Deep slate report — 20260828
Generated 2026-08-29T02:02:26.825188-04:00 · schema deep_slate_diagnosis_v3 · fuses prediction review + forensic deep audit + official starter lines, with a Luna post-mortem.
Process, end to end
1. Settle review
complete
15 games vs official finals
2. Forensic deep audit
complete
per-game probability + pitcher impact
3. Deep diagnosis
built
8 losses attributed to a root cause
4. Luna post-mortem
error
no validated Luna coverage receipt
5. Adjustment
applied_to_production
confidence.moderate_cap
6. Applied to production
champion changed
armed for rollback; next slates measure improve/worsen
Slate summary
Mean total-run bias (losses)
0
What drove the losses
Small edge on the losing side: 1Several errors of similar size: 1Run environment far lower than projected (over-dispersed): 1Run environment far higher than projected (under-dispersed): 1Starting-pitcher projection was wrong: 2Backed the wrong side with real conviction: 2
Evidence coverage
This is the source receipt: a source shown as missing was not silently assumed.
| Source | Status | Evidence supplied | Path | Bytes |
|---|
| prediction review | loaded | official outcomes and scoring errors | /app/output/reports/prediction_review_20260828.json | 373165 |
| forensic deep audit | loaded | probability, provenance, starter and decision contracts | /app/output/reports/PREDICTION_SYSTEM_DEEP_AUDIT_20260828.json | 241183 |
| continuous learning | loaded | pipeline-stage quality, calibration, probability lineage, score distribution, and official starter comparisons | /app/output/learning/continuous_learning_daily_20260828.json | 2424430 |
| detailed report | loaded | full game and simulation rendering | /app/output/reports/detailed_20260828.html | 356746 |
| ultra detailed player audit | loaded | player-level prediction trace | /app/output/reports/20260828_ultra_detailed_player_audit.html | 6285323 |
| structural audit | loaded | all invariant findings | /app/output/reports/audit_report_20260828.html | 374846 |
| integrity revision audit | missing | post-Luna remediation proof for immutable revision artifacts | /app/output/reports/audit_report_20260828_integrity_revision.html | n/a |
High-level causal issue register
Deterministic findings across data integrity, model validation, calibration, score distribution, pipeline stages, simulation, and coverage. Critical is reserved for broken evidence or an unexplained contract violation; high marks material model/process risk.
| Severity | Domain | Finding and evidence | Affected | Action and guardrail |
|---|
| HIGH | artifact_integrity | Blocking canonical artifact error
structural audit totals: {'critical': 0, 'error': 1, 'warning': 0, 'info': 108} | 20260828_SEA@TOR | quarantine the slate from parameter learning and repair the generation boundary
Guardrail: do not attribute prediction accuracy changes to a malformed artifact |
| HIGH | calibration | Published confidence buckets were poorly aligned on this slate
weighted absolute calibration error 0.2179 across 15 games | ARI@SF, BAL@OAK, BOS@NYY, CHW@MIN, CIN@CHC, COL@ATL, HOU@NYM, KC@CLE, LAD@DET, MIA@WAS, PHI@LAA, PIT@STL, SD@TB, SEA@TOR, TEX@MIL | monitor the same buckets over the registered rolling window before changing calibration weights
Guardrail: seven games cannot establish a stable calibration curve |
| HIGH | model_validation | Validated run models explain very little holdout variance
home/away/total R² = 0.0459/0.0247/0.0212 | ARI@SF, BAL@OAK, BOS@NYY, CHW@MIN, CIN@CHC, COL@ATL, HOU@NYM, KC@CLE, LAD@DET, MIA@WAS, PHI@LAA, PIT@STL, SD@TB, SEA@TOR, TEX@MIL | treat exact scores as low-confidence estimates and investigate feature/target signal before tuning tails
Guardrail: positive R² means better than the mean baseline, not that exact scores are reliable |
| INFO | coverage | Every settled game received structural QA
15/15 games structurally audited | ARI@SF, BAL@OAK, BOS@NYY, CHW@MIN, CIN@CHC, COL@ATL, HOU@NYM, KC@CLE, LAD@DET, MIA@WAS, PHI@LAA, PIT@STL, SD@TB, SEA@TOR, TEX@MIL | retain the per-game receipt in every report
Guardrail: complete coverage does not erase findings inside the audited games |
| INFO | probability_lineage | Material calibrated-to-published movements have recorded lineage
5 material movement(s), all explained by named score-coherence and/or recency transformations | ARI@SF, BOS@NYY, CIN@CHC, PIT@STL, SEA@TOR | keep displaying each probability layer and transformation
Guardrail: a higher published value is not automatically a governor bypass |
| INFO | simulation | Simulation produced no winner-direction intervention on this slate
flip candidates 0; published flips 0 | ARI@SF, BAL@OAK, BOS@NYY, CHW@MIN, CIN@CHC, COL@ATL, HOU@NYM, KC@CLE, LAD@DET, MIA@WAS, PHI@LAA, PIT@STL, SD@TB, SEA@TOR, TEX@MIL | measure simulation contribution over rolling slates and distinguish score realization from winner flipping
Guardrail: zero flips is not itself a defect and does not mean simulation did no score-reconciliation work |
Pipeline stage regression
Every stored stage is scored against the same official finals. A later recovery remains visible; it does not erase an unstable intermediate flip.
