Deep slate diagnosis — 20260910

Games: 5 · hits 3 · misses 2 · winner accuracy 0.6
Mean total-run bias on losses: 6.5 (negative = under-projected)
Mean team-run error on losses: 6.5 · mean Brier on losses: 0.289

Loss primary drivers

Run environment far lower than projected (over-dispersed): 1
Starting-pitcher projection was wrong: 1

High-level causal issue register

[CRITICAL] Published probability differs materially without a recorded transformation

Domain: probability_lineage · affected: PIT@CHW
Evidence: 1 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: TEX@SEA
Evidence: historical_ml → variance_applied changed winner accuracy by -0.2000 (-1 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: COL@NYY, HOU@PHI, PIT@CHW, TB@ATL, TEX@SEA
Evidence: home/away/total R² = 0.0118/0.0210/0.0207
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: COL@NYY, HOU@PHI, PIT@CHW, TB@ATL, TEX@SEA
Evidence: 5/5 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: COL@NYY, HOU@PHI, PIT@CHW, TB@ATL, TEX@SEA
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: 20260831, 20260901, 20260902, 20260903, 20260904, 20260905, 20260906, 20260907, 20260908, 20260909, 20260910 (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: flat · series [0.6667, 0.4667, 0.3333, 0.5556, 0.625, 0.5333, 0.6, 0.5455, 0.7333, 0.6667, 0.6]
weighted_calibration_error: improving · series [0.1661, 0.2476, 0.21, 0.1474, 0.1316, 0.1, 0.0937, 0.2611, 0.1825, 0.1069, 0.1171]
margin_bias: improving · series [-3.5, -2.533, -3.067, -1.778, -0.688, -1.467, -1.467, -0.182, -0.867, -2.133, -1.0]
mean_team_run_error_on_losses: regressing · series [8.25, 7.125, 6.4, 6.75, 6.0, 5.286, 5.333, 3.8, 6.75, 5.6, 6.5]

Pipeline stage quality

historical_ml: winner 3/5 (0.6), team-run MAE 2.4915, margin MAE 2.6222
variance_applied: winner 2/5 (0.4), team-run MAE 2.5, margin MAE 2.6
enhanced_prediction: winner 3/5 (0.6), team-run MAE 2.2, margin MAE 2.0
embedded_savant_run_environment: winner 3/5 (0.6), team-run MAE 2.2, margin MAE 1.6
fundamental_edge_target_review: winner 3/5 (0.6), team-run MAE 2.3, margin MAE 1.8
capped_target: winner 3/5 (0.6), team-run MAE 2.3, margin MAE 1.8
simulation_result: winner 3/5 (0.6), team-run MAE 2.3, margin MAE 1.8
published_final: winner 3/5 (0.6), team-run MAE 2.3, margin MAE 1.8

Probability lineage

COL@NYY: raw 0.7103 → calibrated 0.6858 → published 0.58; score coherence followed by matchup-recency guardrail
HOU@PHI: raw 0.6273 → calibrated 0.6031 → published 0.5669; score coherence followed by matchup-recency guardrail
PIT@CHW: raw 0.5 → calibrated 0.5711 → published 0.5283; no recorded transformation explains the publication delta
TB@ATL: raw 0.5 → calibrated 0.508 → published 0.5058; published probability matches the calibrated/governed value
TEX@SEA: raw 0.5634 → calibrated 0.5378 → published 0.5273; score coherence followed by matchup-recency guardrail

Every game, in detail

COL@NYY — ✅ hit

Predicted NYY 4-8 (conf +0.16, moderate) · actual NYY 3-10
Root cause: strong_correct_call
Errors: winner correct · Brier 0.1764 · team-run err 3.0 · total bias -1

HOU@PHI — ❌ miss

Predicted PHI 4-5 (conf +0.13, moderate) · actual HOU 2-1
Root cause: Starting-pitcher projection was wrong
Secondary cause: Run environment far lower than projected (over-dispersed)
Errors: winner WRONG · Brier 0.321329 · team-run err 6.0 · total bias +6
Contributing factors:
total runs over-projected by 6 (9 vs 3)
PHI offense over-projected by 4
Cristian Javier (HOU) allowed 4 fewer ER than projected (1 vs 5)
committed to PHI at 0.13 confidence and it lost

PIT@CHW — ✅ hit

Predicted PIT 4-3 (conf -0.06, lean) · actual PIT 2-0
Root cause: right_winner_wrong_score
Errors: winner correct · Brier 0.222516 · team-run err 5.0 · total bias +5
Contributing factors:
total runs over-projected by 5 (7 vs 2)
CHW offense over-projected by 3
Jared Jones (PIT) allowed 3 fewer ER than projected (0 vs 3)

TB@ATL — ❌ miss

Predicted TB 6-5 (conf -0.01, toss_up) · actual ATL 1-3
Root cause: Run environment far lower than projected (over-dispersed)
Secondary cause: Genuine toss-up that flipped
Errors: winner WRONG · Brier 0.255836 · team-run err 7.0 · total bias +7
Contributing factors:
total runs over-projected by 7 (11 vs 4)
TB offense over-projected by 5

TEX@SEA — ✅ hit

Predicted SEA 4-5 (conf +0.05, lean) · actual SEA 3-4
Root cause: solid_correct_call
Errors: winner correct · Brier 0.223449 · team-run err 2.0 · total bias +2

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 0 hurt
Would help:
HOU@PHI (miss): lost a 0.13-confidence pick; wider tails lower overconfidence

player.pitcher.v4 form weights / player.pitcher_prior_games

Targets: starter_er_mae, starter_outs_mae
Net: 2 helped vs 2 hurt
Would help:
HOU@PHI (miss): starter ER off by 4
PIT@CHW (hit): starter ER off by 3
Would hurt / risk:
COL@NYY (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
TEX@SEA (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: 3 helped vs 1 hurt
Would help:
HOU@PHI (miss): team runs off by 6
PIT@CHW (hit): team runs off by 5
TB@ATL (miss): team runs off by 7
Would hurt / risk:
TEX@SEA (hit): team runs already close (off 2); reweight risks disturbing it