Deep slate diagnosis — 20260916

• Games: 15 · hits 10 · misses 5 · winner accuracy 0.6667
• Mean total-run bias on losses: 3.2 (negative = under-projected)
• Mean team-run error on losses: 6.4 · mean Brier on losses: 0.314

Loss primary drivers

• Small edge on the losing side: 1
• Run environment far lower than projected (over-dispersed): 1
• Starting-pitcher projection was wrong: 1
• Backed the wrong side with real conviction: 2

High-level causal issue register

[HIGH] An intermediate stage materially degraded winner direction

• Domain: pipeline_stages · affected: BOS@TEX, KC@HOU, SEA@LAA
• Evidence: historical_ml → variance_applied changed winner accuracy by -0.2000 (-3 hits)
• Action: List the exact games whose winner flipped at this stage, and keep both the before-stage and after-stage picks in the trace.
• Do not infer: Even if a later stage recovers the pick, this unstable step is still a risk; do not treat it as harmless.

[HIGH] Published confidence buckets were poorly aligned on this slate

• Domain: calibration · affected: ATL@CHC, BAL@NYM, BOS@TEX, CHW@CLE, DET@TOR, KC@HOU, LAD@CIN, MIA@ARI, MIL@PIT, NYY@MIN, OAK@TB, PHI@WAS, SD@COL, SEA@LAA, SF@STL
• Evidence: weighted absolute calibration error 0.1661 across 15 games
• Action: Watch the same confidence buckets over the full rolling window before changing any calibration weight.
• Do not infer: A single slate has far too few games to establish a stable calibration curve.

[HIGH] Validated run models explain very little holdout variance

• Domain: model_validation · affected: ATL@CHC, BAL@NYM, BOS@TEX, CHW@CLE, DET@TOR, KC@HOU, LAD@CIN, MIA@ARI, MIL@PIT, NYY@MIN, OAK@TB, PHI@WAS, SD@COL, SEA@LAA, SF@STL
• Evidence: home/away/total R² = 0.0021/0.0200/0.0035
• Action: Treat exact run scores as low-confidence estimates, and look for stronger features or a better target before tuning the score tails.
• Do not infer: A small positive R² only means the model beats guessing the average; it does not make the exact scores reliable.

[MEDIUM] Most high-raw picks lost before/after shrinkage

• Domain: calibration · affected: LAD@CIN, NYY@MIN, OAK@TB
• Evidence: 1/3 games with raw pick probability ≥0.65 won
• Action: Track how often high-raw-probability picks lose across multiple slates, and keep the raw to calibrated to published lineage.
• Do not infer: This one slate is not enough to justify a new per-bucket shrinkage rule.

[INFO] Every settled game received structural QA

• Domain: coverage · affected: ATL@CHC, BAL@NYM, BOS@TEX, CHW@CLE, DET@TOR, KC@HOU, LAD@CIN, MIA@ARI, MIL@PIT, NYY@MIN, OAK@TB, PHI@WAS, SD@COL, SEA@LAA, SF@STL
• Evidence: 15/15 games structurally audited
• Action: Keep the per-game audit receipt in every report.
• Do not infer: Full audit coverage does not cancel the findings raised inside the audited games.

[INFO] Material calibrated-to-published movements have recorded lineage

• Domain: probability_lineage · affected: ATL@CHC, BAL@NYM, CHW@CLE, DET@TOR, LAD@CIN, MIL@PIT, NYY@MIN, PHI@WAS, SD@COL
• Evidence: 9 material movement(s), all explained by named score-coherence and/or recency transformations
• Action: Keep showing every probability layer and the named transformation between them.
• Do not infer: A published probability above the calibrated one is not a governor bypass when a recorded transformation explains it.

[INFO] Simulation produced no winner-direction intervention on this slate

• Domain: simulation · affected: ATL@CHC, BAL@NYM, BOS@TEX, CHW@CLE, DET@TOR, KC@HOU, LAD@CIN, MIA@ARI, MIL@PIT, NYY@MIN, OAK@TB, PHI@WAS, SD@COL, SEA@LAA, SF@STL
• Evidence: flip candidates 0; published flips 0
• Action: Measure the simulation's contribution across rolling slates, separating score realization from winner flipping.
• Do not infer: Zero winner flips is not a defect and does not mean the simulation did no score-reconciliation work.