| Stage | Winner hits | Accuracy | Team-run MAE | Total-run MAE | Margin MAE |
|---|
| historical_ml | 7/15 | 46.7% | 2.615 | 3.893 | 3.491 |
| variance_applied | 5/15 | 33.3% | 2.6 | 3.867 | 3.467 |
| enhanced_prediction | 7/15 | 46.7% | 2.333 | 3.333 | 3.333 |
| embedded_savant_run_environment | 7/15 | 46.7% | 2.267 | 3.333 | 3.467 |
| fundamental_edge_target_review | 7/15 | 46.7% | 2.267 | 3.333 | 3.467 |
| capped_target | 7/15 | 46.7% | 2.267 | 3.333 | 3.467 |
| simulation_result | 7/15 | 46.7% | 2.267 | 3.333 | 3.467 |
| published_final | 7/15 | 46.7% | 2.267 | 3.333 | 3.467 |
| Transition | Δ accuracy | Δ hits | Δ team-run MAE | Δ margin MAE | Games whose winner/tie changed |
|---|
| historical_ml → variance_applied | -0.1334 | -2 | -0.0154 | -0.0246 | BAL@OAK: BAL → TIE; CHW@MIN: MIN → TIE; LAD@DET: LAD → TIE |
| variance_applied → enhanced_prediction | 0.1334 | 2 | -0.2667 | -0.1334 | BAL@OAK: TIE → BAL; CHW@MIN: TIE → MIN; LAD@DET: TIE → LAD |
| enhanced_prediction → embedded_savant_run_environment | 0 | 0 | -0.0666 | 0.1334 | |
| embedded_savant_run_environment → fundamental_edge_target_review | 0 | 0 | 0 | 0 | |
| fundamental_edge_target_review → capped_target | 0 | 0 | 0 | 0 | |
| capped_target → simulation_result | 0 | 0 | 0 | 0 | |
| simulation_result → published_final | 0 | 0 | 0 | 0 | |
Probability lineage and calibration
All columns are pick-side probabilities, so away favorites are not accidentally compared using home-side semantics.
| Game | Outcome | Raw model | Calibrated | Governor | Score coherence | Published | Explanation |
|---|
| ARI@SF | miss | 0.5085 | 0.5082 | 0.5082 | 0.5466 | 0.5337 | score coherence followed by matchup-recency guardrail |
| BAL@OAK | hit | 0.5495 | 0.5483 | 0.5483 | 0.5666 | 0.5481 | published probability matches the calibrated/governed value |
| BOS@NYY | hit | 0.5085 | 0.5082 | 0.5082 | 0.5466 | 0.53 | score coherence followed by matchup-recency guardrail |
| CHW@MIN | miss | 0.5085 | 0.5082 | 0.5082 | 0.5466 | 0.5269 | score coherence followed by matchup-recency guardrail |
| CIN@CHC | miss | 0.659 | 0.6508 | 0.6508 | 0.6604 | 0.56 | score coherence followed by matchup-recency guardrail |
| COL@ATL | hit | 0.8613 | 0.8553 | 0.8553 | 0.8477 | 0.8477 | score_coherence aligned probability to the published score margin |
| HOU@NYM | miss | 0.5305 | 0.5297 | 0.5297 | 0.5573 | 0.5487 | no recorded transformation explains the publication delta |
| KC@CLE | miss | 0.5922 | 0.5887 | 0.5887 | 0.5869 | 0.5738 | no recorded transformation explains the publication delta |
| LAD@DET | hit | 0.67 | 0.6607 | 0.6607 | 0.6654 | 0.6654 | score_coherence aligned probability to the published score margin |
| MIA@WAS | miss | 0.568 | 0.5661 | 0.5661 | 0.5755 | 0.5546 | score coherence followed by matchup-recency guardrail |
| PHI@LAA | miss | 0.5721 | 0.57 | 0.57 | 0.5775 | 0.5659 | published probability matches the calibrated/governed value |
| PIT@STL | miss | 0.6634 | 0.6547 | 0.6547 | 0.6623 | 0.6 | score coherence followed by matchup-recency guardrail |
| SD@TB | hit | 0.5507 | 0.5495 | 0.5495 | 0.5672 | 0.5572 | no recorded transformation explains the publication delta |
| SEA@TOR | hit | 0.5833 | 0.5798 | 0.5798 | 0.5824 | 0.53 | score coherence followed by matchup-recency guardrail |
| TEX@MIL | hit | 0.8613 | 0.8553 | 0.8553 | 0.8477 | 0.8477 | score_coherence aligned probability to the published score margin |
Score distribution and margin compression
Predicted margins ≤2: 12 · Actual margins ≥4: 5
| Game | Pred margin | Actual margin | Pred − actual |
|---|
| ARI@SF | 1 | 4 | -3 |
| BAL@OAK | 1 | 1 | 0 |
| BOS@NYY | 2 | 1 | 1 |
| CHW@MIN | 1 | 3 | -2 |
| CIN@CHC | 2 | 2 | 0 |
| COL@ATL | 5 | 2 | 3 |
| HOU@NYM | 1 | 2 | -1 |
| KC@CLE | 1 | 3 | -2 |
| LAD@DET | 3 | 1 | 2 |
| MIA@WAS | 1 | 7 | -6 |
| PHI@LAA | 1 | 2 | -1 |
| PIT@STL | 2 | 3 | -1 |
| SD@TB | 1 | 5 | -4 |
| SEA@TOR | 2 | 7 | -5 |
| TEX@MIL | 4 | 5 | -1 |
Per-game evidence coverage
Every game, in depth (15)
ARI@SFMISS
Predicted: SF 5-6 (conf +0.07, lean)Actual: ARI 10-6Brier 0.285total bias -5
Root cause: Starting-pitcher projection was wrong
Evidence-backed diagnosis- published system edge 0.07 (lean); this is not the half-scale distance from 50%
