Deep slate diagnosis — 20260907

• Games: 11 · hits 6 · misses 5 · winner accuracy 0.5455
• Mean total-run bias on losses: -1.0 (negative = under-projected)
• Mean team-run error on losses: 3.8 · mean Brier on losses: 0.354

Loss primary drivers

• Small edge on the losing side: 1
• Several errors of similar size: 1
• Backed the wrong side with real conviction: 3

High-level causal issue register

[CRITICAL] Published probability differs materially without a recorded transformation

• Domain: probability_lineage · affected: ARI@KC, CIN@LAD, LAA@BOS, NYM@MIA
• Evidence: 4 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] Published confidence buckets were poorly aligned on this slate

• Domain: calibration · affected: ARI@KC, ATL@PHI, CHC@MIL, CIN@LAD, CLE@BAL, LAA@BOS, MIN@DET, NYM@MIA, STL@SF, TOR@OAK, WAS@SD
• Evidence: weighted absolute calibration error 0.2611 across 11 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: ARI@KC, ATL@PHI, CHC@MIL, CIN@LAD, CLE@BAL, LAA@BOS, MIN@DET, NYM@MIA, STL@SF, TOR@OAK, WAS@SD
• Evidence: home/away/total R² = 0.0124/0.0205/0.0184
• 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.

[INFO] Every settled game received structural QA

• Domain: coverage · affected: ARI@KC, ATL@PHI, CHC@MIL, CIN@LAD, CLE@BAL, LAA@BOS, MIN@DET, NYM@MIA, STL@SF, TOR@OAK, WAS@SD
• Evidence: 11/11 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@KC, ATL@PHI, CHC@MIL, CIN@LAD, CLE@BAL, LAA@BOS, MIN@DET, NYM@MIA, STL@SF, TOR@OAK, WAS@SD
• 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, 20260906, 20260907 (10 prior slates loaded)
• Highest-confidence pick loss streak: 2 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.
• 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: flat · series [0.4667, 0.9412, 0.5, 0.6667, 0.4667, 0.3333, 0.5556, 0.625, 0.5333, 0.6, 0.5455]
• weighted_calibration_error: regressing · series [0.2179, 0.3961, 0.0693, 0.1661, 0.2476, 0.21, 0.1474, 0.1316, 0.1, 0.0937, 0.2611]
• margin_bias: improving · series [-1.333, -1.882, -2.786, -3.5, -2.533, -3.067, -1.778, -0.688, -1.467, -1.467, -0.182]
• 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, 5.333, 3.8]

Pipeline stage quality

• historical_ml: winner 6/11 (0.5455), team-run MAE 1.5782, margin MAE 2.195
• variance_applied: winner 5/11 (0.4545), team-run MAE 1.5909, margin MAE 2.0909
• enhanced_prediction: winner 6/11 (0.5455), team-run MAE 1.8182, margin MAE 2.0
• embedded_savant_run_environment: winner 6/11 (0.5455), team-run MAE 1.9091, margin MAE 2.1818
• fundamental_edge_target_review: winner 6/11 (0.5455), team-run MAE 1.9091, margin MAE 2.1818
• capped_target: winner 6/11 (0.5455), team-run MAE 1.9091, margin MAE 2.1818
• simulation_result: winner 6/11 (0.5455), team-run MAE 1.9091, margin MAE 2.1818
• published_final: winner 6/11 (0.5455), team-run MAE 1.9091, margin MAE 2.1818

Probability lineage

• ARI@KC: raw 0.5978 → calibrated 0.618 → published 0.5796; no recorded transformation explains the publication delta
• ATL@PHI: raw 0.5995 → calibrated 0.5565 → published 0.53; score coherence followed by matchup-recency guardrail
• CHC@MIL: raw 0.504 → calibrated 0.508 → published 0.5068; published probability matches the calibrated/governed value
• CIN@LAD: raw 0.5322 → calibrated 0.5751 → published 0.5326; no recorded transformation explains the publication delta
• CLE@BAL: raw 0.5133 → calibrated 0.5107 → published 0.5078; published probability matches the calibrated/governed value
• LAA@BOS: raw 0.6371 → calibrated 0.6539 → published 0.6231; no recorded transformation explains the publication delta
• MIN@DET: raw 0.5258 → calibrated 0.5675 → published 0.53; score coherence followed by matchup-recency guardrail
• NYM@MIA: raw 0.582 → calibrated 0.5477 → published 0.5208; no recorded transformation explains the publication delta
• STL@SF: raw 0.5409 → calibrated 0.6268 → published 0.6268; published probability matches the calibrated/governed value
• TOR@OAK: raw 0.7068 → calibrated 0.6996 → published 0.6996; published probability matches the calibrated/governed value
• WAS@SD: raw 0.6299 → calibrated 0.5724 → published 0.5297; score coherence followed by matchup-recency guardrail