Cross-slate trends (persistent-pattern memory)

• Window: 20260906, 20260907, 20260908, 20260909, 20260910, 20260911, 20260912, 20260913, 20260914, 20260915, 20260916 (10 prior slates loaded)
• Highest-confidence pick loss streak: 0 slate(s)
• Actionable pattern detected: True

Escalated persistent patterns

• simulation_inert_on_slate — 11 consecutive slates (11 in window)
• Target lever(s): simulation.allow_score
• Direction: confirm the simulation module is running, not silently no-oping
• Guardrail: simulation_inert_on_slate has persisted 11 consecutive slates; this is no longer single-slate noise. Treat it as an actionable pattern and raise the code/diagnostic action named; do not force a parameter move.
• single_slate_calibration_instability — 3 consecutive slates (7 in window)
• Target lever(s): confidence.empirical_max_weight, confidence.base_cap, confidence.moderate_cap
• Direction: increase empirical shrinkage
• Guardrail: single_slate_calibration_instability has persisted 3 consecutive slates; this is no longer single-slate noise. Treat it as an actionable pattern and propose a bounded shadow experiment against a target lever.
• weak_run_model_signal — 11 consecutive slates (11 in window)
• Target lever(s): training.xgb.n_estimators, training.xgb.max_depth, training.xgb.learning_rate
• Direction: investigate features/target via isolated training backtest (manual)
• Guardrail: weak_run_model_signal has persisted 11 consecutive slates; this is no longer single-slate noise. Treat it as an actionable pattern and raise the code/diagnostic action named; do not force a parameter move.

Metric direction across the window

• winner_accuracy: improving · series [0.6, 0.5455, 0.7333, 0.6667, 0.6, 0.6667, 0.5333, 0.6, 0.7, 0.4, 0.6667]
• weighted_calibration_error: improving · series [0.0937, 0.2611, 0.1825, 0.1069, 0.1171, 0.2091, 0.1711, 0.0827, 0.2812, 0.2795, 0.1661]
• margin_bias: improving · series [-1.467, -0.182, -0.867, -2.133, -1.0, -2.933, -2.667, -1.8, -0.4, -1.667, -1.133]
• mean_team_run_error_on_losses: regressing · series [5.333, 3.8, 6.75, 5.6, 6.5, 5.0, 6.143, 5.0, 6.333, 5.556, 6.4]

Pipeline stage quality

• historical_ml: winner 9/15 (0.6), team-run MAE 1.8428, margin MAE 3.0594
• variance_applied: winner 6/15 (0.4), team-run MAE 1.8667, margin MAE 2.9333
• enhanced_prediction: winner 11/15 (0.7333), team-run MAE 1.9333, margin MAE 3.2
• embedded_savant_run_environment: winner 11/15 (0.7333), team-run MAE 2.0333, margin MAE 3.4
• fundamental_edge_target_review: winner 10/15 (0.6667), team-run MAE 2.1, margin MAE 3.5333
• capped_target: winner 10/15 (0.6667), team-run MAE 2.1, margin MAE 3.5333
• simulation_result: winner 10/15 (0.6667), team-run MAE 2.1, margin MAE 3.5333
• published_final: winner 10/15 (0.6667), team-run MAE 2.1, margin MAE 3.5333

