,0.2,True,False,-0.2,59.5,65.4,67.8,-0.0259,B,True,42.4,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5955,0.6536,-0.0581,0.0713,False,DISAGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,,2/2,,False,v28-pro-max,betting_brain_approved
|
||||
8dmcz3k1u4ze53nvrsoz7eoes,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,1,1,1,0,BTTS,KG Var,1.26,0.0,False,True,0.0,54.9,69.9,57.6,-0.2005,PASS,True,3.3,WATCH_NO_VALUE,odds_below_minimum,inferred_statistical_features;triple_value_not_confirmed;historical_sample_too_low,base_model_playable;value_sniper_override,0.5488,0.7937,-0.2449,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,LOW,0.5033,,1/1,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
8gcbai6m7v7o8piqfram4qe50,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,3,1,2,0,HTFT,1/1,4.59,0.0,False,True,0.0,27.3,27.3,24.6,0.1657,PASS,True,0.0,REJECT,calibrated_confidence_too_low;play_score_too_low;volatile_market_requires_exceptional_evidence,inferred_statistical_features;historical_sample_too_low,base_model_playable,0.2734,0.2179,0.0555,,False,DISAGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,,1/1,,False,v28-pro-max,betting_brain_no_safe_pick
|
||||
3azy3enp78au0zfugc3l1yf4k,2026-05-24,54c65mhi143utomzvvv3q2avh,2,0,1,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.532,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
1d2fb7bt5f8xy5on24w1kj1g4,2026-05-24,54c65mhi143utomzvvv3q2avh,1,0,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.532,missing_full_ms_odds;lineup_probable_not_confirmed;lineup_projection_low_confidence;lineup_incomplete;missing_referee;ai_features_inferred_from_history,LOW,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
pw01xm8v3jlz13fpi3zq0ftg,2026-05-24,3umprqta6ipyann6qjjh07biz,1,1,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.33,missing_full_ms_odds;lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
mjo9k4zr1x884vjlwea2y1hw,2026-05-24,3umprqta6ipyann6qjjh07biz,1,0,1,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.33,missing_full_ms_odds;lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
8d8fm7wli7tfx8hm9w5l8nuhg,2026-05-24,3is4bkgf3loxv9qfg3hm8zfqb,1,1,1,0,BTTS,KG Var,1.72,0.2,True,True,0.144,53.7,69.9,65.2,-0.0712,B,True,28.0,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5371,0.5814,-0.0443,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.5033,,,,False,v28-pro-max,betting_brain_approved
|
||||
oqsq3f0kvic8xfed8dp302z8,2026-05-24,3umprqta6ipyann6qjjh07biz,3,2,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.33,missing_full_ms_odds;lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
o7tn4si7fxvq9c2mg0xs48wk,2026-05-24,3umprqta6ipyann6qjjh07biz,0,1,0,0,,,,1.0,False,,0.0,,,,,,False,,,,,,,,,,False,AGREE,0.33,missing_full_ms_odds;lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,,,,,False,v28-pro-max,no_bet_conditions_met
|
||||
eh9jfegscokidyczxfq691990,2026-05-24,3j81qr7yc4gdnakfwnxf95ovh,2,3,0,1,OU25,Üst,1.44,0.2,True,True,0.088,50.1,57.7,32.9,-0.2537,B,True,17.0,BET,,inferred_statistical_features;trap_market_market_overpriced;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.5009,0.6944,-0.1935,0.0596,True,AGREE,0.51,lineup_unavailable;lineup_incomplete;missing_referee;ai_features_inferred_from_history,MEDIUM,0.8771,,,,False,v28-pro-max,betting_brain_approved
|
||||
dkhhkbwnxwl47e8hybv89mwb8,2026-05-24,5jd0k2txwnq69frs79eulba8j,1,2,0,1,OU25,Üst,1.23,0.0,False,True,0.0,61.4,65.4,61.2,-0.1185,PASS,True,11.4,WATCH_NO_VALUE,odds_below_minimum,base_model_not_playable;inferred_statistical_features;triple_value_not_confirmed,value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6144,0.813,-0.1986,0.0179,False,AGREE,0.74,ai_features_inferred_from_history,LOW,0.9233,,1/1,,False,v28-pro-max,betting_brain_no_value_odds_below_minimum
|
||||
1lknqdz9vmb3hnqu144zkkefo,2026-05-24,1r097lpxe0xn03ihb7wi98kao,1,0,1,0,BTTS,KG Var,1.78,0.2,True,False,-0.2,50.0,61.7,55.6,-0.088,B,True,29.3,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5,0.5618,-0.0618,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,MEDIUM,0.7391,,1/1,,False,v28-pro-max,betting_brain_approved
|
||||
3oazp9kfbyyiatn246k4to6xg,2026-05-24,9ynnnx1qmkizq1o3qr3v0nsuk,1,2,0,1,BTTS,KG Var,1.36,0.2,True,True,0.072,53.7,69.9,61.2,-0.1571,B,True,33.7,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;strong_historical_sample,0.5371,0.7353,-0.1982,,False,AGREE,0.74,live_match_pre_match_features;ai_features_inferred_from_history,LOW,0.554,,2/2,,False,v28-pro-max,betting_brain_approved
|
||||
8cr8t6qh0r6g0mv6ftq0ic1sk,2026-05-24,a9vrdkelbgif0gtu3wxsr75xo,2,1,0,1,OU25,Üst,1.46,0.2,True,True,0.092,61.4,65.4,68.1,-0.0182,B,True,47.8,BET,,inferred_statistical_features;triple_value_not_confirmed,base_model_playable;value_sniper_override;v25_v27_aligned;strong_historical_sample,0.6144,0.6849,-0.0705,0.0535,False,AGREE,0.74,ai_features_inferred_from_history,MEDIUM,0.6618,,,,False,v28-pro-max,betting_brain_approved
|
||||
|
@@ -0,0 +1,220 @@
|
||||
{
|
||||
"args": {
|
||||
"days": 3,
|
||||
"max_matches": 50,
|
||||
"start": null,
|
||||
"end": null,
|
||||
"progress_interval": 50
|
||||
},
|
||||
"aggregate": {
|
||||
"overall": {
|
||||
"n_total": 50,
|
||||
"n_playable_settled": 27,
|
||||
"wins": 15,
|
||||
"losses": 12,
|
||||
"hit_rate_pct": 55.56,
|
||||
"unit_profit": -0.862,
|
||||
"staked": 5.4,
|
||||
"roi_pct": -15.96
|
||||
},
|
||||
"by_market": {
|
||||
"OU25": {
|
||||
"n_total": 13,
|
||||
"n_playable_settled": 13,
|
||||
"wins": 7,
|
||||
"losses": 6,
|
||||
"hit_rate_pct": 53.85,
|
||||
"unit_profit": -0.6,
|
||||
"staked": 2.6,
|
||||
"roi_pct": -23.08
|
||||
},
|
||||
"BTTS": {
|
||||
"n_total": 12,
|
||||
"n_playable_settled": 12,
|
||||
"wins": 6,
|
||||
"losses": 6,
|
||||
"hit_rate_pct": 50.0,
|
||||
"unit_profit": -0.392,
|
||||
"staked": 2.4,
|
||||
"roi_pct": -16.33
|
||||
},
|
||||
"MS": {
|
||||
"n_total": 2,
|
||||
"n_playable_settled": 2,
|
||||
"wins": 2,
|
||||
"losses": 0,
|
||||
"hit_rate_pct": 100.0,
|
||||
"unit_profit": 0.13,
|
||||
"staked": 0.4,
|
||||
"roi_pct": 32.5
|
||||
}
|
||||
},
|
||||
"by_confidence": {
|
||||
"65-70": {
|
||||
"n_total": 22,
|
||||
"n_playable_settled": 22,
|
||||
"wins": 13,
|
||||
"losses": 9,
|
||||
"hit_rate_pct": 59.09,
|
||||
"unit_profit": -0.472,
|
||||
"staked": 4.4,
|
||||
"roi_pct": -10.73
|
||||
},
|
||||
"55-60": {
|
||||
"n_total": 4,
|
||||
"n_playable_settled": 4,
|
||||
"wins": 2,
|
||||
"losses": 2,
|
||||
"hit_rate_pct": 50.0,
|
||||
"unit_profit": -0.19,
|
||||
"staked": 0.8,
|
||||
"roi_pct": -23.75
|
||||
},
|
||||
"60-65": {
|
||||
"n_total": 1,
|
||||
"n_playable_settled": 1,
|
||||
"wins": 0,
|
||||
"losses": 1,
|
||||
"hit_rate_pct": 0.0,
|
||||
"unit_profit": -0.2,
|
||||
"staked": 0.2,
|
||||
"roi_pct": -100.0
|
||||
}
|
||||
},
|
||||
"by_odds": {
|
||||
"1.3-1.5": {
|
||||
"n_total": 11,
|
||||
"n_playable_settled": 11,
|
||||
"wins": 9,
|
||||
"losses": 2,
|
||||
"hit_rate_pct": 81.82,
|
||||
"unit_profit": 0.28,
|
||||
"staked": 2.2,
|
||||
"roi_pct": 12.73
|
||||
},
|
||||
"1.5-1.8": {
|
||||
"n_total": 10,
|
||||
"n_playable_settled": 10,
|
||||
"wins": 5,
|
||||
"losses": 5,
|
||||
"hit_rate_pct": 50.0,
|
||||
"unit_profit": -0.352,
|
||||
"staked": 2.0,
|
||||
"roi_pct": -17.6
|
||||
},
|
||||
"1.8-2.2": {
|
||||
"n_total": 6,
|
||||
"n_playable_settled": 6,
|
||||
"wins": 1,
|
||||
"losses": 5,
|
||||
"hit_rate_pct": 16.67,
|
||||
"unit_profit": -0.79,
|
||||
"staked": 1.2,
|
||||
"roi_pct": -65.83
|
||||
}
|
||||
},
|
||||
"by_grade": {
|
||||
"B": {
|
||||
"n_total": 27,
|
||||
"n_playable_settled": 27,
|
||||
"wins": 15,
|
||||
"losses": 12,
|
||||
"hit_rate_pct": 55.56,
|
||||
"unit_profit": -0.862,
|
||||
"staked": 5.4,
|
||||
"roi_pct": -15.96
|
||||
}
|
||||
},
|
||||
"by_competition": {
|
||||
"league": {
|
||||
"n_total": 27,
|
||||
"n_playable_settled": 27,
|
||||
"wins": 15,
|
||||
"losses": 12,
|
||||
"hit_rate_pct": 55.56,
|
||||
"unit_profit": -0.862,
|
||||
"staked": 5.4,
|
||||
"roi_pct": -15.96
|
||||
}
|
||||
}
|
||||
},
|
||||
"loss_diagnostics": {
|
||||
"n_losses": 12,
|
||||
"total_loss_units": -2.4,
|
||||
"patterns": {
|
||||
"high_htft_reversal_prob (>=0.20)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"cup_match": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"low_league_reliability (<0.45)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"v27_disagree": [
|
||||
3,
|
||||
25.0
|
||||
],
|
||||
"trap_market_flagged": [
|
||||
4,
|
||||
33.33
|
||||
],
|
||||
"low_calibrated_conf (<55)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"high_odds_underdog (>=2.5)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"low_data_quality (<0.55)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"high_risk_level": [
|
||||
3,
|
||||
25.0
|
||||
],
|
||||
"inferred_features": [
|
||||
12,
|
||||
100.0
|
||||
]
|
||||
},
|
||||
"by_market": [
|
||||
[
|
||||
"BTTS",
|
||||
6
|
||||
],
|
||||
[
|
||||
"OU25",
|
||||
6
|
||||
]
|
||||
],
|
||||
"by_league": [
|
||||
[
|
||||
null,
|
||||
12
|
||||
]
|
||||
],
|
||||
"top_bb_issues_in_losses": [
|
||||
[
|
||||
"inferred_statistical_features",
|
||||
12
|
||||
],
|
||||
[
|
||||
"triple_value_not_confirmed",
|
||||
12
|
||||
],
|
||||
[
|
||||
"trap_market_market_overpriced",
|
||||
4
|
||||
]
|
||||
],
|
||||
"top_bb_vetoes_in_losses": []
|
||||
},
|
||||
"recommendations": [],
|
||||
"errors_sample": []
|
||||
}
|
||||
@@ -0,0 +1,71 @@
|
||||
==============================================================================
|
||||
DIAGNOSTIC BACKTEST REPORT
|
||||
==============================================================================
|
||||
Generated: 2026-05-25T02:44:37
|
||||
Sample window: start=-3d, end=now
|
||||
Max matches: 50
|
||||
Excluded days: ['2026-04-29', '2026-05-03']
|
||||
|
||||
OVERALL
|
||||
------------------------------------------------------------------------------
|
||||
n_total : 50
|
||||
n_playable_settled : 27
|
||||
wins : 15
|
||||
losses : 12
|
||||
hit_rate_pct : 55.56
|
||||
unit_profit : -0.862
|
||||
staked : 5.4
|
||||
roi_pct : -15.96
|
||||
|
||||
PER MARKET
|
||||
------------------------------------------------------------------------------
|
||||
market n hit% profit roi%
|
||||
OU25 13 53.85 -0.6 -23.08
|
||||
BTTS 12 50.0 -0.392 -16.33
|
||||
MS 2 100.0 0.13 32.5
|
||||
|
||||
PER CALIBRATED CONFIDENCE BAND
|
||||
------------------------------------------------------------------------------
|
||||
band n hit% roi%
|
||||
55-60 4 50.0 -23.75
|
||||
60-65 1 0.0 -100.0
|
||||
65-70 22 59.09 -10.73
|
||||
|
||||
PER ODDS BAND
|
||||
------------------------------------------------------------------------------
|
||||
band n hit% roi%
|
||||
1.3-1.5 11 81.82 12.73
|
||||
1.5-1.8 10 50.0 -17.6
|
||||
1.8-2.2 6 16.67 -65.83
|
||||
|
||||
LEAGUE vs CUP
|
||||
------------------------------------------------------------------------------
|
||||
league n= 27 hit=55.56% roi=-15.96%
|
||||
|
||||
LOSS DIAGNOSTICS
|
||||
------------------------------------------------------------------------------
|
||||
total losses: 12
|
||||
total lost units: -2.4
|
||||
By market: [('BTTS', 6), ('OU25', 6)]
|
||||
Loss patterns (count, % of losses):
|
||||
high_htft_reversal_prob (>=0.20) 0 (0.0%)
|
||||
cup_match 0 (0.0%)
|
||||
low_league_reliability (<0.45) 0 (0.0%)
|
||||
v27_disagree 3 (25.0%)
|
||||
trap_market_flagged 4 (33.33%)
|
||||
low_calibrated_conf (<55) 0 (0.0%)
|
||||
high_odds_underdog (>=2.5) 0 (0.0%)
|
||||
low_data_quality (<0.55) 0 (0.0%)
|
||||
high_risk_level 3 (25.0%)
|
||||
inferred_features 12 (100.0%)
|
||||
Top betting_brain issues seen in losses:
|
||||
inferred_statistical_features 12
|
||||
triple_value_not_confirmed 12
|
||||
trap_market_market_overpriced 4
|
||||
