Deep slate diagnosis — 20260905

• Games: 15 · hits 8 · misses 7 · winner accuracy 0.5333
• Mean total-run bias on losses: 1.286 (negative = under-projected)
• Mean team-run error on losses: 5.286 · mean Brier on losses: 0.301

Loss primary drivers

• Several errors of similar size: 1
• Run environment far lower than projected (over-dispersed): 1
• Starting-pitcher projection was wrong: 3
• Genuine toss-up that flipped: 1
• Backed the wrong side with real conviction: 1

High-level causal issue register

[CRITICAL] Published probability differs materially without a recorded transformation

• Domain: probability_lineage · affected: ATL@PHI, LAA@PIT, WAS@LAD
• Evidence: 3 game(s) have an unexplained ≥2pp calibrated-to-published delta
• Action: Block any promotion until the missing probability transformation is recorded and reconciled.
• Do not infer: Do not assume the probability governor was bypassed until you have traced the pick-side math.

[HIGH] An intermediate stage materially degraded winner direction

• Domain: pipeline_stages · affected: NYY@SD, SF@NYM, TB@TEX
• 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] Validated run models explain very little holdout variance

• Domain: model_validation · affected: ARI@HOU, ATL@PHI, BOS@BAL, CHC@MIA, DET@CLE, LAA@PIT, MIL@CIN, MIN@CHW, NYY@SD, OAK@SEA, SF@NYM, STL@COL, TB@TEX, TOR@KC, WAS@LAD
• Evidence: home/away/total R² = 0.0106/0.0203/0.0176
• 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: CHC@MIA, LAA@PIT, MIL@CIN
• 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: ARI@HOU, ATL@PHI, BOS@BAL, CHC@MIA, DET@CLE, LAA@PIT, MIL@CIN, MIN@CHW, NYY@SD, OAK@SEA, SF@NYM, STL@COL, TB@TEX, TOR@KC, WAS@LAD
• 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] Simulation produced no winner-direction intervention on this slate

• Domain: simulation · affected: ARI@HOU, ATL@PHI, BOS@BAL, CHC@MIA, DET@CLE, LAA@PIT, MIL@CIN, MIN@CHW, NYY@SD, OAK@SEA, SF@NYM, STL@COL, TB@TEX, TOR@KC, WAS@LAD
• 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: 20260828, 20260829, 20260830, 20260831, 20260901, 20260902, 20260903, 20260904, 20260905 (8 prior slates loaded)
• Highest-confidence pick loss streak: 0 slate(s)
• Actionable pattern detected: True

Escalated persistent patterns

• simulation_inert_on_slate — 9 consecutive slates (9 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 9 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.
• weak_run_model_signal — 9 consecutive slates (9 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 9 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: regressing · series [0.4667, 0.9412, 0.5, 0.6667, 0.4667, 0.3333, 0.5556, 0.625, 0.5333]
• weighted_calibration_error: improving · series [0.2179, 0.3961, 0.0693, 0.1661, 0.2476, 0.21, 0.1474, 0.1316, 0.1]
• margin_bias: improving · series [-1.333, -1.882, -2.786, -3.5, -2.533, -3.067, -1.778, -0.688, -1.467]
• mean_team_run_error_on_losses: improving · series [5.0, 3.0, 7.0, 8.25, 7.125, 6.4, 6.75, 6.0, 5.286]

Pipeline stage quality

• historical_ml: winner 8/15 (0.5333), team-run MAE 1.8729, margin MAE 3.0583
• variance_applied: winner 5/15 (0.3333), team-run MAE 1.8333, margin MAE 3.0
• enhanced_prediction: winner 8/15 (0.5333), team-run MAE 1.9667, margin MAE 3.2667
• embedded_savant_run_environment: winner 8/15 (0.5333), team-run MAE 2.0, margin MAE 3.3333
• fundamental_edge_target_review: winner 8/15 (0.5333), team-run MAE 2.0, margin MAE 3.3333
• capped_target: winner 8/15 (0.5333), team-run MAE 2.0, margin MAE 3.3333
• simulation_result: winner 8/15 (0.5333), team-run MAE 2.0, margin MAE 3.3333
• published_final: winner 8/15 (0.5333), team-run MAE 2.0, margin MAE 3.3333

