Deep slate diagnosis — 20260918

• Games: 15 · hits 10 · misses 5 · winner accuracy 0.6667
• Mean total-run bias on losses: 1.2 (negative = under-projected)
• Mean team-run error on losses: 5.2 · mean Brier on losses: 0.283

Loss primary drivers

• Small edge on the losing side: 1
• Starting-pitcher projection was wrong: 2
• 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: KC@PIT, MIL@BAL
• Evidence: 2 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: ATL@HOU, DET@CHW, WAS@STL
• 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: ATL@HOU, BOS@TB, CHC@CIN, DET@CHW, KC@PIT, MIA@SD, MIL@BAL, MIN@LAA, NYY@ARI, OAK@CLE, PHI@NYM, SEA@COL, SF@LAD, TOR@TEX, WAS@STL
• Evidence: home/away/total R² = 0.0157/0.0327/0.0073
• 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: ATL@HOU, BOS@TB, CHC@CIN, DET@CHW, KC@PIT, MIA@SD, MIL@BAL, MIN@LAA, NYY@ARI, OAK@CLE, PHI@NYM, SEA@COL, SF@LAD, TOR@TEX, WAS@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] Simulation produced no winner-direction intervention on this slate

• Domain: simulation · affected: ATL@HOU, BOS@TB, CHC@CIN, DET@CHW, KC@PIT, MIA@SD, MIL@BAL, MIN@LAA, NYY@ARI, OAK@CLE, PHI@NYM, SEA@COL, SF@LAD, TOR@TEX, WAS@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: 20260908, 20260909, 20260910, 20260911, 20260912, 20260913, 20260914, 20260915, 20260916, 20260917, 20260918 (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.
• 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 [0.7333, 0.6667, 0.6, 0.6667, 0.5333, 0.6, 0.7, 0.4, 0.6667, 0.4444, 0.6667]
• weighted_calibration_error: improving · series [0.1825, 0.1069, 0.1171, 0.2091, 0.1711, 0.0827, 0.2812, 0.2795, 0.1661, 0.1016, 0.1247]
• margin_bias: regressing · series [-0.867, -2.133, -1.0, -2.933, -2.667, -1.8, -0.4, -1.667, -1.133, -2.111, -2.0]
• mean_team_run_error_on_losses: improving · series [6.75, 5.6, 6.5, 5.0, 6.143, 5.0, 6.333, 5.556, 6.4, 3.6, 5.2]

Pipeline stage quality

• historical_ml: winner 11/15 (0.7333), team-run MAE 2.2167, margin MAE 3.5339
• variance_applied: winner 8/15 (0.5333), team-run MAE 2.2333, margin MAE 3.4
• enhanced_prediction: winner 10/15 (0.6667), team-run MAE 2.4, margin MAE 3.7333
• embedded_savant_run_environment: winner 9/15 (0.6), team-run MAE 2.4, margin MAE 3.7333
• fundamental_edge_target_review: winner 10/15 (0.6667), team-run MAE 2.4, margin MAE 3.7333
• capped_target: winner 10/15 (0.6667), team-run MAE 2.4, margin MAE 3.7333
• simulation_result: winner 10/15 (0.6667), team-run MAE 2.4, margin MAE 3.7333
• published_final: winner 10/15 (0.6667), team-run MAE 2.4, margin MAE 3.7333