Top betting_brain vetoes (in losses — i.e. veto fired but bet still went through value-sniper override):
|
||||
|
||||
RECOMMENDATIONS
|
||||
------------------------------------------------------------------------------
|
||||
(none surfaced — sample too small or no clear pattern)
|
||||
|
||||
==============================================================================
|
||||
File diff suppressed because it is too large
Load Diff
@@ -0,0 +1,335 @@
|
||||
{
|
||||
"args": {
|
||||
"days": 14,
|
||||
"max_matches": 1000,
|
||||
"start": null,
|
||||
"end": null,
|
||||
"progress_interval": 50
|
||||
},
|
||||
"aggregate": {
|
||||
"overall": {
|
||||
"n_total": 1000,
|
||||
"n_playable_settled": 524,
|
||||
"wins": 287,
|
||||
"losses": 237,
|
||||
"hit_rate_pct": 54.77,
|
||||
"unit_profit": -17.897,
|
||||
"staked": 107.0,
|
||||
"roi_pct": -16.73
|
||||
},
|
||||
"by_market": {
|
||||
"OU25": {
|
||||
"n_total": 236,
|
||||
"n_playable_settled": 236,
|
||||
"wins": 134,
|
||||
"losses": 102,
|
||||
"hit_rate_pct": 56.78,
|
||||
"unit_profit": -6.271,
|
||||
"staked": 48.5,
|
||||
"roi_pct": -12.93
|
||||
},
|
||||
"BTTS": {
|
||||
"n_total": 205,
|
||||
"n_playable_settled": 205,
|
||||
"wins": 105,
|
||||
"losses": 100,
|
||||
"hit_rate_pct": 51.22,
|
||||
"unit_profit": -8.89,
|
||||
"staked": 41.1,
|
||||
"roi_pct": -21.63
|
||||
},
|
||||
"MS": {
|
||||
"n_total": 76,
|
||||
"n_playable_settled": 76,
|
||||
"wins": 44,
|
||||
"losses": 32,
|
||||
"hit_rate_pct": 57.89,
|
||||
"unit_profit": -2.396,
|
||||
"staked": 16.0,
|
||||
"roi_pct": -14.98
|
||||
},
|
||||
"OU35": {
|
||||
"n_total": 3,
|
||||
"n_playable_settled": 3,
|
||||
"wins": 0,
|
||||
"losses": 3,
|
||||
"hit_rate_pct": 0.0,
|
||||
"unit_profit": -0.6,
|
||||
"staked": 0.6,
|
||||
"roi_pct": -100.0
|
||||
},
|
||||
"DC": {
|
||||
"n_total": 4,
|
||||
"n_playable_settled": 4,
|
||||
"wins": 4,
|
||||
"losses": 0,
|
||||
"hit_rate_pct": 100.0,
|
||||
"unit_profit": 0.26,
|
||||
"staked": 0.8,
|
||||
"roi_pct": 32.5
|
||||
}
|
||||
},
|
||||
"by_confidence": {
|
||||
"65-70": {
|
||||
"n_total": 420,
|
||||
"n_playable_settled": 420,
|
||||
"wins": 233,
|
||||
"losses": 187,
|
||||
"hit_rate_pct": 55.48,
|
||||
"unit_profit": -14.057,
|
||||
"staked": 85.1,
|
||||
"roi_pct": -16.52
|
||||
},
|
||||
"60-65": {
|
||||
"n_total": 33,
|
||||
"n_playable_settled": 33,
|
||||
"wins": 16,
|
||||
"losses": 17,
|
||||
"hit_rate_pct": 48.48,
|
||||
"unit_profit": -1.61,
|
||||
"staked": 6.6,
|
||||
"roi_pct": -24.39
|
||||
},
|
||||
"55-60": {
|
||||
"n_total": 52,
|
||||
"n_playable_settled": 52,
|
||||
"wins": 28,
|
||||
"losses": 24,
|
||||
"hit_rate_pct": 53.85,
|
||||
"unit_profit": -0.668,
|
||||
"staked": 10.5,
|
||||
"roi_pct": -6.36
|
||||
},
|
||||
"50-55": {
|
||||
"n_total": 5,
|
||||
"n_playable_settled": 5,
|
||||
"wins": 2,
|
||||
"losses": 3,
|
||||
"hit_rate_pct": 40.0,
|
||||
"unit_profit": -0.64,
|
||||
"staked": 1.3,
|
||||
"roi_pct": -49.23
|
||||
},
|
||||
"45-50": {
|
||||
"n_total": 8,
|
||||
"n_playable_settled": 8,
|
||||
"wins": 4,
|
||||
"losses": 4,
|
||||
"hit_rate_pct": 50.0,
|
||||
"unit_profit": -0.382,
|
||||
"staked": 1.9,
|
||||
"roi_pct": -20.11
|
||||
},
|
||||
"70-80": {
|
||||
"n_total": 6,
|
||||
"n_playable_settled": 6,
|
||||
"wins": 4,
|
||||
"losses": 2,
|
||||
"hit_rate_pct": 66.67,
|
||||
"unit_profit": -0.54,
|
||||
"staked": 1.6,
|
||||
"roi_pct": -33.75
|
||||
}
|
||||
},
|
||||
"by_odds": {
|
||||
"1.3-1.5": {
|
||||
"n_total": 241,
|
||||
"n_playable_settled": 241,
|
||||
"wins": 148,
|
||||
"losses": 93,
|
||||
"hit_rate_pct": 61.41,
|
||||
"unit_profit": -7.408,
|
||||
"staked": 49.0,
|
||||
"roi_pct": -15.12
|
||||
},
|
||||
"1.5-1.8": {
|
||||
"n_total": 221,
|
||||
"n_playable_settled": 221,
|
||||
"wins": 115,
|
||||
"losses": 106,
|
||||
"hit_rate_pct": 52.04,
|
||||
"unit_profit": -6.926,
|
||||
"staked": 44.3,
|
||||
"roi_pct": -15.63
|
||||
},
|
||||
"1.8-2.2": {
|
||||
"n_total": 56,
|
||||
"n_playable_settled": 56,
|
||||
"wins": 23,
|
||||
"losses": 33,
|
||||
"hit_rate_pct": 41.07,
|
||||
"unit_profit": -2.789,
|
||||
"staked": 12.2,
|
||||
"roi_pct": -22.86
|
||||
},
|
||||
"2.2-3.0": {
|
||||
"n_total": 5,
|
||||
"n_playable_settled": 5,
|
||||
"wins": 1,
|
||||
"losses": 4,
|
||||
"hit_rate_pct": 20.0,
|
||||
"unit_profit": -0.574,
|
||||
"staked": 1.3,
|
||||
"roi_pct": -44.15
|
||||
},
|
||||
"3.0-5.0": {
|
||||
"n_total": 1,
|
||||
"n_playable_settled": 1,
|
||||
"wins": 0,
|
||||
"losses": 1,
|
||||
"hit_rate_pct": 0.0,
|
||||
"unit_profit": -0.2,
|
||||
"staked": 0.2,
|
||||
"roi_pct": -100.0
|
||||
}
|
||||
},
|
||||
"by_grade": {
|
||||
"B": {
|
||||
"n_total": 518,
|
||||
"n_playable_settled": 518,
|
||||
"wins": 285,
|
||||
"losses": 233,
|
||||
"hit_rate_pct": 55.02,
|
||||
"unit_profit": -16.931,
|
||||
"staked": 105.3,
|
||||
"roi_pct": -16.08
|
||||
},
|
||||
"A": {
|
||||
"n_total": 6,
|
||||
"n_playable_settled": 6,
|
||||
"wins": 2,
|
||||
"losses": 4,
|
||||
"hit_rate_pct": 33.33,
|
||||
"unit_profit": -0.966,
|
||||
"staked": 1.7,
|
||||
"roi_pct": -56.82
|
||||
}
|
||||
},
|
||||
"by_competition": {
|
||||
"league": {
|
||||
"n_total": 524,
|
||||
"n_playable_settled": 524,
|
||||
"wins": 287,
|
||||
"losses": 237,
|
||||
"hit_rate_pct": 54.77,
|
||||
"unit_profit": -17.897,
|
||||
"staked": 107.0,
|
||||
"roi_pct": -16.73
|
||||
}
|
||||
}
|
||||
},
|
||||
"loss_diagnostics": {
|
||||
"n_losses": 237,
|
||||
"total_loss_units": -48.7,
|
||||
"patterns": {
|
||||
"high_htft_reversal_prob (>=0.20)": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"cup_match": [
|
||||
0,
|
||||
0.0
|
||||
],
|
||||
"low_league_reliability (<0.45)": [
|
||||
42,
|
||||
17.72
|
||||
],
|
||||
"v27_disagree": [
|
||||
60,
|
||||
25.32
|
||||
],
|
||||
"trap_market_flagged": [
|
||||
81,
|
||||
34.18
|
||||
],
|
||||
"low_calibrated_conf (<55)": [
|
||||
7,
|
||||
2.95
|
||||
],
|
||||
"high_odds_underdog (>=2.5)": [
|
||||
4,
|
||||
1.69
|
||||
],
|
||||
"low_data_quality (<0.55)": [
|
||||
40,
|
||||
16.88
|
||||
],
|
||||
"high_risk_level": [
|
||||
20,
|
||||
8.44
|
||||
],
|
||||
"inferred_features": [
|
||||
0,
|
||||
0.0
|
||||
]
|
||||
},
|
||||
"by_market": [
|
||||
[
|
||||
"OU25",
|
||||
102
|
||||
],
|
||||
[
|
||||
"BTTS",
|
||||
100
|
||||
],
|
||||
[
|
||||
"MS",
|
||||
32
|
||||
],
|
||||
[
|
||||
"OU35",
|
||||
3
|
||||
]
|
||||
],
|
||||
"by_league": [
|
||||
[
|
||||
null,
|
||||
237
|
||||
]
|
||||
],
|
||||
"top_bb_issues_in_losses": [
|
||||
[
|
||||
"triple_value_not_confirmed",
|
||||
230
|
||||
],
|
||||
[
|
||||
"trap_market_market_overpriced",
|
||||
81
|
||||
],
|
||||
[
|
||||
"low_reliability_league",
|
||||
40
|
||||
],
|
||||
[
|
||||
"v25_v27_soft_disagreement",
|
||||
10
|
||||
],
|
||||
[
|
||||
"engine_consensus_disagree",
|
||||
5
|
||||
],
|
||||
[
|
||||
"historical_sample_too_low",
|
||||
3
|
||||
],
|
||||
[
|
||||
"very_low_reliability_league",
|
||||
2
|
||||
],
|
||||
[
|
||||
"htft_reversal_prob_minor=0.13",
|
||||
1
|
||||
]
|
||||
],
|
||||
"top_bb_vetoes_in_losses": []
|
||||
},
|
||||
"recommendations": [
|
||||
{
|
||||
"type": "raise_confidence_threshold",
|
||||
"confidence_band": "65-70",
|
||||
"evidence": "n=420, roi=-16.52%",
|
||||
"suggested_fix": "Raise MIN_BET_SCORE or market_min_conf above 65"
|
||||
}
|
||||
],
|
||||
"errors_sample": []
|
||||
}
|
||||
@@ -0,0 +1,86 @@
|
||||
==============================================================================
|
||||
DIAGNOSTIC BACKTEST REPORT
|
||||
==============================================================================
|
||||
Generated: 2026-05-25T03:56:49
|
||||
Sample window: start=-14d, end=now
|
||||
Max matches: 1000
|
||||
Excluded days: ['2026-04-29', '2026-05-03']
|
||||
|
||||
OVERALL
|
||||
------------------------------------------------------------------------------
|
||||
n_total : 1000
|
||||
n_playable_settled : 524
|
||||
wins : 287
|
||||
losses : 237
|
||||
hit_rate_pct : 54.77
|
||||
unit_profit : -17.897
|
||||
staked : 107.0
|
||||
roi_pct : -16.73
|
||||
|
||||
PER MARKET
|
||||
------------------------------------------------------------------------------
|
||||
market n hit% profit roi%
|
||||
OU25 236 56.78 -6.271 -12.93
|
||||
BTTS 205 51.22 -8.89 -21.63
|
||||
MS 76 57.89 -2.396 -14.98
|
||||
DC 4 100.0 0.26 32.5
|
||||
OU35 3 0.0 -0.6 -100.0
|
||||
|
||||
PER CALIBRATED CONFIDENCE BAND
|
||||
------------------------------------------------------------------------------
|
||||
band n hit% roi%
|
||||
45-50 8 50.0 -20.11
|
||||
50-55 5 40.0 -49.23
|
||||
55-60 52 53.85 -6.36
|
||||
60-65 33 48.48 -24.39
|
||||
65-70 420 55.48 -16.52
|
||||
70-80 6 66.67 -33.75
|
||||
|
||||
PER ODDS BAND
|
||||
------------------------------------------------------------------------------
|
||||
band n hit% roi%
|
||||
1.3-1.5 241 61.41 -15.12
|
||||
1.5-1.8 221 52.04 -15.63
|
||||
1.8-2.2 56 41.07 -22.86
|
||||
2.2-3.0 5 20.0 -44.15
|
||||
3.0-5.0 1 0.0 -100.0
|
||||
|
||||
LEAGUE vs CUP
|
||||
------------------------------------------------------------------------------
|
||||
league n= 524 hit=54.77% roi=-16.73%
|
||||
|
||||
LOSS DIAGNOSTICS
|
||||
------------------------------------------------------------------------------
|
||||
total losses: 237
|
||||
total lost units: -48.7
|
||||
By market: [('OU25', 102), ('BTTS', 100), ('MS', 32), ('OU35', 3)]
|
||||
Loss patterns (count, % of losses):
|
||||
high_htft_reversal_prob (>=0.20) 0 (0.0%)
|
||||
cup_match 0 (0.0%)
|
||||
low_league_reliability (<0.45) 42 (17.72%)
|
||||
v27_disagree 60 (25.32%)
|
||||
trap_market_flagged 81 (34.18%)
|
||||
low_calibrated_conf (<55) 7 (2.95%)
|
||||
high_odds_underdog (>=2.5) 4 (1.69%)
|
||||
low_data_quality (<0.55) 40 (16.88%)
|
||||
high_risk_level 20 (8.44%)
|
||||
inferred_features 0 (0.0%)
|
||||
Top betting_brain issues seen in losses:
|
||||
triple_value_not_confirmed 230
|
||||
trap_market_market_overpriced 81
|
||||
low_reliability_league 40
|
||||
v25_v27_soft_disagreement 10
|
||||
engine_consensus_disagree 5
|
||||
historical_sample_too_low 3
|
||||
very_low_reliability_league 2
|
||||
htft_reversal_prob_minor=0.13 1
|
||||
Top betting_brain vetoes (in losses — i.e. veto fired but bet still went through value-sniper override):
|
||||
|
||||
RECOMMENDATIONS
|
||||
------------------------------------------------------------------------------
|
||||
• [raise_confidence_threshold]
|
||||
confidence_band: 65-70
|
||||
evidence: n=420, roi=-16.52%
|
||||
suggested_fix: Raise MIN_BET_SCORE or market_min_conf above 65
|
||||
|
||||
==============================================================================
|
||||
@@ -0,0 +1,38 @@
|
||||
{
|
||||
"BTTS": {
|
||||
"min_calibrated_confidence": 65,
|
||||
"min_ev_edge": -1.0,
|
||||
"max_ev_edge": 0.1,
|
||||
"min_odds": 1.4,
|
||||
"max_odds": 10.0,
|
||||
"min_odds_reliability": 0.55,
|
||||
"require_v27_agree": true,
|
||||
"expected_n_bets": 54,
|
||||
"expected_hit_pct": 55.56,
|
||||
"expected_roi_pct": -10.96
|
||||
},
|
||||
"MS": {
|
||||
"min_calibrated_confidence": 0,
|
||||
"min_ev_edge": -0.05,
|
||||
"max_ev_edge": 0.15,
|
||||
"min_odds": 1.2,
|
||||
"max_odds": 10.0,
|
||||
"min_odds_reliability": 0.0,
|
||||
"require_v27_agree": true,
|
||||
"expected_n_bets": 21,
|
||||
"expected_hit_pct": 61.9,
|
||||
"expected_roi_pct": 8.23
|
||||
},
|
||||
"OU25": {
|
||||
"min_calibrated_confidence": 0,
|
||||
"min_ev_edge": -1.0,
|
||||
"max_ev_edge": 0.15,
|
||||
"min_odds": 1.8,
|
||||
"max_odds": 10.0,
|
||||
"min_odds_reliability": 0.0,
|
||||
"require_v27_agree": false,
|
||||
"expected_n_bets": 20,
|
||||
"expected_hit_pct": 65.0,
|
||||
"expected_roi_pct": 28.91
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,227 @@
|
||||
"""
|
||||
Deep root-cause analysis on diagnostic_backtest CSV.