Probability lineage

• ATL@CHC: raw 0.6244 → calibrated 0.5695 → published 0.53; score coherence followed by matchup-recency guardrail
• BAL@NYM: raw 0.6169 → calibrated 0.595 → published 0.53; score coherence followed by matchup-recency guardrail
• BOS@TEX: raw 0.5432 → calibrated 0.5328 → published 0.5279; published probability matches the calibrated/governed value
• CHW@CLE: raw 0.6098 → calibrated 0.5787 → published 0.5565; score coherence followed by matchup-recency guardrail
• DET@TOR: raw 0.6469 → calibrated 0.6531 → published 0.58; score coherence followed by matchup-recency guardrail
• KC@HOU: raw 0.5 → calibrated 0.5076 → published 0.5055; published probability matches the calibrated/governed value
• LAD@CIN: raw 0.7529 → calibrated 0.7429 → published 0.58; score coherence followed by matchup-recency guardrail
• MIA@ARI: raw 0.5818 → calibrated 0.5642 → published 0.5545; score coherence followed by matchup-recency guardrail
• MIL@PIT: raw 0.6029 → calibrated 0.6672 → published 0.56; score coherence followed by matchup-recency guardrail
• NYY@MIN: raw 0.6828 → calibrated 0.6906 → published 0.58; score coherence followed by matchup-recency guardrail
• OAK@TB: raw 0.6849 → calibrated 0.7669 → published 0.7669; published probability matches the calibrated/governed value
• PHI@WAS: raw 0.5885 → calibrated 0.5506 → published 0.5098; score coherence followed by matchup-recency guardrail
• SD@COL: raw 0.6166 → calibrated 0.6367 → published 0.53; score coherence followed by matchup-recency guardrail
• SEA@LAA: raw 0.5039 → calibrated 0.5076 → published 0.5052; published probability matches the calibrated/governed value
• SF@STL: raw 0.5 → calibrated 0.5274 → published 0.5136; no recorded transformation explains the publication delta

Every game, in detail

ATL@CHC — ✅ hit

• Predicted CHC 5-7 (conf +0.06, lean) · actual CHC 4-8
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.2209 · team-run err 2.0 · total bias +0
• Contributing factors:
• Shota Imanaga (CHC) allowed 4 fewer ER than projected (1 vs 5)

BAL@NYM — ❌ miss

• Predicted NYM 5-7 (conf +0.06, lean) · actual BAL 7-1
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.2809 · team-run err 8.0 · total bias +4
• Contributing factors:
• total runs over-projected by 4 (12 vs 8)
• NYM offense over-projected by 6
• Chris Bassitt (BAL) allowed 3 fewer ER than projected (1 vs 4)
• Xzavion Curry (NYM) allowed 5 more ER than projected (6 vs 1)

BOS@TEX — ❌ miss

• Predicted BOS 6-5 (conf -0.06, lean) · actual TEX 3-7
• Root cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.278649 · team-run err 5.0 · total bias +1
• Contributing factors:
• BOS offense over-projected by 3

CHW@CLE — ✅ hit

• Predicted CLE 4-5 (conf +0.11, moderate) · actual CLE 3-6
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.196681 · team-run err 2.0 · total bias +0
• Contributing factors:
• Anthony Kay (CHW) allowed 5 fewer ER than projected (0 vs 5)

DET@TOR — ❌ miss

• Predicted DET 7-4 (conf -0.16, moderate) · actual TOR 1-5
• Root cause: Run environment far lower than projected (over-dispersed)
• Secondary cause: Backed the wrong side with real conviction
• Errors: winner WRONG · Brier 0.3364 · team-run err 7.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (11 vs 6)
• DET offense over-projected by 6
• committed to DET at 0.16 confidence and it lost

KC@HOU — ✅ hit

• Predicted KC 5-4 (conf -0.01, toss_up) · actual KC 5-2
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.244518 · team-run err 2.0 · total bias +2

LAD@CIN — ❌ miss

• Predicted LAD 8-4 (conf -0.16, moderate) · actual CIN 2-6
• Root cause: Backed the wrong side with real conviction
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.3364 · team-run err 8.0 · total bias +4
• Contributing factors:
• total runs over-projected by 4 (12 vs 8)
• LAD offense over-projected by 6
• committed to LAD at 0.16 confidence and it lost

MIA@ARI — ✅ hit

• Predicted MIA 7-5 (conf -0.11, moderate) · actual MIA 4-3
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.198427 · team-run err 5.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (12 vs 7)
• MIA offense over-projected by 3

MIL@PIT — ✅ hit

• Predicted MIL 5-3 (conf -0.12, moderate) · actual MIL 5-4
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.1936 · team-run err 1.0 · total bias -1

NYY@MIN — ❌ miss

• Predicted NYY 7-4 (conf -0.16, moderate) · actual MIN 4-5
• Root cause: Backed the wrong side with real conviction
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.3364 · team-run err 4.0 · total bias +2
• Contributing factors:
• NYY offense over-projected by 3
• Zebby Matthews (MIN) allowed 3 fewer ER than projected (0 vs 3)
• committed to NYY at 0.16 confidence and it lost