Every game, in detail

ARI@KC — ❌ miss

• Predicted KC 3-5 (conf +0.16, moderate) · actual ARI 5-4
• Root cause: Backed the wrong side with real conviction
• Errors: winner WRONG · Brier 0.335963 · team-run err 3.0 · total bias -1
• Contributing factors:
• committed to KC at 0.16 confidence and it lost

ATL@PHI — ✅ hit

• Predicted PHI 3-4 (conf +0.06, lean) · actual PHI 0-1
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.2209 · team-run err 6.0 · total bias +6
• Contributing factors:
• total runs over-projected by 6 (7 vs 1)
• ATL offense over-projected by 3
• PHI offense over-projected by 3

CHC@MIL — ✅ hit

• Predicted MIL 5-7 (conf +0.01, toss_up) · actual MIL 3-4
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.243228 · team-run err 5.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (12 vs 7)
• MIL offense over-projected by 3
• Robert Gasser (MIL) allowed 3 fewer ER than projected (1 vs 4)

CIN@LAD — ✅ hit

• Predicted LAD 5-6 (conf +0.07, lean) · actual LAD 3-6
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.218499 · team-run err 2.0 · total bias +2
• Contributing factors:
• Chase Burns (CIN) allowed 3 fewer ER than projected (0 vs 3)

CLE@BAL — ✅ hit

• Predicted BAL 3-4 (conf +0.02, toss_up) · actual BAL 4-6
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.242295 · team-run err 3.0 · total bias -3
• Contributing factors:
• total runs under-projected by 3 (7 vs 10)

LAA@BOS — ✅ hit

• Predicted BOS 3-5 (conf +0.25, strong) · actual BOS 2-5
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.142016 · team-run err 1.0 · total bias +1

MIN@DET — ❌ miss

• Predicted MIN 5-4 (conf -0.06, lean) · actual DET 4-5
• Root cause: Several errors of similar size
• Errors: winner WRONG · Brier 0.2809 · team-run err 2.0 · total bias +0

NYM@MIA — ❌ miss

• Predicted MIA 4-5 (conf +0.04, lean) · actual NYM 9-4
• Root cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.271202 · team-run err 6.0 · total bias -4
• Contributing factors:
• total runs under-projected by 4 (9 vs 13)
• NYM offense under-projected by 5

STL@SF — ❌ miss

• Predicted STL 7-5 (conf -0.25, strong) · actual SF 4-5
• Root cause: Backed the wrong side with real conviction
• Errors: winner WRONG · Brier 0.392833 · team-run err 3.0 · total bias +3
• Contributing factors:
• total runs over-projected by 3 (12 vs 9)
• STL offense over-projected by 3
• Logan Webb (SF) allowed 3 fewer ER than projected (1 vs 4)
• committed to STL at 0.25 confidence and it lost

TOR@OAK — ❌ miss

• Predicted TOR 6-2 (conf -0.40, strong) · actual OAK 5-6
• Root cause: Backed the wrong side with real conviction
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.489381 · team-run err 5.0 · total bias -3
• Contributing factors:
• total runs under-projected by 3 (8 vs 11)
• OAK offense under-projected by 4
• committed to TOR at 0.40 confidence and it lost

WAS@SD — ✅ hit

• Predicted SD 5-6 (conf +0.06, lean) · actual SD 2-3
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.221225 · team-run err 6.0 · total bias +6
• Contributing factors:
• total runs over-projected by 6 (11 vs 5)
• WAS offense over-projected by 3
• SD offense over-projected by 3
• Nick Pivetta (SD) allowed 4 fewer ER than projected (0 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:
• ARI@KC (miss): lost a 0.16-confidence pick; wider tails lower overconfidence
• STL@SF (miss): lost a 0.25-confidence pick; wider tails lower overconfidence
• TOR@OAK (miss): lost a 0.40-confidence pick; wider tails lower overconfidence
• Would hurt / risk:
• LAA@BOS (hit): strong correct call at 0.25; widening softens a good edge

player.pitcher.v4 form weights / player.pitcher_prior_games

• Targets: starter_er_mae, starter_outs_mae
• Net: 4 helped vs 0 hurt
• Would help:
• CHC@MIL (hit): starter ER off by 3
• CIN@LAD (hit): starter ER off by 3
• STL@SF (miss): starter ER off by 3
• WAS@SD (hit): starter ER off by 4

recent. / season. run-mean weighting

• Targets: team_run_mae, total_run_mae, accuracy
• Net: 5 helped vs 3 hurt
• Would help:
• ATL@PHI (hit): team runs off by 6
• CHC@MIL (hit): team runs off by 5
• NYM@MIA (miss): team runs off by 6
• TOR@OAK (miss): team runs off by 5
• WAS@SD (hit): team runs off by 6
• Would hurt / risk:
• CIN@LAD (hit): team runs already close (off 2); reweight risks disturbing it
• LAA@BOS (hit): team runs already close (off 1); reweight risks disturbing it
• MIN@DET (miss): team runs already close (off 2); reweight risks disturbing it