|
||||
Tests specific hypotheses with hard numbers and proposes actionable
|
||||
filter rules with estimated impact (units saved, ROI shift).
|
||||
"""
|
||||
|
||||
import sys, os, glob
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
|
||||
REPORTS_DIR = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "reports")
|
||||
|
||||
def latest_csv():
|
||||
files = sorted(glob.glob(os.path.join(REPORTS_DIR, "diagnostic_backtest_*.csv")),
|
||||
key=os.path.getmtime, reverse=True)
|
||||
return files[0] if files else None
|
||||
|
||||
def fmt_pct(x):
|
||||
return f"{x:>6.2f}%" if pd.notna(x) else " ----"
|
||||
|
||||
def cell(df, label, mask):
|
||||
sub = df[mask]
|
||||
n = len(sub)
|
||||
if n == 0:
|
||||
return f" {label:<60} n=0"
|
||||
wins = (sub["won"] == True).sum()
|
||||
losses = (sub["won"] == False).sum()
|
||||
settled = wins + losses
|
||||
hr = 100.0 * wins / settled if settled else 0
|
||||
profit = sub["unit_profit"].sum()
|
||||
staked = sub["stake_units"].sum()
|
||||
roi = 100.0 * profit / staked if staked else 0
|
||||
return (f" {label:<60} n={n:>4} hit={hr:>6.2f}% "
|
||||
f"profit={profit:>+7.2f}u roi={roi:>+7.2f}%")
|
||||
|
||||
def hypothesis_block(title, rows):
|
||||
print(f"\n{'─' * 78}")
|
||||
print(f" {title}")
|
||||
print(f"{'─' * 78}")
|
||||
for row in rows:
|
||||
print(row)
|
||||
|
||||
def main():
|
||||
csv_path = latest_csv()
|
||||
if not csv_path:
|
||||
print("No backtest CSV found")
|
||||
return
|
||||
print(f"Reading {csv_path}")
|
||||
df = pd.read_csv(csv_path)
|
||||
print(f"Loaded {len(df)} rows")
|
||||
|
||||
# Filter only playable + settled
|
||||
pdf = df[(df["playable"] == True) & (df["won"].notna())].copy()
|
||||
pdf["won"] = pdf["won"].astype(bool)
|
||||
print(f"Playable + settled: {len(pdf)}")
|
||||
|
||||
overall_hr = (pdf["won"].sum() / len(pdf)) * 100
|
||||
overall_roi = 100.0 * pdf["unit_profit"].sum() / pdf["stake_units"].sum()
|
||||
print(f"\nOVERALL: hit={overall_hr:.2f}% roi={overall_roi:.2f}%")
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H1: TRIPLE VALUE CONFIRMATION
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
triple_confirmed_mask = ~pdf["bb_issues"].fillna("").str.contains(
|
||||
"triple_value_not_confirmed", na=False
|
||||
)
|
||||
hypothesis_block(
|
||||
"H1: TRIPLE VALUE CONFIRMED vs NOT CONFIRMED",
|
||||
[
|
||||
cell(pdf, "triple_value CONFIRMED", triple_confirmed_mask),
|
||||
cell(pdf, "triple_value NOT CONFIRMED", ~triple_confirmed_mask),
|
||||
]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H2: TRAP MARKET FLAG
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
trap_mask = pdf["bb_trap_market"] == True
|
||||
hypothesis_block(
|
||||
"H2: TRAP MARKET FLAG (model says band rate < implied → market overpriced)",
|
||||
[
|
||||
cell(pdf, "trap_market_flag = TRUE (model warned)", trap_mask),
|
||||
cell(pdf, "trap_market_flag = FALSE", ~trap_mask),
|
||||
]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H3: V25/V27 CONSENSUS
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
agree_mask = pdf["v27_consensus"] == "AGREE"
|
||||
disagree_mask = pdf["v27_consensus"] == "DISAGREE"
|
||||
hypothesis_block(
|
||||
"H3: V25 ↔ V27 CONSENSUS",
|
||||
[
|
||||
cell(pdf, "AGREE", agree_mask),
|
||||
cell(pdf, "DISAGREE", disagree_mask),
|
||||
cell(pdf, "neither/null", ~(agree_mask | disagree_mask)),
|
||||
]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H4: ODDS RELIABILITY (league quality)
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
pdf["rel_band"] = pd.cut(
|
||||
pdf["odds_reliability"].fillna(0.35),
|
||||
[0, 0.30, 0.45, 0.55, 1.0],
|
||||
labels=["<0.30 verylow", "0.30-0.45 low", "0.45-0.55 mid", ">=0.55 high"]
|
||||
)
|
||||
hypothesis_block(
|
||||
"H4: LEAGUE ODDS RELIABILITY",
|
||||
[cell(pdf, str(b), pdf["rel_band"] == b) for b in pdf["rel_band"].cat.categories]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H5: CALIBRATOR IMPACT (raw vs calibrated)
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
pdf["calib_delta"] = pdf["calibrated_confidence"] - pdf["raw_confidence"]
|
||||
pdf["delta_band"] = pd.cut(
|
||||
pdf["calib_delta"].fillna(0),
|
||||
[-100, -10, -3, 3, 10, 100],
|
||||
labels=["cal<<raw (-10+)", "cal<raw (-3..-10)", "≈equal (±3)",
|
||||
"cal>raw (3..10)", "cal>>raw (+10+)"]
|
||||
)
|
||||
hypothesis_block(
|
||||
"H5: CALIBRATOR DELTA (calibrated_conf - raw_conf)",
|
||||
[cell(pdf, str(b), pdf["delta_band"] == b) for b in pdf["delta_band"].cat.categories]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H6: EV EDGE
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
pdf["edge_band"] = pd.cut(
|
||||
pdf["ev_edge"].fillna(0),
|
||||
[-10, -0.05, 0.0, 0.05, 0.10, 0.20, 10],
|
||||
labels=["edge<-5%", "-5%-0%", "0-5%", "5-10%", "10-20%", ">20%"]
|
||||
)
|
||||
hypothesis_block(
|
||||
"H6: EV EDGE (model_prob - implied_prob)",
|
||||
[cell(pdf, str(b), pdf["edge_band"] == b) for b in pdf["edge_band"].cat.categories]
|
||||
)
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H7: ODDS x MARKET cross
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
pdf["odds_band"] = pd.cut(
|
||||
pdf["odds"].fillna(0),
|
||||
[0, 1.30, 1.50, 1.80, 2.20, 3.00, 100],
|
||||
labels=["<1.30", "1.30-1.50", "1.50-1.80", "1.80-2.20", "2.20-3.00", ">3.00"]
|
||||
)
|
||||
|
||||
print(f"\n{'─' * 78}")
|
||||
print(f" H7: ODDS BAND × MARKET (per cell hit% / roi% / n)")
|
||||
print(f"{'─' * 78}")
|
||||
pivot_n = pdf.pivot_table(index="market", columns="odds_band",
|
||||
values="match_id", aggfunc="count", fill_value=0,
|
||||
observed=False)
|
||||
pivot_roi = pdf.pivot_table(index="market", columns="odds_band",
|
||||
values="unit_profit", aggfunc="sum", fill_value=0,
|
||||
observed=False)
|
||||
pivot_stake = pdf.pivot_table(index="market", columns="odds_band",
|
||||
values="stake_units", aggfunc="sum", fill_value=0,
|
||||
observed=False)
|
||||
pivot_roi_pct = (100.0 * pivot_roi / pivot_stake.replace(0, np.nan)).round(1)
|
||||
print("\n Bet count per cell:")
|
||||
print(pivot_n.to_string())
|
||||
print("\n ROI% per cell:")
|
||||
print(pivot_roi_pct.to_string())
|
||||
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
# H8: COMBINED FILTER SIMULATION
|
||||
# ─────────────────────────────────────────────────────────────────────
|
||||
print(f"\n{'─' * 78}")
|
||||
print(" H8: COMBINED FILTER SIMULATION (what if we add rules)")
|
||||
print(f"{'─' * 78}")
|
||||
|
||||
def simulate(filter_name, keep_mask):
|
||||
kept = pdf[keep_mask]
|
||||
rejected = pdf[~keep_mask]
|
||||
if len(kept) == 0:
|
||||
return f" {filter_name:<55} → 0 bet remain"
|
||||
kept_hr = 100.0 * kept["won"].sum() / len(kept)
|
||||
kept_profit = kept["unit_profit"].sum()
|
||||
kept_staked = kept["stake_units"].sum()
|
||||
kept_roi = 100.0 * kept_profit / kept_staked if kept_staked else 0
|
||||
saved = -rejected["unit_profit"].sum() # money we WOULD HAVE LOST
|
||||
return (f" {filter_name:<55} keep={len(kept):>3} hit={kept_hr:>5.1f}% "
|
||||
f"roi={kept_roi:>+6.2f}% saved={saved:>+6.2f}u")
|
||||
|
||||
print(simulate("BASELINE (no extra filter)", pd.Series([True] * len(pdf), index=pdf.index)))
|
||||
print(simulate("REJECT triple_value_not_confirmed",
|
||||
~pdf["bb_issues"].fillna("").str.contains("triple_value_not_confirmed")))
|
||||
print(simulate("REJECT trap_market_flag",
|
||||
~(pdf["bb_trap_market"] == True)))
|
||||
print(simulate("REJECT v27 DISAGREE",
|
||||
pdf["v27_consensus"] != "DISAGREE"))
|
||||
print(simulate("REJECT odds_reliability < 0.45",
|
||||
pdf["odds_reliability"].fillna(1.0) >= 0.45))
|
||||
print(simulate("REJECT odds in 1.80-2.20",
|
||||
(pdf["odds"].fillna(0) < 1.80) | (pdf["odds"].fillna(0) >= 2.20)))
|
||||
print(simulate("REJECT ev_edge < 0",
|
||||
pdf["ev_edge"].fillna(0) >= 0))
|
||||
print(simulate("REJECT ev_edge < 0.05",
|
||||
pdf["ev_edge"].fillna(0) >= 0.05))
|
||||
print()
|
||||
print(" COMBINED rules:")
|
||||
# Stack 1: drop triple_not_confirmed + trap_market + DISAGREE
|
||||
s1 = (
|
||||
~pdf["bb_issues"].fillna("").str.contains("triple_value_not_confirmed")
|
||||
& ~(pdf["bb_trap_market"] == True)
|
||||
& (pdf["v27_consensus"] != "DISAGREE")
|
||||
)
|
||||
print(simulate("STACK1: !triple_not_conf & !trap & !disagree", s1))
|
||||
# Stack 2: + edge>=0
|
||||
s2 = s1 & (pdf["ev_edge"].fillna(0) >= 0)
|
||||
print(simulate("STACK2: STACK1 + edge >= 0", s2))
|
||||
# Stack 3: + reliability>=0.45
|
||||
s3 = s2 & (pdf["odds_reliability"].fillna(1.0) >= 0.45)
|
||||
print(simulate("STACK3: STACK2 + reliability >= 0.45", s3))
|
||||
# Stack 4: + odds outside 1.80-2.20
|
||||
s4 = s3 & ((pdf["odds"].fillna(0) < 1.80) | (pdf["odds"].fillna(0) >= 2.20))
|
||||
print(simulate("STACK4: STACK3 + odds NOT in 1.80-2.20", s4))
|
||||
|
||||
print(f"\n{'─' * 78}")
|
||||
print("DONE.")