OAK@TB — ✅ hit

• Predicted TB 2-6 (conf +0.53, strong) · actual TB 3-4
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.054354 · team-run err 3.0 · total bias +1
• Contributing factors:
• Brady Basso (OAK) allowed 3 more ER than projected (4 vs 1)

PHI@WAS — ✅ hit

• Predicted PHI 5-4 (conf -0.02, toss_up) · actual PHI 3-0
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.240264 · team-run err 6.0 · total bias +6
• Contributing factors:
• total runs over-projected by 6 (9 vs 3)
• WAS offense over-projected by 4
• Zack Wheeler (PHI) allowed 3 fewer ER than projected (0 vs 3)

SD@COL — ✅ hit

• Predicted SD 7-5 (conf -0.06, lean) · actual SD 9-3
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.2209 · team-run err 4.0 · total bias +0

SEA@LAA — ✅ hit

• Predicted SEA 5-4 (conf -0.01, toss_up) · actual SEA 7-2
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.244811 · team-run err 4.0 · total bias +0

SF@STL — ✅ hit

• Predicted SF 5-4 (conf -0.03, toss_up) · actual SF 6-5
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.236604 · team-run err 2.0 · total bias -2
• Contributing factors:
• Matthew Liberatore (STL) allowed 3 fewer ER than projected (1 vs 4)

Cross-game lever tension

Each candidate fix, with the games it would help and the games it would hurt. A lever that repairs one game and breaks two is visible here before it is proposed.

score.variance_multiplier / score.variance_floor (widen run distributions)

• Targets: brier_score, calibration_error
• Net: 3 helped vs 1 hurt
• Would help:
• DET@TOR (miss): lost a 0.16-confidence pick; wider tails lower overconfidence
• LAD@CIN (miss): lost a 0.16-confidence pick; wider tails lower overconfidence
• NYY@MIN (miss): lost a 0.16-confidence pick; wider tails lower overconfidence
• Would hurt / risk:
• OAK@TB (hit): strong correct call at 0.53; widening softens a good edge

player.pitcher.v4 form weights / player.pitcher_prior_games

• Targets: starter_er_mae, starter_outs_mae
• Net: 7 helped vs 6 hurt
• Would help:
• ATL@CHC (hit): starter ER off by 4
• BAL@NYM (miss): starter ER off by 5
• CHW@CLE (hit): starter ER off by 5
• NYY@MIN (miss): starter ER off by 3
• OAK@TB (hit): starter ER off by 3
• PHI@WAS (hit): starter ER off by 3
• SF@STL (hit): starter ER off by 3
• Would hurt / risk:
• BOS@TEX (miss): starter projection was accurate (ER off 0); reweight risks disturbing it
• KC@HOU (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• LAD@CIN (miss): starter projection was accurate (ER off 0); reweight risks disturbing it
• MIA@ARI (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• SD@COL (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• SEA@LAA (hit): starter projection was accurate (ER off 1); reweight risks disturbing it

recent. / season. run-mean weighting

• Targets: team_run_mae, total_run_mae, accuracy
• Net: 9 helped vs 5 hurt
• Would help:
• BAL@NYM (miss): team runs off by 8
• BOS@TEX (miss): team runs off by 5
• DET@TOR (miss): team runs off by 7
• LAD@CIN (miss): team runs off by 8
• MIA@ARI (hit): team runs off by 5
• NYY@MIN (miss): team runs off by 4
• PHI@WAS (hit): team runs off by 6
• SD@COL (hit): team runs off by 4
• SEA@LAA (hit): team runs off by 4
• Would hurt / risk:
• ATL@CHC (hit): team runs already close (off 2); reweight risks disturbing it
• CHW@CLE (hit): team runs already close (off 2); reweight risks disturbing it
• KC@HOU (hit): team runs already close (off 2); reweight risks disturbing it
• MIL@PIT (hit): team runs already close (off 1); reweight risks disturbing it
• SF@STL (hit): team runs already close (off 2); reweight risks disturbing it