Probability lineage

• ARI@HOU: raw 0.5085 → calibrated 0.5083 → published 0.5171; score coherence followed by matchup-recency guardrail
• ATL@PHI: raw 0.5028 → calibrated 0.505 → published 0.538; no recorded transformation explains the publication delta
• BOS@BAL: raw 0.5052 → calibrated 0.5051 → published 0.5323; score coherence followed by matchup-recency guardrail
• CHC@MIA: raw 0.7723 → calibrated 0.7404 → published 0.7395; score_coherence aligned probability to the published score margin
• DET@CLE: raw 0.568 → calibrated 0.5663 → published 0.5544; score coherence followed by matchup-recency guardrail
• LAA@PIT: raw 0.67 → calibrated 0.6591 → published 0.6357; no recorded transformation explains the publication delta
• MIL@CIN: raw 0.7085 → calibrated 0.6861 → published 0.54; score coherence followed by matchup-recency guardrail
• MIN@CHW: raw 0.5085 → calibrated 0.5083 → published 0.53; score coherence followed by matchup-recency guardrail
• NYY@SD: raw 0.551 → calibrated 0.5499 → published 0.5485; published probability matches the calibrated/governed value
• OAK@SEA: raw 0.5085 → calibrated 0.5083 → published 0.53; score coherence followed by matchup-recency guardrail
• SF@NYM: raw 0.5016 → calibrated 0.505 → published 0.5236; no recorded transformation explains the publication delta
• STL@COL: raw 0.5085 → calibrated 0.5083 → published 0.5092; published probability matches the calibrated/governed value
• TB@TEX: raw 0.5085 → calibrated 0.5083 → published 0.5335; score coherence followed by matchup-recency guardrail
• TOR@KC: raw 0.6256 → calibrated 0.6202 → published 0.57; score coherence followed by matchup-recency guardrail
• WAS@LAD: raw 0.6275 → calibrated 0.6219 → published 0.5734; no recorded transformation explains the publication delta

Every game, in detail

ARI@HOU — ✅ hit

• Predicted ARI 5-4 (conf -0.03, toss_up) · actual ARI 4-3
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.233146 · team-run err 2.0 · total bias +2
• Contributing factors:
• Ethan Pecko (HOU) allowed 3 fewer ER than projected (1 vs 4)

ATL@PHI — ❌ miss

• Predicted ATL 5-4 (conf -0.08, lean) · actual PHI 2-4
• Root cause: Several errors of similar size
• Errors: winner WRONG · Brier 0.289421 · team-run err 3.0 · total bias +3
• Contributing factors:
• total runs over-projected by 3 (9 vs 6)
• ATL offense over-projected by 3

BOS@BAL — ✅ hit

• Predicted BOS 5-4 (conf -0.06, lean) · actual BOS 5-0
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.218738 · team-run err 4.0 · total bias +4
• Contributing factors:
• total runs over-projected by 4 (9 vs 5)
• BAL offense over-projected by 4
• Sonny Gray (BOS) allowed 3 fewer ER than projected (0 vs 3)

CHC@MIA — ✅ hit

• Predicted CHC 6-3 (conf -0.48, strong) · actual CHC 6-5
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.067845 · team-run err 2.0 · total bias -2

DET@CLE — ❌ miss

• Predicted CLE 3-4 (conf +0.11, moderate) · actual DET 6-0
• Root cause: Backed the wrong side with real conviction
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.307378 · team-run err 7.0 · total bias +1
• Contributing factors:
• DET offense under-projected by 3
• CLE offense over-projected by 4
• committed to CLE at 0.11 confidence and it lost

LAA@PIT — ❌ miss

• Predicted PIT 4-6 (conf +0.27, strong) · actual LAA 6-1
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Backed the wrong side with real conviction
• Errors: winner WRONG · Brier 0.404151 · team-run err 7.0 · total bias +3
• Contributing factors:
• total runs over-projected by 3 (10 vs 7)
• PIT offense over-projected by 5
• Yusei Kikuchi (LAA) allowed 6 fewer ER than projected (0 vs 6)
• committed to PIT at 0.27 confidence and it lost

MIL@CIN — ❌ miss

• Predicted MIL 8-5 (conf -0.08, lean) · actual CIN 3-5
• Root cause: Run environment far lower than projected (over-dispersed)
• Errors: winner WRONG · Brier 0.2916 · team-run err 5.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (13 vs 8)
• MIL offense over-projected by 5