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,134 @@
|
||||
"""
|
||||
Compare two diagnostic_backtest CSV outputs side-by-side.
|
||||
Used to validate that a filter change actually improved ROI vs the
|
||||
baseline run — and to detect overfitting (in-sample success but
|
||||
out-of-sample collapse).
|
||||
|
||||
Usage:
|
||||
python scripts/compare_backtests.py <baseline.csv> <validation.csv>
|
||||
python scripts/compare_backtests.py (auto-picks 2 most recent CSVs)
|
||||
"""
|
||||
|
||||
import sys, os, glob
|
||||
import pandas as pd
|
||||
from typing import Dict
|
||||
|
||||
REPORTS_DIR = os.path.join(os.path.dirname(os.path.dirname(os.path.abspath(__file__))), "reports")
|
||||
|
||||
|
||||
def load(path: str) -> pd.DataFrame:
|
||||
df = pd.read_csv(path)
|
||||
df["won_bool"] = df["won"].map(
|
||||
{True: True, False: False, "True": True, "False": False, 1: True, 0: False}
|
||||
)
|
||||
return df
|
||||
|
||||
|
||||
def stats(df: pd.DataFrame, mask=None) -> Dict:
|
||||
if mask is not None:
|
||||
df = df[mask]
|
||||
playable = df[(df["playable"] == True) & (df["won_bool"].notna())]
|
||||
if len(playable) == 0:
|
||||
return {"n_total": len(df), "n_playable": 0, "hit": 0, "profit": 0,
|
||||
"staked": 0, "roi": 0}
|
||||
wins = playable["won_bool"].sum()
|
||||
profit = playable["unit_profit"].sum()
|
||||
staked = playable["stake_units"].sum()
|
||||
return {
|
||||
"n_total": int(len(df)),
|
||||
"n_playable": int(len(playable)),
|
||||
"wins": int(wins),
|
||||
"losses": int(len(playable) - wins),
|
||||
"hit": round(100.0 * wins / len(playable), 2),
|
||||
"profit": round(profit, 2),
|
||||
"staked": round(staked, 2),
|
||||
"roi": round(100.0 * profit / staked, 2) if staked else 0,
|
||||
}
|
||||
|
||||
|
||||
def line(label: str, a: Dict, b: Dict, suffix: str = ""):
|
||||
fields = ["n_total", "n_playable", "hit", "profit", "staked", "roi"]
|
||||
parts = [f"{label:<28}"]
|
||||
for f in fields:
|
||||
va = a.get(f, "-")
|
||||
vb = b.get(f, "-")
|
||||
parts.append(f"{f}: {str(va):>8} → {str(vb):>8}")
|
||||
print(" " + " | ".join(parts) + suffix)
|
||||
|
||||
|
||||
def main():
|
||||
if len(sys.argv) == 3:
|
||||
a_path, b_path = sys.argv[1], sys.argv[2]
|
||||
else:
|
||||
files = sorted(glob.glob(os.path.join(REPORTS_DIR, "diagnostic_backtest_*.csv")),
|
||||
key=os.path.getmtime, reverse=True)
|
||||
if len(files) < 2:
|
||||
print("Need at least 2 backtest CSVs in reports/. Pass paths manually.")
|
||||
return
|
||||
b_path, a_path = files[0], files[1] # newest first as "validation"
|
||||
|
||||
print(f"Baseline A: {os.path.basename(a_path)}")
|
||||
print(f"Validation B: {os.path.basename(b_path)}")
|
||||
|
||||
a = load(a_path)
|
||||
b = load(b_path)
|
||||
|
||||
print(f"\n{'=' * 100}")
|
||||
print(f" OVERALL")
|
||||
print(f"{'=' * 100}")
|
||||
line("ALL", stats(a), stats(b))
|
||||
|
||||
print(f"\n{'─' * 100}")
|
||||
print(f" PER MARKET")
|
||||
print(f"{'─' * 100}")
|
||||
markets = sorted(set(a["market"].dropna().unique()) | set(b["market"].dropna().unique()))
|
||||
for m in markets:
|
||||
line(f"market={m}",
|
||||
stats(a, a["market"] == m),
|
||||
stats(b, b["market"] == m))
|
||||
|
||||
# New veto family check — did MUTED_MARKETS actually mute?
|
||||
print(f"\n{'─' * 100}")
|
||||
print(f" NEW VETO IMPACT (look for new veto names in betting_brain.vetoes)")
|
||||
print(f"{'─' * 100}")
|
||||
new_vetoes = ["market_muted_by_backtest", "negative_ev_edge", "ev_edge_too_high_trap",
|
||||
"outside_envelope_edge_low", "outside_envelope_edge_high",
|
||||
"outside_envelope_odds_low", "outside_envelope_v27_must_agree"]
|
||||
for veto in new_vetoes:
|
||||
a_hits = a["bb_vetoes"].fillna("").str.contains(veto).sum()
|
||||
b_hits = b["bb_vetoes"].fillna("").str.contains(veto).sum()
|
||||
print(f" {veto:<45} A={a_hits:>4} B={b_hits:>4}")
|
||||
|
||||
# Top issue tags
|
||||
print(f"\n{'─' * 100}")
|
||||
print(f" BTTS MUTE CHECK — should be ~0 playable in validation")
|
||||
print(f"{'─' * 100}")
|
||||
a_btts_play = ((a["market"] == "BTTS") & (a["playable"] == True)).sum()
|
||||
b_btts_play = ((b["market"] == "BTTS") & (b["playable"] == True)).sum()
|
||||
print(f" BTTS playable bets: A={a_btts_play} → B={b_btts_play} "
|
||||
f"(should be 0 in B if MUTE works)")
|
||||
|
||||
# Verdict
|
||||
print(f"\n{'=' * 100}")
|
||||
a_s = stats(a)
|
||||
b_s = stats(b)
|
||||
roi_delta = b_s["roi"] - a_s["roi"]
|
||||
if b_s["n_playable"] < 20:
|
||||
verdict = "TOO FEW BETS — sample insufficient"
|
||||
elif roi_delta > 5 and b_s["roi"] > 0:
|
||||
verdict = "✅ FILTERS WORK — ROI improved AND positive"
|
||||
elif roi_delta > 5:
|
||||
verdict = "🟡 PARTIAL — ROI improved but still negative"
|
||||
elif roi_delta > 0:
|
||||
verdict = "🟡 SLIGHT IMPROVEMENT"
|
||||
elif roi_delta < -5:
|
||||
verdict = "❌ OVERFITTING — validation ROI collapsed"
|
||||
else:
|
||||
verdict = "❌ NO MATERIAL CHANGE"
|
||||
print(f" VERDICT: {verdict}")
|
||||
print(f" ROI: {a_s['roi']}% → {b_s['roi']}% (Δ {roi_delta:+.2f}pp)")
|
||||
print(f"{'=' * 100}")
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,674 @@
|
||||
"""
|
||||
Diagnostic Backtest
|
||||
===================
|
||||
Run the full V28 orchestrator (in-process — no HTTP) on a window of completed
|
||||
matches, capture the recommendation + key signal features + the actual outcome,
|
||||
and produce a *diagnostic* report: not just "what was the hit rate" but
|
||||
"which feature clusters drive the losing bets".
|
||||
|
||||
Outputs:
|
||||
- reports/diagnostic_backtest_YYYYMMDD.csv (per-bet detail)
|
||||
- reports/diagnostic_backtest_YYYYMMDD.json (aggregate metrics)
|
||||
- reports/diagnostic_backtest_YYYYMMDD.txt (human-readable summary)
|
||||
|
||||
Usage:
|
||||
python scripts/diagnostic_backtest.py --days 14 --max-matches 2000
|
||||
python scripts/diagnostic_backtest.py --start 2026-05-10 --end 2026-05-24
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
import time
|
||||
import traceback
|
||||
from collections import defaultdict, Counter
|
||||
from datetime import datetime, timedelta, timezone
|
||||
from typing import Any, Dict, List, Optional, Tuple
|
||||
|
||||
import psycopg2
|
||||
from psycopg2.extras import RealDictCursor
|
||||
|
||||
# Path bootstrap so we can import the ai-engine package from anywhere
|
||||
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
|
||||
AI_ENGINE_DIR = os.path.dirname(SCRIPT_DIR)
|
||||
sys.path.insert(0, AI_ENGINE_DIR)
|
||||
|
||||
from data.db import get_clean_dsn
|
||||
from services.single_match_orchestrator import get_single_match_orchestrator
|
||||
|
||||
REPORTS_DIR = os.path.join(AI_ENGINE_DIR, "reports")
|
||||
os.makedirs(REPORTS_DIR, exist_ok=True)
|
||||
|
||||
# Days with confirmed feeder gaps — exclude from sample
|
||||
EXCLUDED_DATES = {"2026-05-03", "2026-04-29"}
|
||||
|
||||
|
||||
# ── Outcome resolution ────────────────────────────────────────────────
|
||||
def _norm_pick(pick: Optional[str]) -> str:
|
||||
return str(pick or "").strip().casefold()
|
||||
|
||||
|
||||
def resolve_outcome(market: str, pick: str, sh: int, sa: int,
|
||||
htsh: Optional[int], htsa: Optional[int]) -> Optional[bool]:
|
||||
"""Mirror of prediction-settlement.market-resolver.ts (TS side).
|
||||
Returns True/False on settle, None if cannot resolve."""
|
||||
m = (market or "").upper().replace(" ", "").replace("-", "_")
|
||||
p = _norm_pick(pick)
|
||||
|
||||
if m in ("MS", "ML", "1X2"):
|
||||
outcome = "1" if sh > sa else "2" if sa > sh else "x"
|
||||
return p in {outcome, outcome.upper(), outcome.lower(), "0" if outcome == "x" else outcome}
|
||||
|
||||
if m in ("HT", "IY"):
|
||||
if htsh is None or htsa is None:
|
||||
return None
|
||||
outcome = "1" if htsh > htsa else "2" if htsa > htsh else "x"
|
||||
return p in {outcome, "0" if outcome == "x" else outcome}
|
||||
|
||||
if m in ("OU05", "OU15", "OU25", "OU35", "OU45", "TOTAL"):
|
||||
line = {"OU05": 0.5, "OU15": 1.5, "OU25": 2.5, "OU35": 3.5,
|
||||
"OU45": 4.5, "TOTAL": 2.5}[m]
|
||||
total = sh + sa
|
||||
if total == line:
|
||||
return None
|
||||
is_over = total > line
|
||||
if "over" in p or "üst" in p or "ust" in p:
|
||||
return is_over
|
||||
if "alt" in p or "under" in p:
|
||||
return not is_over
|
||||
return None
|
||||
|
||||
if m in ("OU05_HT", "OU15_HT", "OU25_HT", "HT_OU05", "HT_OU15", "HT_OU25"):
|
||||
if htsh is None or htsa is None:
|
||||
return None
|
||||
line = {"OU05_HT": 0.5, "OU15_HT": 1.5, "OU25_HT": 2.5,
|
||||
"HT_OU05": 0.5, "HT_OU15": 1.5, "HT_OU25": 2.5}[m]
|
||||
total = htsh + htsa
|
||||
if total == line:
|
||||
return None
|
||||
is_over = total > line
|
||||
if "over" in p or "üst" in p or "ust" in p:
|
||||
return is_over
|
||||
if "alt" in p or "under" in p:
|
||||
return not is_over
|
||||
return None
|
||||
|
||||
if m in ("BTTS", "KG"):
|
||||
both = sh > 0 and sa > 0
|
||||
if "yes" in p or "var" in p:
|
||||
return both
|
||||
if "no" in p or "yok" in p:
|
||||
return not both
|
||||
return None
|
||||
|
||||
if m in ("HTFT", "IYMS"):
|
||||
if htsh is None or htsa is None or "/" not in p:
|
||||
return None
|
||||
ht_p, ft_p = p.split("/", 1)
|
||||
ht_actual = "1" if htsh > htsa else "2" if htsa > htsh else "x"
|
||||
ft_actual = "1" if sh > sa else "2" if sa > sh else "x"
|
||||
return ht_p.strip() == ht_actual and ft_p.strip() == ft_actual
|
||||
|
||||
if m in ("DC", "CIFTE_SANS"):
|
||||
ft = "1" if sh > sa else "2" if sa > sh else "X"
|
||||
raw = p.upper().replace("-", "").replace("/", "")
|
||||
if raw in ("1X", "X1"):
|
||||
pair = ["1", "X"]
|
||||
elif raw in ("X2", "2X"):
|
||||
pair = ["X", "2"]
|
||||
elif raw in ("12", "21"):
|
||||
pair = ["1", "2"]
|
||||
else:
|
||||
return None
|
||||
return ft in pair
|
||||
|
||||
if m in ("OE", "TEKCIFT"):
|
||||
is_odd = (sh + sa) % 2 == 1
|
||||
if "tek" in p or "odd" in p:
|
||||
return is_odd
|
||||
if "cift" in p or "çift" in p or "even" in p:
|
||||
return not is_odd
|
||||
return None
|
||||
|
||||
return None
|
||||
|
||||
|
||||
def compute_unit_profit(won: Optional[bool], stake: float, odds: Optional[float]) -> float:
|
||||
if won is None:
|
||||
return 0.0
|
||||
if not won:
|
||||
return -abs(stake) if stake else -1.0
|
||||
if not odds or odds <= 1.0:
|
||||
return 0.0
|
||||
return round(stake * (odds - 1.0), 4)
|
||||
|
||||
|
||||
# ── Data fetch ────────────────────────────────────────────────────────
|
||||
def fetch_match_window(args) -> List[Dict]:
|
||||
dsn = get_clean_dsn()
|
||||
if "?schema=" in dsn:
|
||||
dsn = dsn.split("?schema=")[0]
|
||||
|
||||
if args.start and args.end:
|
||||
start = datetime.strptime(args.start, "%Y-%m-%d")
|
||||
end = datetime.strptime(args.end, "%Y-%m-%d") + timedelta(days=1)
|
||||
else:
|
||||
end = datetime.now(timezone.utc).replace(tzinfo=None)
|
||||
start = end - timedelta(days=args.days)
|
||||
|
||||
start_ms = int(start.timestamp() * 1000)
|
||||
end_ms = int(end.timestamp() * 1000)
|
||||
|
||||
excluded = sorted(EXCLUDED_DATES)
|
||||
excluded_clause = ""
|
||||
if excluded:
|
||||
ex_csv = ",".join(f"'{d}'" for d in excluded)
|
||||
excluded_clause = (
|
||||
f" AND to_timestamp(mst_utc/1000)::date "
|
||||
f"NOT IN ({ex_csv})"
|
||||
)
|
||||
|
||||
with psycopg2.connect(dsn) as conn:
|
||||
with conn.cursor(cursor_factory=RealDictCursor) as cur:
|
||||
cur.execute(
|
||||
f"""
|
||||
SELECT id AS match_id,
|
||||
score_home, score_away,
|
||||
ht_score_home, ht_score_away,
|
||||
league_id,
|
||||
to_timestamp(mst_utc/1000)::date AS match_date
|
||||
FROM matches
|
||||
WHERE sport='football'
|
||||
AND status='FT'
|
||||
AND score_home IS NOT NULL
|
||||
AND score_away IS NOT NULL
|
||||
AND mst_utc >= %s
|
||||
AND mst_utc < %s
|
||||
{excluded_clause}
|
||||
ORDER BY mst_utc DESC
|
||||
LIMIT %s
|
||||
""",
|
||||
(start_ms, end_ms, args.max_matches),
|
||||
)
|
||||
return cur.fetchall()
|
||||
|
||||
|
||||
# ── Per-bet capture ───────────────────────────────────────────────────
|
||||
def capture_bet_row(match: Dict, package: Dict) -> Dict[str, Any]:
|
||||
"""Distill orchestrator response + ground truth into one analytic row."""