Probability lineage

• ATL@HOU: raw 0.541 → calibrated 0.5725 → published 0.5616; no recorded transformation explains the publication delta
• BOS@TB: raw 0.618 → calibrated 0.6244 → published 0.5723; score coherence followed by matchup-recency guardrail
• CHC@CIN: raw 0.5404 → calibrated 0.6582 → published 0.53; score coherence followed by matchup-recency guardrail
• DET@CHW: raw 0.5763 → calibrated 0.6129 → published 0.5435; score coherence followed by matchup-recency guardrail
• KC@PIT: raw 0.6541 → calibrated 0.627 → published 0.5862; no recorded transformation explains the publication delta
• MIA@SD: raw 0.6803 → calibrated 0.6605 → published 0.58; score coherence followed by matchup-recency guardrail
• MIL@BAL: raw 0.6786 → calibrated 0.7376 → published 0.6657; no recorded transformation explains the publication delta
• MIN@LAA: raw 0.508 → calibrated 0.5594 → published 0.5343; score coherence followed by matchup-recency guardrail
• NYY@ARI: raw 0.5 → calibrated 0.505 → published 0.505; published probability matches the calibrated/governed value
• OAK@CLE: raw 0.7538 → calibrated 0.698 → published 0.58; score coherence followed by matchup-recency guardrail
• PHI@NYM: raw 0.5233 → calibrated 0.516 → published 0.5115; published probability matches the calibrated/governed value
• SEA@COL: raw 0.776 → calibrated 0.7605 → published 0.7605; published probability matches the calibrated/governed value
• SF@LAD: raw 0.626 → calibrated 0.6 → published 0.5383; score coherence followed by matchup-recency guardrail
• TOR@TEX: raw 0.5 → calibrated 0.505 → published 0.505; published probability matches the calibrated/governed value
• WAS@STL: raw 0.5093 → calibrated 0.511 → published 0.5094; published probability matches the calibrated/governed value

Every game, in detail

ATL@HOU — ✅ hit

• Predicted ATL 4-3 (conf -0.12, moderate) · actual ATL 6-2
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.192156 · team-run err 3.0 · total bias -1

BOS@TB — ❌ miss

• Predicted TB 4-5 (conf +0.14, moderate) · actual BOS 4-2
• Root cause: Backed the wrong side with real conviction
• Errors: winner WRONG · Brier 0.327531 · team-run err 3.0 · total bias +3
• Contributing factors:
• total runs over-projected by 3 (9 vs 6)
• TB offense over-projected by 3
• Ian Seymour (TB) allowed 3 more ER than projected (4 vs 1)
• committed to TB at 0.14 confidence and it lost

CHC@CIN — ❌ miss

• Predicted CHC 6-4 (conf -0.06, lean) · actual CIN 4-6
• Root cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.2809 · team-run err 4.0 · total bias +0
• Contributing factors:
• Clay Holmes (CHC) allowed 3 more ER than projected (6 vs 3)

DET@CHW — ✅ hit

• Predicted DET 6-4 (conf -0.09, lean) · actual DET 11-8
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.208402 · team-run err 9.0 · total bias -9
• Contributing factors:
• total runs under-projected by 9 (10 vs 19)
• DET offense under-projected by 5
• CHW offense under-projected by 4

KC@PIT — ✅ hit

• Predicted PIT 3-4 (conf +0.17, moderate) · actual PIT 5-8
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.171227 · team-run err 6.0 · total bias -6
• Contributing factors:
• total runs under-projected by 6 (7 vs 13)
• PIT offense under-projected by 4
• Paul Skenes (PIT) allowed 3 more ER than projected (5 vs 2)

MIA@SD — ✅ hit

• Predicted SD 4-6 (conf +0.16, moderate) · actual SD 2-8
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.1764 · team-run err 4.0 · total bias +0

MIL@BAL — ✅ hit

• Predicted MIL 7-4 (conf -0.33, strong) · actual MIL 6-5
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.111777 · team-run err 2.0 · total bias +0

MIN@LAA — ❌ miss

• Predicted LAA 4-5 (conf +0.07, lean) · actual MIN 3-0
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Run environment far lower than projected (over-dispersed)
• Errors: winner WRONG · Brier 0.285524 · team-run err 6.0 · total bias +6
• Contributing factors:
• total runs over-projected by 6 (9 vs 3)
• LAA offense over-projected by 5
• Connor Prielipp (MIN) allowed 4 fewer ER than projected (0 vs 4)

NYY@ARI — ✅ hit

• Predicted NYY 5-4 (conf -0.01, toss_up) · actual NYY 9-2
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.245025 · team-run err 6.0 · total bias -2
• Contributing factors:
• NYY offense under-projected by 4
• Eduardo Rodriguez (ARI) allowed 3 more ER than projected (7 vs 4)