MIN@CHW — ✅ hit

• Predicted MIN 6-5 (conf -0.06, lean) · actual MIN 6-4
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.2209 · team-run err 1.0 · total bias +1

NYY@SD — ✅ hit

• Predicted NYY 4-3 (conf -0.10, lean) · actual NYY 5-1
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.203847 · team-run err 3.0 · total bias +1

OAK@SEA — ❌ miss

• Predicted SEA 6-8 (conf +0.06, lean) · actual OAK 6-2
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Run environment far lower than projected (over-dispersed)
• Errors: winner WRONG · Brier 0.2809 · team-run err 6.0 · total bias +6
• Contributing factors:
• total runs over-projected by 6 (14 vs 8)
• SEA offense over-projected by 6
• Jeffrey Springs (OAK) allowed 5 fewer ER than projected (0 vs 5)
• George Kirby (SEA) allowed 3 fewer ER than projected (3 vs 6)

SF@NYM — ❌ miss

• Predicted NYM 4-5 (conf +0.05, lean) · actual SF 9-5
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Run environment far higher than projected (under-dispersed)
• Errors: winner WRONG · Brier 0.274134 · team-run err 5.0 · total bias -5
• Contributing factors:
• total runs under-projected by 5 (9 vs 14)
• SF offense under-projected by 5
• Zac Thornton (NYM) allowed 5 more ER than projected (6 vs 1)

STL@COL — ❌ miss

• Predicted STL 7-6 (conf -0.02, toss_up) · actual COL 7-10
• Root cause: Genuine toss-up that flipped
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.259286 · team-run err 4.0 · total bias -4
• Contributing factors:
• total runs under-projected by 4 (13 vs 17)
• COL offense under-projected by 4

TB@TEX — ✅ hit

• Predicted TB 6-5 (conf -0.07, lean) · actual TB 6-3
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.217661 · team-run err 2.0 · total bias +2

TOR@KC — ✅ hit

• Predicted TOR 7-5 (conf -0.14, moderate) · actual TOR 4-3
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.1849 · team-run err 5.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (12 vs 7)
• TOR offense over-projected by 3

WAS@LAD — ✅ hit

• Predicted LAD 3-4 (conf +0.15, moderate) · actual LAD 5-6
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.181961 · team-run err 4.0 · total bias -4
• Contributing factors:
• total runs under-projected by 4 (7 vs 11)

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: 2 helped vs 1 hurt
• Would help:
• DET@CLE (miss): lost a 0.11-confidence pick; wider tails lower overconfidence
• LAA@PIT (miss): lost a 0.27-confidence pick; wider tails lower overconfidence
• Would hurt / risk:
• CHC@MIA (hit): strong correct call at 0.48; widening softens a good edge

player.pitcher.v4 form weights / player.pitcher_prior_games

• Targets: starter_er_mae, starter_outs_mae
• Net: 5 helped vs 5 hurt
• Would help:
• ARI@HOU (hit): starter ER off by 3
• BOS@BAL (hit): starter ER off by 3
• LAA@PIT (miss): starter ER off by 6
• OAK@SEA (miss): starter ER off by 5
• SF@NYM (miss): starter ER off by 5
• Would hurt / risk:
• CHC@MIA (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• MIL@CIN (miss): starter projection was accurate (ER off 1); reweight risks disturbing it
• MIN@CHW (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• STL@COL (miss): starter projection was accurate (ER off 1); reweight risks disturbing it
• TOR@KC (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 4 hurt
• Would help:
• BOS@BAL (hit): team runs off by 4
• DET@CLE (miss): team runs off by 7
• LAA@PIT (miss): team runs off by 7
• MIL@CIN (miss): team runs off by 5
• OAK@SEA (miss): team runs off by 6
• SF@NYM (miss): team runs off by 5
• STL@COL (miss): team runs off by 4
• TOR@KC (hit): team runs off by 5
• WAS@LAD (hit): team runs off by 4
• Would hurt / risk:
• ARI@HOU (hit): team runs already close (off 2); reweight risks disturbing it
• CHC@MIA (hit): team runs already close (off 2); reweight risks disturbing it
• MIN@CHW (hit): team runs already close (off 1); reweight risks disturbing it
• TB@TEX (hit): team runs already close (off 2); reweight risks disturbing it