|
||||
main = package.get("main_pick") or {}
|
||||
bb = main.get("betting_brain") or {}
|
||||
advice = package.get("bet_advice") or {}
|
||||
v27 = package.get("v27_engine") or {}
|
||||
triple = (v27.get("triple_value") or {})
|
||||
risk = package.get("risk") or {}
|
||||
quality = package.get("data_quality") or {}
|
||||
htft_payload = ((package.get("market_board") or {}).get("HTFT") or {})
|
||||
htft_probs = htft_payload.get("probs") or {}
|
||||
|
||||
sh, sa = match["score_home"], match["score_away"]
|
||||
htsh, htsa = match["ht_score_home"], match["ht_score_away"]
|
||||
|
||||
market = main.get("market")
|
||||
pick = main.get("pick")
|
||||
odds_val = _f(main.get("odds"))
|
||||
stake = _f(main.get("stake_units"), 1.0)
|
||||
playable = bool(main.get("playable")) and bool(advice.get("playable"))
|
||||
|
||||
won = resolve_outcome(market, pick, sh, sa, htsh, htsa) if market and pick else None
|
||||
profit = compute_unit_profit(won, stake, odds_val) if playable else 0.0
|
||||
|
||||
# Reversal context (only meaningful for MS picks)
|
||||
rev_prob = None
|
||||
if market == "MS" and pick in ("1", "2"):
|
||||
if pick == "1":
|
||||
rev_prob = _f(htft_probs.get("1/2"), 0.0) + _f(htft_probs.get("1/X"), 0.0)
|
||||
else:
|
||||
rev_prob = _f(htft_probs.get("2/1"), 0.0) + _f(htft_probs.get("2/X"), 0.0)
|
||||
|
||||
return {
|
||||
"match_id": match["match_id"],
|
||||
"match_date": str(match["match_date"]),
|
||||
"league_id": match.get("league_id"),
|
||||
"score_home": sh,
|
||||
"score_away": sa,
|
||||
"ht_score_home": htsh,
|
||||
"ht_score_away": htsa,
|
||||
"market": market,
|
||||
"pick": pick,
|
||||
"odds": odds_val,
|
||||
"stake_units": stake,
|
||||
"playable": playable,
|
||||
"won": won,
|
||||
"unit_profit": profit,
|
||||
"raw_confidence": _f(main.get("raw_confidence")),
|
||||
"calibrated_confidence": _f(main.get("calibrated_confidence")),
|
||||
"play_score": _f(main.get("play_score")),
|
||||
"ev_edge": _f(main.get("ev_edge")),
|
||||
"bet_grade": main.get("bet_grade"),
|
||||
"is_value_sniper": bool(main.get("is_value_sniper")),
|
||||
"bb_score": _f(bb.get("score")),
|
||||
"bb_action": bb.get("action"),
|
||||
"bb_vetoes": ";".join(bb.get("vetoes") or []),
|
||||
"bb_issues": ";".join(bb.get("issues") or []),
|
||||
"bb_positives": ";".join(bb.get("positives") or []),
|
||||
"bb_model_prob": _f(bb.get("model_prob")),
|
||||
"bb_implied_prob": _f(bb.get("implied_prob")),
|
||||
"bb_model_market_gap": _f(bb.get("model_market_gap")),
|
||||
"bb_divergence": _f(bb.get("divergence")),
|
||||
"bb_trap_market": bool(bb.get("trap_market_flag")),
|
||||
"v27_consensus": v27.get("consensus"),
|
||||
"data_quality_score": _f(quality.get("score")),
|
||||
"data_quality_flags": ";".join(quality.get("flags") or []),
|
||||
"risk_level": (risk.get("level") if isinstance(risk, dict) else None),
|
||||
"odds_reliability": _f(main.get("odds_reliability")),
|
||||
"htft_reversal_prob": rev_prob,
|
||||
"htft_top_pick": _argmax(htft_probs),
|
||||
"league_name": (package.get("match_info") or {}).get("league_name"),
|
||||
"is_cup": _is_cup((package.get("match_info") or {}).get("league_name") or ""),
|
||||
"model_version": package.get("model_version"),
|
||||
"decision_reason": main.get("pick_reason") or advice.get("reason"),
|
||||
}
|
||||
|
||||
|
||||
def _f(x: Any, default: Optional[float] = None) -> Optional[float]:
|
||||
try:
|
||||
return float(x) if x is not None else default
|
||||
except (TypeError, ValueError):
|
||||
return default
|
||||
|
||||
|
||||
def _argmax(d: Dict[str, Any]) -> Optional[str]:
|
||||
best, val = None, -1.0
|
||||
for k, v in d.items():
|
||||
fv = _f(v, 0.0) or 0.0
|
||||
if fv > val:
|
||||
best, val = k, fv
|
||||
return best
|
||||
|
||||
|
||||
_CUP_KEYWORDS = ("kupa", "cup", "coupe", "copa", "coppa", "pokal", "trophy",
|
||||
"shield", "ziraat", "süper kupa", "super cup", "beker", "taça", "taca")
|
||||
|
||||
|
||||
def _is_cup(name: str) -> bool:
|
||||
n = (name or "").lower()
|
||||
return any(kw in n for kw in _CUP_KEYWORDS)
|
||||
|
||||
|
||||
# ── Aggregation helpers ────────────────────────────────────────────────
|
||||
def _bucket(value: Optional[float], edges: List[float]) -> Optional[str]:
|
||||
if value is None:
|
||||
return None
|
||||
for i, edge in enumerate(edges):
|
||||
if value < edge:
|
||||
if i == 0:
|
||||
return f"<{edge}"
|
||||
return f"{edges[i-1]}-{edge}"
|
||||
return f">={edges[-1]}"
|
||||
|
||||
|
||||
def _summary_stats(rows: List[Dict]) -> Dict[str, Any]:
|
||||
if not rows:
|
||||
return {"n": 0}
|
||||
settled = [r for r in rows if r["playable"] and r["won"] is not None]
|
||||
won = sum(1 for r in settled if r["won"])
|
||||
lost = sum(1 for r in settled if not r["won"])
|
||||
profit = sum(float(r["unit_profit"]) for r in settled)
|
||||
staked = sum(float(r["stake_units"]) for r in settled)
|
||||
return {
|
||||
"n_total": len(rows),
|
||||
"n_playable_settled": len(settled),
|
||||
"wins": won,
|
||||
"losses": lost,
|
||||
"hit_rate_pct": round(100.0 * won / len(settled), 2) if settled else None,
|
||||
"unit_profit": round(profit, 3),
|
||||
"staked": round(staked, 3),
|
||||
"roi_pct": round(100.0 * profit / staked, 2) if staked else None,
|
||||
}
|
||||
|
||||
|
||||
def aggregate(rows: List[Dict]) -> Dict[str, Any]:
|
||||
out: Dict[str, Any] = {"overall": _summary_stats(rows)}
|
||||
|
||||
by = lambda key_fn: defaultdict(list)
|
||||
market_buckets = by(None)
|
||||
conf_buckets = by(None)
|
||||
odds_buckets = by(None)
|
||||
grade_buckets = by(None)
|
||||
cup_buckets = by(None)
|
||||
motivation_buckets = by(None)
|
||||
|
||||
for r in rows:
|
||||
if r["playable"]:
|
||||
market_buckets[r["market"] or "?"].append(r)
|
||||
conf_buckets[_bucket(r["calibrated_confidence"],
|
||||
[45, 50, 55, 60, 65, 70, 80])].append(r)
|
||||
odds_buckets[_bucket(r["odds"], [1.3, 1.5, 1.8, 2.2, 3.0, 5.0])].append(r)
|
||||
grade_buckets[r["bet_grade"] or "?"].append(r)
|
||||
cup_buckets["cup" if r["is_cup"] else "league"].append(r)
|
||||
|
||||
out["by_market"] = {k: _summary_stats(v) for k, v in market_buckets.items()}
|
||||
out["by_confidence"] = {k: _summary_stats(v) for k, v in conf_buckets.items() if k}
|
||||
out["by_odds"] = {k: _summary_stats(v) for k, v in odds_buckets.items() if k}
|
||||
out["by_grade"] = {k: _summary_stats(v) for k, v in grade_buckets.items()}
|
||||
out["by_competition"] = {k: _summary_stats(v) for k, v in cup_buckets.items()}
|
||||
return out
|
||||
|
||||
|
||||
def loss_diagnostics(rows: List[Dict]) -> Dict[str, Any]:
|
||||
losses = [r for r in rows if r["playable"] and r["won"] is False]
|
||||
if not losses:
|
||||
return {"n_losses": 0}
|
||||
n = len(losses)
|
||||
|
||||
def share(predicate) -> Tuple[int, float]:
|
||||
c = sum(1 for r in losses if predicate(r))
|
||||
return c, round(100.0 * c / n, 2)
|
||||
|
||||
diagnostics = {
|
||||
"n_losses": n,
|
||||
"total_loss_units": round(sum(float(r["unit_profit"]) for r in losses), 3),
|
||||
"patterns": {
|
||||
"high_htft_reversal_prob (>=0.20)": share(
|
||||
lambda r: (r.get("htft_reversal_prob") or 0) >= 0.20
|
||||
),
|
||||
"cup_match": share(lambda r: r["is_cup"]),
|
||||
"low_league_reliability (<0.45)": share(
|
||||
lambda r: (r.get("odds_reliability") or 1) < 0.45
|
||||
),
|
||||
"v27_disagree": share(lambda r: r.get("v27_consensus") == "DISAGREE"),
|
||||
"trap_market_flagged": share(lambda r: r.get("bb_trap_market")),
|
||||
"low_calibrated_conf (<55)": share(
|
||||
lambda r: (r.get("calibrated_confidence") or 0) < 55
|
||||
),
|
||||
"high_odds_underdog (>=2.5)": share(
|
||||
lambda r: (r.get("odds") or 0) >= 2.5
|
||||
),
|
||||
"low_data_quality (<0.55)": share(
|
||||
lambda r: (r.get("data_quality_score") or 1) < 0.55
|
||||
),
|
||||
"high_risk_level": share(
|
||||
lambda r: r.get("risk_level") in ("HIGH", "EXTREME")
|
||||
),
|
||||
"inferred_features": share(
|
||||
lambda r: "ai_features_inferred_from_history" in (r.get("data_quality_flags") or "")
|
||||
),
|
||||
},
|
||||
"by_market": Counter(r["market"] for r in losses).most_common(),
|
||||
"by_league": Counter(r.get("league_name") for r in losses).most_common(10),
|
||||
}
|
||||
|
||||
# Top issue tags from betting_brain across losses
|
||||
issue_counter = Counter()
|
||||
veto_counter = Counter()
|
||||
for r in losses:
|
||||
for tag in (r.get("bb_issues") or "").split(";"):
|
||||
if tag:
|
||||
issue_counter[tag] += 1
|
||||
for tag in (r.get("bb_vetoes") or "").split(";"):
|
||||
if tag:
|
||||
veto_counter[tag] += 1
|
||||
diagnostics["top_bb_issues_in_losses"] = issue_counter.most_common(15)
|
||||
diagnostics["top_bb_vetoes_in_losses"] = veto_counter.most_common(15)
|
||||
return diagnostics
|
||||
|
||||
|
||||
# ── Recommendations ────────────────────────────────────────────────────
|
||||
def make_recommendations(rows: List[Dict], agg: Dict[str, Any],
|
||||
diag: Dict[str, Any]) -> List[Dict[str, Any]]:
|
||||
recs: List[Dict[str, Any]] = []
|
||||
overall = agg.get("overall") or {}
|
||||
if not overall.get("n_playable_settled"):