OAK@CLE — ✅ hit

• Predicted CLE 3-7 (conf +0.16, moderate) · actual CLE 3-5
• Root cause: strong_correct_call
• Errors: winner correct · Brier 0.1764 · team-run err 2.0 · total bias +2

PHI@NYM — ❌ miss

• Predicted PHI 5-4 (conf -0.02, toss_up) · actual NYM 3-6
• Root cause: Genuine toss-up that flipped
• Secondary cause: Small edge on the losing side
• Errors: winner WRONG · Brier 0.26166 · team-run err 4.0 · total bias +0

SEA@COL — ✅ hit

• Predicted SEA 9-5 (conf -0.52, strong) · actual SEA 5-4
• Root cause: right_winner_wrong_score
• Errors: winner correct · Brier 0.057375 · team-run err 5.0 · total bias +5
• Contributing factors:
• total runs over-projected by 5 (14 vs 9)
• SEA offense over-projected by 4
• Gabriel Hughes (COL) allowed 3 fewer ER than projected (3 vs 6)

SF@LAD — ✅ hit

• Predicted LAD 2-4 (conf +0.08, lean) · actual LAD 2-8
• Root cause: solid_correct_call
• Errors: winner correct · Brier 0.213184 · team-run err 4.0 · total bias -4
• Contributing factors:
• total runs under-projected by 4 (6 vs 10)
• LAD offense under-projected by 4

TOR@TEX — ✅ hit

• Predicted TEX 4-5 (conf +0.01, toss_up) · actual TEX 1-7
• Root cause: toss_up_went_our_way
• Errors: winner correct · Brier 0.245025 · team-run err 5.0 · total bias +1
• Contributing factors:
• TOR offense over-projected by 3

WAS@STL — ❌ miss

• Predicted STL 3-4 (conf +0.02, toss_up) · actual WAS 9-1
• Root cause: Starting-pitcher projection was wrong
• Secondary cause: Genuine toss-up that flipped
• Errors: winner WRONG · Brier 0.259463 · team-run err 9.0 · total bias -3
• Contributing factors:
• total runs under-projected by 3 (7 vs 10)
• WAS offense under-projected by 6
• STL offense over-projected by 3
• Cade Cavalli (WAS) allowed 3 fewer ER than projected (0 vs 3)
• Kyle Leahy (STL) allowed 5 more ER than projected (6 vs 1)

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: 1 helped vs 2 hurt
• Would help:
• BOS@TB (miss): lost a 0.14-confidence pick; wider tails lower overconfidence
• Would hurt / risk:
• MIL@BAL (hit): strong correct call at 0.33; widening softens a good edge
• SEA@COL (hit): strong correct call at 0.52; 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 2 hurt
• Would help:
• BOS@TB (miss): starter ER off by 3
• CHC@CIN (miss): starter ER off by 3
• KC@PIT (hit): starter ER off by 3
• MIN@LAA (miss): starter ER off by 4
• NYY@ARI (hit): starter ER off by 3
• SEA@COL (hit): starter ER off by 3
• WAS@STL (miss): starter ER off by 5
• Would hurt / risk:
• MIL@BAL (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
• PHI@NYM (miss): 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: 11 helped vs 2 hurt
• Would help:
• CHC@CIN (miss): team runs off by 4
• DET@CHW (hit): team runs off by 9
• KC@PIT (hit): team runs off by 6
• MIA@SD (hit): team runs off by 4
• MIN@LAA (miss): team runs off by 6
• NYY@ARI (hit): team runs off by 6
• PHI@NYM (miss): team runs off by 4
• SEA@COL (hit): team runs off by 5
• SF@LAD (hit): team runs off by 4
• TOR@TEX (hit): team runs off by 5
• WAS@STL (miss): team runs off by 9
• Would hurt / risk:
• MIL@BAL (hit): team runs already close (off 2); reweight risks disturbing it
• OAK@CLE (hit): team runs already close (off 2); reweight risks disturbing it