|
||||
return recs
|
||||
|
||||
# Cross-reference market hit rate vs overall — flag chronic losers.
|
||||
overall_hit = overall.get("hit_rate_pct") or 0.0
|
||||
for market, stats in (agg.get("by_market") or {}).items():
|
||||
n = stats.get("n_playable_settled") or 0
|
||||
hit = stats.get("hit_rate_pct")
|
||||
roi = stats.get("roi_pct")
|
||||
if n < 30:
|
||||
continue
|
||||
if hit is not None and roi is not None and roi < -10 and hit < overall_hit - 10:
|
||||
recs.append({
|
||||
"type": "drop_market",
|
||||
"market": market,
|
||||
"evidence": f"hit={hit}%, roi={roi}%, n={n} — chronic loser",
|
||||
"suggested_fix": f"Add veto in betting_brain when market=={market} unless overwhelming evidence",
|
||||
"estimated_loss_prevented_units": round(-(stats.get("unit_profit") or 0), 2),
|
||||
})
|
||||
|
||||
# Confidence band tuning — flag bands where ROI < 0 despite passing eşik
|
||||
for band, stats in (agg.get("by_confidence") or {}).items():
|
||||
n = stats.get("n_playable_settled") or 0
|
||||
roi = stats.get("roi_pct")
|
||||
if n >= 40 and roi is not None and roi < -8:
|
||||
recs.append({
|
||||
"type": "raise_confidence_threshold",
|
||||
"confidence_band": band,
|
||||
"evidence": f"n={n}, roi={roi}%",
|
||||
"suggested_fix": f"Raise MIN_BET_SCORE or market_min_conf above {band.split('-')[0]}",
|
||||
})
|
||||
|
||||
# Loss diagnostic — if cup matches dominate losses, recommend cup-aware filter
|
||||
patterns = (diag.get("patterns") or {})
|
||||
cup_share = patterns.get("cup_match", (0, 0))[1]
|
||||
if cup_share >= 25:
|
||||
recs.append({
|
||||
"type": "cup_match_filter",
|
||||
"evidence": f"{cup_share}% of losses are cup matches",
|
||||
"suggested_fix": "Tighten betting_brain thresholds for is_cup_match=True picks",
|
||||
})
|
||||
|
||||
rev_share = patterns.get("high_htft_reversal_prob (>=0.20)", (0, 0))[1]
|
||||
if rev_share >= 15:
|
||||
recs.append({
|
||||
"type": "tighten_reversal_check",
|
||||
"evidence": f"{rev_share}% of losses had HTFT reversal prob >=0.20 (already partial fix)",
|
||||
"suggested_fix": "Lower reversal threshold in betting_brain from 0.25 to 0.20 for veto trigger",
|
||||
})
|
||||
|
||||
rel_share = patterns.get("low_league_reliability (<0.45)", (0, 0))[1]
|
||||
if rel_share >= 20:
|
||||
recs.append({
|
||||
"type": "league_reliability_filter",
|
||||
"evidence": f"{rel_share}% of losses in low-reliability leagues (<0.45)",
|
||||
"suggested_fix": "Add hard veto when odds_reliability<0.45 for non-value-sniper picks",
|
||||
})
|
||||
|
||||
return recs
|
||||
|
||||
|
||||
# ── CSV / report writers ───────────────────────────────────────────────
|
||||
def write_csv(rows: List[Dict], path: str):
|
||||
if not rows:
|
||||
return
|
||||
import csv
|
||||
fields = list(rows[0].keys())
|
||||
with open(path, "w", newline="", encoding="utf-8") as f:
|
||||
w = csv.DictWriter(f, fieldnames=fields)
|
||||
w.writeheader()
|
||||
for r in rows:
|
||||
w.writerow(r)
|
||||
|
||||
|
||||
def write_text_summary(rows: List[Dict], agg: Dict, diag: Dict,
|
||||
recs: List[Dict], path: str, args):
|
||||
lines: List[str] = []
|
||||
push = lines.append
|
||||
push("=" * 78)
|
||||
push("DIAGNOSTIC BACKTEST REPORT")
|
||||
push("=" * 78)
|
||||
push(f"Generated: {datetime.now().isoformat(timespec='seconds')}")
|
||||
push(f"Sample window: start={args.start or f'-{args.days}d'}, end={args.end or 'now'}")
|
||||
push(f"Max matches: {args.max_matches}")
|
||||
push(f"Excluded days: {sorted(EXCLUDED_DATES)}")
|
||||
push("")
|
||||
push("OVERALL")
|
||||
push("-" * 78)
|
||||
overall = agg.get("overall") or {}
|
||||
for k in ("n_total", "n_playable_settled", "wins", "losses",
|
||||
"hit_rate_pct", "unit_profit", "staked", "roi_pct"):
|
||||
push(f" {k:25}: {overall.get(k)}")
|
||||
push("")
|
||||
push("PER MARKET")
|
||||
push("-" * 78)
|
||||
push(f" {'market':<8} {'n':>6} {'hit%':>7} {'profit':>9} {'roi%':>7}")
|
||||
for market, s in sorted((agg.get("by_market") or {}).items(),
|
||||
key=lambda kv: -(kv[1].get("n_playable_settled") or 0)):
|
||||
push(f" {market:<8} {s.get('n_playable_settled',0):>6} "
|
||||
f"{str(s.get('hit_rate_pct','')):>7} "
|
||||
f"{str(s.get('unit_profit','')):>9} "
|
||||
f"{str(s.get('roi_pct','')):>7}")
|
||||
push("")
|
||||
push("PER CALIBRATED CONFIDENCE BAND")
|
||||
push("-" * 78)
|
||||
push(f" {'band':<10} {'n':>6} {'hit%':>7} {'roi%':>7}")
|
||||
for band, s in sorted((agg.get("by_confidence") or {}).items()):
|
||||
push(f" {band:<10} {s.get('n_playable_settled',0):>6} "
|
||||
f"{str(s.get('hit_rate_pct','')):>7} "
|
||||
f"{str(s.get('roi_pct','')):>7}")
|
||||
push("")
|
||||
push("PER ODDS BAND")
|
||||
push("-" * 78)
|
||||
push(f" {'band':<10} {'n':>6} {'hit%':>7} {'roi%':>7}")
|
||||
for band, s in sorted((agg.get("by_odds") or {}).items()):
|
||||
push(f" {band:<10} {s.get('n_playable_settled',0):>6} "
|
||||
f"{str(s.get('hit_rate_pct','')):>7} "
|
||||
f"{str(s.get('roi_pct','')):>7}")
|
||||
push("")
|
||||
push("LEAGUE vs CUP")
|
||||
push("-" * 78)
|
||||
for k, s in (agg.get("by_competition") or {}).items():
|
||||
push(f" {k:<8} n={s.get('n_playable_settled',0):>4} "
|
||||
f"hit={s.get('hit_rate_pct','-')}% roi={s.get('roi_pct','-')}%")
|
||||
push("")
|
||||
push("LOSS DIAGNOSTICS")
|
||||
push("-" * 78)
|
||||
push(f" total losses: {diag.get('n_losses')}")
|
||||
push(f" total lost units: {diag.get('total_loss_units')}")
|
||||
push(f" By market: {diag.get('by_market')}")
|
||||
push(" Loss patterns (count, % of losses):")
|
||||
for pattern, (c, pct) in (diag.get("patterns") or {}).items():
|
||||
push(f" {pattern:<55} {c:>4} ({pct}%)")
|
||||
push(" Top betting_brain issues seen in losses:")
|
||||
for issue, c in (diag.get("top_bb_issues_in_losses") or []):
|
||||
push(f" {issue:<55} {c}")
|
||||
push(" Top betting_brain vetoes (in losses — i.e. veto fired but bet still went through value-sniper override):")
|
||||
for veto, c in (diag.get("top_bb_vetoes_in_losses") or []):
|
||||
push(f" {veto:<55} {c}")
|
||||
push("")
|
||||
push("RECOMMENDATIONS")
|
||||
push("-" * 78)
|
||||
if not recs:
|
||||
push(" (none surfaced — sample too small or no clear pattern)")
|
||||
for r in recs:
|
||||
push(f" • [{r['type']}]")
|
||||
for k, v in r.items():
|
||||
if k == "type":
|
||||
continue
|
||||
push(f" {k}: {v}")
|
||||
push("")
|
||||
push("=" * 78)
|
||||
with open(path, "w", encoding="utf-8") as f:
|
||||
f.write("\n".join(lines))
|
||||
|
||||
|
||||
# ── Main loop ─────────────────────────────────────────────────────────
|
||||
def main():
|
||||
parser = argparse.ArgumentParser(description=__doc__)
|
||||
parser.add_argument("--days", type=int, default=14,
|
||||
help="Backwards window from now (default 14)")
|
||||
parser.add_argument("--max-matches", type=int, default=2000,
|
||||
help="Hard cap on matches processed (default 2000)")
|
||||
parser.add_argument("--start", help="Start date YYYY-MM-DD (overrides --days)")
|
||||
parser.add_argument("--end", help="End date YYYY-MM-DD")
|
||||
parser.add_argument("--progress-interval", type=int, default=50)
|
||||
args = parser.parse_args()
|
||||
|
||||
print("=" * 70)
|
||||
print("DIAGNOSTIC BACKTEST")
|
||||
print("=" * 70)
|
||||
print(f"Loading orchestrator...")
|
||||
orch = get_single_match_orchestrator()
|
||||
# Warm V25 + V27 + basketball loaders so the first match doesn't pay it
|
||||
try:
|
||||
orch._get_v25_predictor()
|
||||
except Exception as e:
|
||||
print(f" v25 warmup: {e}")
|
||||
try:
|
||||
orch._get_v27_predictor()
|
||||
except Exception as e:
|
||||
print(f" v27 warmup: {e}")
|
||||
|
||||
print(f"Fetching match window...")
|
||||
matches = fetch_match_window(args)
|
||||
n = len(matches)
|
||||
print(f" {n} matches selected")
|
||||
|
||||
if not matches:
|
||||
print("No matches to process. Exiting.")
|
||||
return
|
||||
|
||||
rows: List[Dict[str, Any]] = []
|
||||
errors: List[Tuple[str, str]] = []
|
||||
t0 = time.time()
|
||||
|
||||
for i, m in enumerate(matches, start=1):
|
||||
mid = str(m["match_id"])
|
||||
try:
|
||||
pkg = orch.analyze_match(mid)
|
||||
if pkg is None:
|
||||
continue
|
||||
row = capture_bet_row(m, pkg)
|
||||
rows.append(row)
|
||||
except KeyboardInterrupt:
|
||||
print("\nInterrupted, writing partial results...")
|
||||
break
|
||||
except Exception as e:
|
||||
errors.append((mid, str(e)))
|
||||
if len(errors) <= 5:
|
||||
traceback.print_exc()
|
||||
|
||||
if i % args.progress_interval == 0:
|
||||
elapsed = time.time() - t0
|
||||
rate = i / elapsed
|
||||
eta = (n - i) / rate if rate else 0
|
||||
playable_so_far = sum(1 for r in rows if r["playable"])
|
||||
print(f" [{i}/{n}] rate={rate:.1f}/s eta={eta/60:.1f}min "
|
||||
f"playable={playable_so_far} errors={len(errors)}")
|
||||
|
||||
print(f"\nProcessed {len(rows)} rows in {(time.time()-t0):.1f}s "
|
||||
f"({len(errors)} errors)")
|
||||
|
||||
# Aggregate
|
||||
print("Aggregating...")
|
||||
agg = aggregate(rows)
|
||||
diag = loss_diagnostics(rows)
|
||||
recs = make_recommendations(rows, agg, diag)
|
||||
|
||||
stamp = datetime.now().strftime("%Y%m%d_%H%M%S")
|
||||
csv_path = os.path.join(REPORTS_DIR, f"diagnostic_backtest_{stamp}.csv")
|
||||
json_path = os.path.join(REPORTS_DIR, f"diagnostic_backtest_{stamp}.json")
|
||||
txt_path = os.path.join(REPORTS_DIR, f"diagnostic_backtest_{stamp}.txt")
|
||||
|
||||
write_csv(rows, csv_path)
|
||||
with open(json_path, "w", encoding="utf-8") as f:
|
||||
json.dump({"args": vars(args), "aggregate": agg, "loss_diagnostics": diag,
|
||||
"recommendations": recs, "errors_sample": errors[:20]},
|
||||
f, indent=2, default=str)
|
||||
write_text_summary(rows, agg, diag, recs, txt_path, args)
|
||||
|
||||
print(f"\nOutputs:")
|
||||
print(f" CSV: {csv_path}")
|
||||
print(f" JSON: {json_path}")
|
||||
print(f" TXT: {txt_path}")
|
||||
print("\nOverall:", agg.get("overall"))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,254 @@
|
||||
"""
|
||||
Filter Optimizer
|
||||
================
|
||||
Grid-search over filter thresholds (per market) using the existing
|
||||
diagnostic_backtest CSV. Finds the (confidence, edge, odds, reliability)
|
||||
combination that maximizes ROI while keeping bet volume reasonable.
|
||||
|
||||
No re-prediction needed — pure offline simulation on the bets already
|
||||
captured. Output: per-market optimal thresholds + projected ROI lift +
|
||||
JSON patch ready to drop into config/market_thresholds.json.
|
||||
|
||||
Usage:
|
||||
python scripts/optimize_filters.py
|
||||
python scripts/optimize_filters.py --csv reports/diagnostic_backtest_X.csv
|
||||
python scripts/optimize_filters.py --min-bets 20 --apply
|
||||
"""
|
||||
|
||||
import argparse
|
||||
import json
|
||||
import os
|
||||
import sys
|
||||
import glob
|
||||
import itertools
|
||||
from typing import List, Dict, Tuple, Optional
|
||||
import pandas as pd
|
||||
import numpy as np
|
||||
|
||||
SCRIPT_DIR = os.path.dirname(os.path.abspath(__file__))
|
||||
AI_ENGINE_DIR = os.path.dirname(SCRIPT_DIR)
|
||||
sys.path.insert(0, AI_ENGINE_DIR)
|
||||
|
||||
REPORTS_DIR = os.path.join(AI_ENGINE_DIR, "reports")
|
||||
CONFIG_PATH = os.path.join(AI_ENGINE_DIR, "config", "market_thresholds.json")
|
||||
|
||||
|
||||
def latest_csv() -> Optional[str]:
|
||||
files = sorted(glob.glob(os.path.join(REPORTS_DIR, "diagnostic_backtest_*.csv")),
|
||||
key=os.path.getmtime, reverse=True)
|
||||
return files[0] if files else None
|
||||
|
||||
|
||||
def load_backtest(path: str) -> pd.DataFrame:
|
||||
df = pd.read_csv(path)
|
||||
# Keep only playable + settled bets — these are what the SYSTEM
|
||||
# actually placed and got an outcome on.
|
||||
pdf = df[(df["playable"] == True) & (df["won"].notna())].copy()
|
||||
pdf["won"] = pdf["won"].astype(bool)
|
||||
pdf["calibrated_confidence"] = pdf["calibrated_confidence"].fillna(0)
|
||||
pdf["ev_edge"] = pdf["ev_edge"].fillna(0)
|
||||
pdf["odds"] = pdf["odds"].fillna(0)
|
||||
pdf["odds_reliability"] = pdf["odds_reliability"].fillna(0.35)
|
||||
return pdf
|
||||
|
||||
|
||||
def evaluate(pdf: pd.DataFrame, mask) -> Dict:
|
||||
kept = pdf[mask]
|
||||
if len(kept) == 0:
|
||||
return {"n": 0, "hit_pct": 0, "profit": 0, "staked": 0, "roi_pct": 0}
|
||||
wins = kept["won"].sum()
|
||||
profit = kept["unit_profit"].sum()
|
||||
staked = kept["stake_units"].sum()
|
||||
return {
|
||||
"n": int(len(kept)),
|
||||
"hit_pct": round(100.0 * wins / len(kept), 2),
|
||||
"profit": round(profit, 3),
|
||||
"staked": round(staked, 3),
|
||||
"roi_pct": round(100.0 * profit / staked, 2) if staked else 0,
|
||||
}
|
||||
|
||||
|
||||
def grid_search_market(
|
||||
market_df: pd.DataFrame,
|
||||
market: str,
|
||||
min_bets: int = 15,
|
||||
) -> List[Dict]:
|
||||
"""Try a wide grid of (min_conf, min_edge, max_edge, min_odds, max_odds,
|
||||
min_reliability) combinations. Return all candidates with n >= min_bets,
|
||||
sorted by ROI descending."""
|
||||
|
||||
conf_options = [0, 45, 50, 55, 60, 65, 70]
|
||||
min_edge_options = [-1.0, -0.05, 0.0, 0.03, 0.05, 0.08]
|
||||
max_edge_options = [10.0, 0.30, 0.20, 0.15, 0.10]
|
||||
min_odds_options = [1.20, 1.30, 1.40, 1.50, 1.60, 1.80]
|
||||
max_odds_options = [10.0, 3.0, 2.5, 2.2, 2.0]
|
||||
rel_options = [0.0, 0.30, 0.45, 0.55]
|
||||
consensus_options = ["any", "agree_or_null"]
|
||||
|
||||
candidates: List[Dict] = []
|
||||
for mc, mine, maxe, mino, maxo, mrel, cons in itertools.product(
|
||||
conf_options, min_edge_options, max_edge_options,
|
||||
min_odds_options, max_odds_options, rel_options, consensus_options,
|
||||
):
|
||||
if mine >= maxe or mino >= maxo:
|
||||
continue
|
||||
mask = (
|
||||
(market_df["calibrated_confidence"] >= mc)
|
||||
& (market_df["ev_edge"] >= mine)
|
||||
& (market_df["ev_edge"] <= maxe)
|
||||
& (market_df["odds"] >= mino)
|
||||
& (market_df["odds"] <= maxo)
|
||||
& (market_df["odds_reliability"] >= mrel)
|
||||
)
|
||||
if cons == "agree_or_null":
|
||||
mask &= market_df["v27_consensus"] != "DISAGREE"
|
||||
result = evaluate(market_df, mask)
|
||||
if result["n"] >= min_bets:
|
||||
candidates.append({
|
||||
"market": market,
|
||||
"min_conf": mc,
|
||||
"min_edge": mine,
|
||||
"max_edge": maxe,
|
||||
"min_odds": mino,
|
||||
"max_odds": maxo,
|
||||
"min_reliability": mrel,
|
||||
"consensus": cons,
|
||||
**result,
|
||||
})
|
||||
candidates.sort(key=lambda r: (r["roi_pct"], r["n"]), reverse=True)
|
||||
return candidates
|
||||
|
||||
|
||||
def baseline(pdf: pd.DataFrame, market: str) -> Dict:
|
||||
m = pdf[pdf["market"] == market]
|
||||
return evaluate(m, pd.Series([True] * len(m), index=m.index))
|
||||
|
||||
|
||||
def main():
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument("--csv", default=None, help="Override CSV path")
|
||||
parser.add_argument("--min-bets", type=int, default=15,
|
||||
help="Min bet count to consider a config valid")
|
||||
parser.add_argument("--top-k", type=int, default=3,
|
||||
help="Show top K configs per market")
|
||||
parser.add_argument("--apply", action="store_true",
|
||||
help="Patch config/market_thresholds.json with winners")
|
||||
args = parser.parse_args()
|
||||
|
||||
csv_path = args.csv or latest_csv()
|
||||
if not csv_path or not os.path.exists(csv_path):
|
||||
print("No backtest CSV found.")
|
||||
return
|
||||
print(f"Loading: {csv_path}")
|
||||
pdf = load_backtest(csv_path)
|
||||
print(f"Playable + settled bets: {len(pdf)}")
|
||||
|
||||
markets = sorted(pdf["market"].dropna().unique())
|
||||
print(f"Markets: {markets}\n")
|
||||
|
||||
all_winners: Dict[str, Dict] = {}
|
||||
|
||||
for market in markets:
|
||||
market_df = pdf[pdf["market"] == market]
|
||||
n_total = len(market_df)
|
||||
base = baseline(pdf, market)
|
||||
|
||||
print(f"\n{'=' * 78}")
|
||||
print(f"MARKET: {market} (n={n_total} baseline_roi={base['roi_pct']}%)")
|
||||
print(f"{'=' * 78}")
|
||||
|
||||
if n_total < args.min_bets * 2:
|
||||
print(f" Sample too small to grid-search reliably (n={n_total}). Skip.")
|
||||
continue
|
||||
|
||||
candidates = grid_search_market(market_df, market, args.min_bets)
|
||||
if not candidates:
|
||||
print(f" No config kept >= {args.min_bets} bets. Skip.")
|
||||
continue
|
||||
|
||||
# Pareto-ish: show top-K by ROI but also one that keeps higher bet count
|
||||
winners = candidates[:args.top_k]
|
||||
keep_high_volume = None
|
||||
for c in candidates:
|
||||
if c["n"] >= max(40, n_total // 3) and c["roi_pct"] > base["roi_pct"]:
|
||||
keep_high_volume = c
|
||||
break
|
||||
|
||||
print(f" {'rank':<5}{'n':>5}{'hit%':>7}{'roi%':>8} "
|
||||
f"{'min_conf':>9}{'min_edge':>10}{'max_edge':>10}"
|
||||
f"{'min_odds':>10}{'max_odds':>10}{'min_rel':>9}{'cons':>15}")
|
||||
for i, w in enumerate(winners, 1):
|
||||
print(f" {i:<5}{w['n']:>5}{w['hit_pct']:>7}{w['roi_pct']:>+8}"
|
||||
f" {w['min_conf']:>9}{w['min_edge']:>+10.3f}{w['max_edge']:>+10.3f}"
|
||||
f"{w['min_odds']:>10.2f}{w['max_odds']:>10.2f}"
|
||||
f"{w['min_reliability']:>9.2f}{w['consensus']:>15}")
|
||||
if keep_high_volume and keep_high_volume not in winners:
|
||||
print(f" high {keep_high_volume['n']:>5}{keep_high_volume['hit_pct']:>7}"
|
||||
f"{keep_high_volume['roi_pct']:>+8}"
|
||||
f" {keep_high_volume['min_conf']:>9}"
|
||||
f"{keep_high_volume['min_edge']:>+10.3f}"
|
||||
f"{keep_high_volume['max_edge']:>+10.3f}"
|
||||
f"{keep_high_volume['min_odds']:>10.2f}"
|
||||
f"{keep_high_volume['max_odds']:>10.2f}"
|
||||
f"{keep_high_volume['min_reliability']:>9.2f}"
|
||||
f"{keep_high_volume['consensus']:>15}")
|
||||
|
||||
# Pick a "good" recommendation: best ROI with n >= min_bets
|
||||
# If best ROI is still negative, flag the market as unprofitable.
|
||||
best = winners[0]
|
||||
all_winners[market] = best
|
||||
if best["roi_pct"] <= 0:
|
||||
print(f" ⚠️ Best config still loses money (ROI={best['roi_pct']}%) "
|
||||
f"— consider muting this market entirely.")
|
||||
else:
|
||||
print(f" ✅ Best config: ROI={best['roi_pct']}% on {best['n']} bets "
|
||||
f"(vs baseline {base['roi_pct']}% on {n_total}).")
|
||||
|
||||
# ─── Aggregate impact ────────────────────────────────────────────────
|
||||
print(f"\n{'=' * 78}")
|
||||
print("AGGREGATE IMPACT (if we apply each market's best config)")
|
||||
print(f"{'=' * 78}")
|
||||
total_old_bets = total_old_profit = total_old_staked = 0
|
||||
total_new_bets = total_new_profit = total_new_staked = 0
|
||||
for market, win in all_winners.items():
|
||||
base = baseline(pdf, market)
|
||||
total_old_bets += base["n"]
|
||||
total_old_profit += base["profit"]
|
||||
total_old_staked += base["staked"]
|
||||
total_new_bets += win["n"]
|
||||
total_new_profit += win["profit"]
|
||||
total_new_staked += win["staked"]
|
||||
base_roi = 100.0 * total_old_profit / total_old_staked if total_old_staked else 0
|
||||
new_roi = 100.0 * total_new_profit / total_new_staked if total_new_staked else 0
|
||||
print(f" Baseline: {total_old_bets:>4} bets, "
|
||||
f"profit={total_old_profit:+.2f}u, ROI={base_roi:+.2f}%")
|
||||
print(f" Optimized: {total_new_bets:>4} bets, "
|
||||
f"profit={total_new_profit:+.2f}u, ROI={new_roi:+.2f}%")
|
||||
print(f" Δ: {total_new_bets - total_old_bets:+d} bets, "
|
||||
f"{total_new_profit - total_old_profit:+.2f}u, "
|
||||
f"{new_roi - base_roi:+.2f}pp")
|
||||
|
||||
# ─── Write JSON patch ────────────────────────────────────────────────
|
||||
patch_path = os.path.join(REPORTS_DIR, "filter_optimization_patch.json")
|
||||
patch = {market: {
|
||||
"min_calibrated_confidence": win["min_conf"],
|
||||
"min_ev_edge": win["min_edge"],
|
||||
"max_ev_edge": win["max_edge"],
|
||||
"min_odds": win["min_odds"],
|
||||
"max_odds": win["max_odds"],
|
||||
"min_odds_reliability": win["min_reliability"],
|
||||
"require_v27_agree": win["consensus"] == "agree_or_null",
|
||||
"expected_n_bets": win["n"],
|
||||
"expected_hit_pct": win["hit_pct"],
|
||||
"expected_roi_pct": win["roi_pct"],
|
||||
} for market, win in all_winners.items()}
|
||||
with open(patch_path, "w", encoding="utf-8") as f:
|
||||
json.dump(patch, f, indent=2)
|
||||
print(f"\nPatch saved: {patch_path}")
|
||||
|
||||
if args.apply:
|
||||
print("\n--apply flag set. Patching not implemented yet — "
|
||||
"review the patch JSON and update config/market_thresholds.json manually.")
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
@@ -0,0 +1,54 @@
|
||||
"""Smoke test for the score-coherence filter using the LAFC vs Sounders
|
||||
1-0 scenario from production. Verifies that markets that contradict the
|
||||
predicted score are correctly excluded from the coherent set, and that
|
||||
the markets the model got right are all included.
|
||||
"""
|
||||
import os, sys
|
||||
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
||||
from services.betting_brain import BettingBrain
|
||||
|
||||
brain = BettingBrain()
|
||||
pkg = {
|
||||
"score_prediction": {"ft": "1-0", "ht": "0-0"},
|
||||
}
|
||||
coh = brain._score_consistent_markets(pkg)
|
||||
print(f"Predicted: 1-0 (HT 0-0)")
|
||||
print(f"Coherent set size: {len(coh)}")
|
||||
print()
|
||||
|
||||
# Each pick the system actually offered for the LAFC match, with whether
|
||||
# it was the *actual* winning pick.
|
||||
test_picks = [
|
||||
("MS", "1", True, "correct"),
|
||||
("MS", "2", False, "wrong"),
|
||||
("MS", "X", False, "wrong"),
|
||||
("DC", "1X", True, "correct"),
|
||||
("DC", "12", True, "correct"),
|
||||
("DC", "X2", False, "wrong"),
|
||||
("OU25", "Üst", False, "WRONG — system featured this"),
|
||||
("OU25", "Alt", True, "correct"),
|
||||
("OU35", "Alt", True, "correct"),
|
||||
("OU35", "Üst", False, "wrong"),
|
||||
("BTTS", "Var", False, "wrong"),
|
||||
("BTTS", "Yok", True, "correct"),
|
||||
("HT", "X", True, "correct"),
|
||||
("HT", "1", False, "wrong"),
|
||||
("HTFT", "X/1", True, "correct"),
|
||||
("HTFT", "1/1", False, "wrong (HT was 0-0)"),
|
||||
("HT_OU05", "Üst", False, "wrong"),
|
||||
("HT_OU05", "Alt", True, "correct"),
|
||||
("OE", "Çift", False, "wrong (1 is odd)"),
|
||||
("OE", "Tek", True, "correct"),
|
||||
]
|
||||
|
||||
print(f"{'market':<10}{'pick':<10}{'real-win?':<12}{'in-coherent?':<14}{'match?'}")
|
||||
print("-" * 60)
|
||||
ok = 0
|
||||
for market, pick, would_win, note in test_picks:
|
||||
in_coh = (market, pick) in coh
|
||||
match = "✓" if in_coh == would_win else "✗ MISMATCH"
|
||||
if in_coh == would_win: ok += 1
|
||||
print(f"{market:<10}{pick:<10}{str(would_win):<12}{str(in_coh):<14}{match} {note}")
|
||||
|
||||
print()
|
||||
print(f"Result: {ok}/{len(test_picks)} picks correctly classified")
|
||||
@@ -39,6 +39,31 @@ class BettingBrain:
|
||||
|
||||
SNIPER_BLOCKED_MARKETS = {"HT", "HTFT", "OE", "CARDS", "HT_OU05", "HT_OU15"}
|
||||
|
||||
# Markets that lose money under every filter combination per the
|
||||
# diagnostic backtest (1000 matches). Until calibration is rebuilt for
|
||||
# these specifically, force NO_BET. Re-evaluate after each backtest run.
|
||||
MUTED_MARKETS = {"BTTS"}
|
||||
|
||||
# Per-market optimal filter envelopes derived from the diagnostic
|
||||
# backtest grid search (reports/filter_optimization_patch.json). Any
|
||||
# pick falling OUTSIDE this envelope is vetoed. Tightens the playable
|
||||
# band to the ROI-positive zone identified empirically.
|
||||
#
|
||||
# Each entry: {min_conf, min_edge, max_edge, min_odds, max_odds,
|
||||
# min_reliability, require_v27_agree}
|
||||
MARKET_OPTIMAL_FILTERS = {
|
||||
"MS": {
|
||||
"min_edge": -0.05, "max_edge": 0.15,
|
||||
"min_odds": 1.20, "max_odds": 10.0,
|
||||
"min_reliability": 0.0, "require_v27_agree": True,
|
||||
},
|
||||
"OU25": {
|
||||
"min_edge": -1.0, "max_edge": 0.15,
|
||||
"min_odds": 1.80, "max_odds": 10.0,
|
||||
"min_reliability": 0.0, "require_v27_agree": False,
|
||||
},
|
||||
}
|
||||
|
||||
MARKET_PRIORS = {
|
||||
"DC": 4.0,
|
||||
"OU15": 3.0,
|
||||
@@ -86,6 +111,36 @@ class BettingBrain:
|
||||
watchlist.sort(key=self._candidate_sort_key, reverse=True)
|
||||
no_value.sort(key=self._candidate_sort_key, reverse=True)
|
||||
|
||||
# ── SCORE COHERENCE FILTER ──────────────────────────────────────
|
||||
# If the model also produced a score prediction (e.g. 1-0), pick
|
||||
# main_pick from the subset of candidates that would WIN at that
|
||||
# score. Stops the system from recommending OU25 Üst while also
|
||||
# predicting 1-0 (only 1 goal). Falls back to original list if no
|
||||
# coherent candidate exists.
|
||||
coherent_set = self._score_consistent_markets(guarded)
|
||||
coherent_flag = False
|
||||
if coherent_set:
|
||||
def is_coherent(row: Dict[str, Any]) -> bool:
|
||||
m = str(row.get("market") or "")
|
||||
p = str(row.get("pick") or "")
|
||||
return (m, p) in coherent_set
|
||||
|
||||
approved_coh = [r for r in approved if is_coherent(r)]
|
||||
watchlist_coh = [r for r in watchlist if is_coherent(r)]
|
||||
|
||||
if approved_coh:
|
||||
approved = approved_coh
|
||||
coherent_flag = True
|
||||
elif watchlist_coh:
|
||||
# No coherent BET candidates — at least promote a coherent
|
||||
# watch over an incoherent BET.
|
||||
watchlist = watchlist_coh + [r for r in watchlist if not is_coherent(r)]
|
||||
coherent_flag = True
|
||||
# Tag every row so the UI/diagnostics can see what happened
|
||||
for row in judged_rows.values():
|
||||
row.setdefault("betting_brain", {})
|
||||
row["betting_brain"]["score_coherent"] = is_coherent(row)
|
||||
|
||||
original_main = guarded.get("main_pick") or {}
|
||||
main_pick = None
|
||||
decision = "NO_BET"
|
||||
@@ -142,10 +197,11 @@ class BettingBrain:
|
||||
|
||||
rejected = [d for d in decisions if d.get("action") == "REJECT"]
|
||||
guarded["betting_brain"] = {
|
||||
"version": "judge-v1",
|
||||
"version": "judge-v2-score-coherent",
|
||||
"decision": decision,
|
||||
"reason": decision_reason,
|
||||
"main_pick_key": main_key or None,
|
||||
"score_coherent_filter_applied": coherent_flag,
|
||||
"approved_count": len(approved),
|
||||
"watchlist_count": len(watchlist),
|
||||
"rejected_count": len(rejected),
|
||||
@@ -243,6 +299,44 @@ class BettingBrain:
|
||||
if play_score < 50.0 and not is_value_sniper:
|
||||
vetoes.append("play_score_too_low")
|
||||
|
||||
# ── HARD EV-EDGE VETO ───────────────────────────────────────────
|
||||
# Diagnostic backtest (1000 maç, 524 settled bet) gösterdi ki
|
||||
# ev_edge < 0 olan bahisler %76 of all picks ve ROI yaklaşık -%16.
|
||||
# ev_edge < 0 = "model market'in altında olasılık veriyor" = vig'i
|
||||
# yiyemeyeceğimiz negative-EV bahis. Hard veto: oynama.
|
||||
# Sniper override hâlâ geçer (yüksek convicted alternatif pick'ler).
|
||||
if ev_edge < 0.0 and not is_value_sniper:
|
||||
vetoes.append("negative_ev_edge")
|
||||
issues.append(f"ev_edge={ev_edge:.3f}_below_zero")
|
||||
# Trap edge: bizim diagnostic backtest'te ev_edge >= 0.20 olan tüm
|
||||
# bahisler kaybediyordu (n=10, hepsi -%25+ ROI). Model market'i bu
|
||||
# kadar yanlış buluyorsa muhtemelen modelin kendisinin yanlış olduğu
|
||||
# bir senaryo (eksik info, tuhaf maç, vs.) — oynama.
|
||||
if ev_edge >= 0.20 and not is_value_sniper:
|
||||
vetoes.append("ev_edge_too_high_trap")
|
||||
issues.append(f"ev_edge={ev_edge:.3f}_trap_range")
|
||||
|
||||
# ── MUTED MARKETS (grid search showed no profitable config) ──
|
||||
if market in self.MUTED_MARKETS and not is_value_sniper:
|
||||
vetoes.append("market_muted_by_backtest")
|
||||
issues.append(f"market_{market}_muted")
|
||||
|
||||
# ── PER-MARKET OPTIMAL ENVELOPE (from grid search) ──
|
||||
envelope = self.MARKET_OPTIMAL_FILTERS.get(market)
|
||||
if envelope and not is_value_sniper:
|
||||
if ev_edge < envelope["min_edge"]:
|
||||
vetoes.append("outside_envelope_edge_low")
|
||||
if ev_edge > envelope["max_edge"]:
|
||||
vetoes.append("outside_envelope_edge_high")
|
||||
if odds and odds < envelope["min_odds"]:
|
||||
vetoes.append("outside_envelope_odds_low")
|
||||
if odds and odds > envelope["max_odds"]:
|
||||
vetoes.append("outside_envelope_odds_high")
|
||||
if odds_rel < envelope["min_reliability"]:
|
||||
vetoes.append("outside_envelope_reliability_low")
|
||||
if envelope["require_v27_agree"] and consensus != "AGREE":
|
||||
vetoes.append("outside_envelope_v27_must_agree")
|
||||
|
||||
if divergence is not None:
|
||||
if divergence >= self.HARD_DIVERGENCE and not is_value_sniper:
|
||||
score -= 42.0
|
||||
@@ -635,6 +729,112 @@ class BettingBrain:
|
||||
return self._safe_float(ou25.get(key)) if key else None
|
||||
return None
|
||||
|
||||
def _score_consistent_markets(self, package: Dict[str, Any]) -> Optional[set]:
|
||||
"""Build the set of (market, pick) tuples that WOULD WIN if the
|
||||
model's own score prediction came true. We use this as a coherence
|
||||
gate: if the model is confident about a 1-0 outcome but also wants
|
||||
to play OU25 Üst, those two beliefs contradict each other — and the
|
||||
score prediction is the more informative one because it aggregates
|
||||
all market signals into a single most-likely scenario.
|
||||
|
||||
Returns None if the score prediction is missing or malformed; in
|
||||
that case we skip the coherence check.
|
||||
"""
|
||||
score_pred = package.get("score_prediction") or {}
|
||||
ft_raw = str(score_pred.get("ft") or score_pred.get("full_time") or "").strip()
|
||||
ht_raw = str(score_pred.get("ht") or score_pred.get("half_time") or "").strip()
|
||||
|
||||
def parse(s: str) -> Optional[Tuple[int, int]]:
|
||||
for sep in ("-", ":", "–"):
|
||||
if sep in s:
|
||||
parts = s.split(sep, 1)
|
||||
try:
|
||||
return int(parts[0].strip()), int(parts[1].strip())
|
||||
except (ValueError, IndexError):
|
||||
return None
|
||||
return None
|
||||
|
||||
ft = parse(ft_raw)
|
||||
if ft is None:
|
||||
return None
|
||||
ht = parse(ht_raw)
|
||||
|
||||
fh, fa = ft
|
||||
total = fh + fa
|
||||
consistent: set = set()
|
||||
|
||||
# MS / 1X2 — single outcome
|
||||
if fh > fa:
|
||||
consistent.add(("MS", "1"))
|
||||
consistent.add(("ML", "1"))
|
||||
elif fh < fa:
|
||||
consistent.add(("MS", "2"))
|
||||
consistent.add(("ML", "2"))
|
||||
else:
|
||||
consistent.add(("MS", "X"))
|
||||
consistent.add(("ML", "X"))
|
||||
|
||||
# DC — two of three legs win at any score
|
||||
if fh >= fa:
|
||||
consistent.add(("DC", "1X"))
|
||||
if fh <= fa:
|
||||
consistent.add(("DC", "X2"))
|
||||
if fh != fa:
|
||||
consistent.add(("DC", "12"))
|
||||
|
||||
# Over/Under main lines
|
||||
for line, market in ((0.5, "OU05"), (1.5, "OU15"),
|
||||
(2.5, "OU25"), (3.5, "OU35"), (4.5, "OU45")):
|
||||
if total > line:
|
||||
for p in ("Üst", "Ust", "Over", "OVER"):
|
||||
consistent.add((market, p))
|
||||
elif total < line:
|
||||
for p in ("Alt", "Under", "UNDER"):
|
||||
consistent.add((market, p))
|
||||
# total == line → push, neither side wins → don't add
|
||||
|
||||
# BTTS — both teams score
|
||||
if fh > 0 and fa > 0:
|
||||
for p in ("Var", "KG Var", "Yes", "YES"):
|
||||
consistent.add(("BTTS", p))
|
||||
else:
|
||||
for p in ("Yok", "KG Yok", "No", "NO"):
|
||||
consistent.add(("BTTS", p))
|
||||
|
||||
# OE — total goals odd/even
|
||||
if total % 2 == 1:
|
||||
for p in ("Tek", "Odd", "ODD"):
|
||||
consistent.add(("OE", p))
|
||||
else:
|
||||
for p in ("Çift", "Cift", "Even", "EVEN"):
|
||||
consistent.add(("OE", p))
|
||||
|
||||
# HT-only markets (need HT score)
|
||||
if ht is not None:
|
||||
hh, ha = ht
|
||||
ht_total = hh + ha
|
||||
if hh > ha:
|
||||
consistent.add(("HT", "1"))
|
||||
elif hh < ha:
|
||||
consistent.add(("HT", "2"))
|
||||
else:
|
||||
consistent.add(("HT", "X"))
|
||||
for line, market in ((0.5, "HT_OU05"), (1.5, "HT_OU15"), (2.5, "HT_OU25")):
|
||||
if ht_total > line:
|
||||
for p in ("Üst", "Ust", "Over"):
|
||||
consistent.add((market, p))
|
||||
elif ht_total < line:
|
||||
for p in ("Alt", "Under"):
|
||||
consistent.add((market, p))
|
||||
|
||||
# HTFT — single combo
|
||||
ht_o = "1" if hh > ha else "2" if hh < ha else "X"
|
||||
ft_o = "1" if fh > fa else "2" if fh < fa else "X"
|
||||
consistent.add(("HTFT", f"{ht_o}/{ft_o}"))
|
||||
consistent.add(("HTFT", f"{ht_o}{ft_o}"))
|
||||
|
||||
return consistent
|
||||
|
||||
def _triple_value(self, package: Dict[str, Any], key: Optional[str]) -> Optional[Dict[str, Any]]:
|
||||
if not key:
|
||||
return None
|
||||
|
||||
@@ -449,6 +449,12 @@ class DataLoaderMixin:
|
||||
return 1.5, 1.2
|
||||
return weighted_for / total_weight, weighted_against / total_weight
|
||||
|
||||
# Approximate European season window — Eredivisie/PL/La Liga start late
|
||||
# July / mid-August, end May. Using 300 days as a buffer covers most
|
||||
# competitions while excluding "career points" from previous seasons.
|
||||
# When a proper seasons table lands this should query season boundaries.
|
||||
_SEASON_LOOKBACK_MS = 300 * 24 * 60 * 60 * 1000
|
||||
|
||||
def _estimate_league_position(
|
||||
self,
|
||||
cur: RealDictCursor,
|
||||
@@ -458,6 +464,7 @@ class DataLoaderMixin:
|
||||
) -> int:
|
||||
if not team_id or not league_id:
|
||||
return 10
|
||||
season_start_ms = before_date_ms - self._SEASON_LOOKBACK_MS
|
||||
try:
|
||||
cur.execute(
|
||||
"""
|
||||
@@ -478,6 +485,7 @@ class DataLoaderMixin:
|
||||
AND m.score_home IS NOT NULL
|
||||
AND m.score_away IS NOT NULL
|
||||
AND m.mst_utc < %s
|
||||
AND m.mst_utc >= %s
|
||||
UNION ALL
|
||||
SELECT
|
||||
m.away_team_id AS team_id,
|
||||
@@ -492,11 +500,15 @@ class DataLoaderMixin:
|
||||
AND m.score_home IS NOT NULL
|
||||
AND m.score_away IS NOT NULL
|
||||
AND m.mst_utc < %s
|
||||
AND m.mst_utc >= %s
|
||||
) tm
|
||||
GROUP BY tm.team_id
|
||||
ORDER BY points DESC
|
||||
""",
|
||||
(league_id, before_date_ms, league_id, before_date_ms),
|
||||
(
|
||||
league_id, before_date_ms, season_start_ms,
|
||||
league_id, before_date_ms, season_start_ms,
|
||||
),
|
||||
)
|
||||
rows = cur.fetchall()
|
||||
if not rows:
|
||||
|
||||
@@ -225,20 +225,43 @@ class FeatureBuilderMixin:
|
||||
if enrichment_failures:
|
||||
print(f"⚠️ Enrichment partial failures for {data.match_id}: {', '.join(enrichment_failures)}")
|
||||
|
||||
# ── Cup game detection (used by upset engine + elo dampening below) ──
|
||||
_league_name_lower = (getattr(data, 'league_name', '') or '').lower()
|
||||
_cup_keywords = ("kupa", "cup", "coupe", "copa", "coppa", "pokal",
|
||||
"trophy", "shield", "ziraat", "süper kupa", "super cup",
|
||||
"beker", "taça", "taca")
|
||||
_is_cup_match = any(kw in _league_name_lower for kw in _cup_keywords)
|
||||
|
||||
# ── League size hint: top European leagues 18-20 teams, lower 16-24 ──
|
||||
# We don't have a per-league team count, so fall back to 20 (standard).
|
||||
# When standings infra lands this should pull from seasons table.
|
||||
_league_total_teams = 20
|
||||
|
||||
# Upset engine features
|
||||
upset_atmosphere, upset_motivation, upset_fatigue = 0.0, 0.0, 0.0
|
||||
try:
|
||||
upset_engine = get_upset_engine()
|
||||
# Use the real position estimates from data_loader; fall back to mid-
|
||||
# table (10) only when the loader couldn't compute one. Hardcoding 10
|
||||
# for every team made motivation_score collapse to 0 for everyone.
|
||||
_home_pos = getattr(data, 'home_position', None)
|
||||
_away_pos = getattr(data, 'away_position', None)
|
||||
if _home_pos is None or _home_pos <= 0:
|
||||
_home_pos = 10
|
||||
if _away_pos is None or _away_pos <= 0:
|
||||
_away_pos = 10
|
||||
upset_feats = upset_engine.get_features(
|
||||
home_team_name=getattr(data, 'home_team_name', '') or '',
|
||||
home_team_id=data.home_team_id,
|
||||
away_team_name=getattr(data, 'away_team_name', '') or '',
|
||||
league_name=getattr(data, 'league_name', '') or '',
|
||||
home_position=10,
|
||||
away_position=10,
|
||||
home_position=_home_pos,
|
||||
away_position=_away_pos,
|
||||
match_date_ms=data.match_date_ms,
|
||||
is_cup_match=_is_cup_match,
|
||||
home_days_rest=int(home_rest),
|
||||
away_days_rest=int(away_rest),
|
||||
total_teams=_league_total_teams,
|
||||
)
|
||||
upset_atmosphere = upset_feats.get('upset_atmosphere', 0.0)
|
||||
upset_motivation = upset_feats.get('upset_motivation', 0.0)
|
||||
@@ -276,15 +299,10 @@ class FeatureBuilderMixin:
|
||||
is_season_start = 1.0 if match_month in (7, 8, 9) else 0.0
|
||||
is_season_end = 1.0 if match_month in (5, 6) else 0.0
|
||||
|
||||
# ── Cup game detection: dampen home advantage in feature space ──
|
||||
_league_name = (getattr(data, 'league_name', '') or '').lower()
|
||||
_cup_keywords = ("kupa", "cup", "coupe", "copa", "coppa", "pokal",
|
||||
"trophy", "shield", "ziraat", "süper kupa", "super cup")
|
||||
_is_cup = any(kw in _league_name for kw in _cup_keywords)
|
||||
|
||||
# ── Derived / Interaction features (V27) ──
|
||||
# Cup games: home ELO advantage is ~30% weaker (rotation, lower motivation)
|
||||
elo_diff = (home_elo - away_elo) * (0.70 if _is_cup else 1.0)
|
||||
# Uses _is_cup_match computed earlier (before upset engine call).
|
||||
elo_diff = (home_elo - away_elo) * (0.70 if _is_cup_match else 1.0)
|
||||
form_elo_diff = home_form_elo_val - away_form_elo_val
|
||||
attack_vs_defense_home = data.home_goals_avg - data.away_conceded_avg
|
||||
attack_vs_defense_away = data.away_goals_avg - data.home_conceded_avg
|
||||
|
||||
Reference in New Issue
Block a user