- team-run absolute error 5; total bias -5
- pregame home-minus-away evidence: season -0.418, recent +0.014, pitcher -0.025, composite -0.101
- confidence pipeline reached 0.508 before later publication/recency controls; published home probability was 0.534
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs under-projected by 5 (11 vs 16)
- ARI offense under-projected by 5
- Blade Tidwell (SF) allowed 5 more ER than projected (6 vs 1)
| Team | Starter | Projected | Actual | ER err |
|---|
| ARI | Kohl Drake | 6 ER / 13 outs | 5 ER / 12 outs | -1 |
| SF | Blade Tidwell | 1 ER / 15 outs | 6 ER / 15 outs | 5 |
Decision internals: run factors home×0.974 / away×0.999; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.418 | positive favors home; negative favors away |
| Recent form | 0.014 | positive favors home; negative favors away |
| Starting-pitcher matchup | -0.025 | positive favors home; negative favors away |
| Composite | -0.101 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.89-5.28 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-6 | 0.534 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-6 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-6 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-6 | n/a | Capped at max 12 realistic runs |
| simulation_result | 5-6 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.89-5.28 | 4-5 | -0.389 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 5-6 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-6 | 5-6 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-6 | 5-6 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-6 | 5-6 | 0 | Capped at max 12 realistic runs |
| simulation_result | 5-6 | 5-6 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| SF | Andrew Knizner | batter / C | better | predicted_RBI 3.67 |
| ARI | Tim Tawa | batter / 1B | better | predicted_RBI 1.7 |
| SF | Shay Whitcomb | batter / 3B | worse | predicted_AB 1.5 |
| ARI | James McCann | batter / C | worse | predicted_RBI 1.49 |
| SF | Osleivis Basabe | batter / 2B | worse | predicted_H 1.25 |
| ARI | Nolan Arenado | batter / 3B | worse | predicted_SO 1.192 |
| SF | Rafael Devers | batter / 1B | worse | predicted_H -1.14 |
| SF | Turner Hill | batter / LF | worse | predicted_AB 1.111 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | ARI pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | ARI |
| info | CURRENT_PITCHER_DUPLICATION | SF pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | SF |
| info | CURRENT_FINALIZED_MATCH_BOX | ARI model-prediction batting totals differ from final simulation box score (AB: current=35, sim=34; H: current=8, sim=7; 2B: current=2, sim=0; RBI: current=4, sim=5; R: current=4, sim=5; BB: current=3, sim=4; SO: current=7, sim=5). | ARI |
| info | CURRENT_PLAYER_MATCH_BOX | ARI current finalized hitter lines are model predictions and differ from the simulation box score: Gabriel Moreno: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Nolan Arenado: HR current=1 box=0, RBI current=1 box=0, SO current=1 box=2; Geraldo Perdomo: AB current=4 box=3, RBI current=0 box=1, R current=0 box=1, SO current=0 box=1; Lars Nootbaar: R current=0 box=1, BB current=0 box=1; Ildemaro Vargas: H current=1 box=0, RBI current=1 box=0; Corbin Carroll: AB current=5 box=4, H current=1 box=0, 2B current=1 box=0, R current=1 box=0; James McCann: RBI current=1 box=2, R current=0 box=1; Tim Tawa: H current=0 box=2, HR current=0 box=1, RBI current=0 box=2, R current=0 box=1 | ARI |
| info | CURRENT_FINALIZED_MATCH_BOX | SF model-prediction batting totals differ from final simulation box score (AB: current=33, sim=32; 2B: current=2, sim=1; 3B: current=0, sim=1; HR: current=1, sim=3; RBI: current=4, sim=6; R: current=4, sim=6; SO: current=9, sim=7). | SF |
| info | CURRENT_PLAYER_MATCH_BOX | SF current finalized hitter lines are model predictions and differ from the simulation box score: Jonah Cox: H current=1 box=0, RBI current=1 box=0, BB current=0 box=1, SO current=2 box=1; Andrew Knizner: AB current=3 box=4, H current=1 box=3, 3B current=0 box=1, HR current=0 box=1; Rafael Devers: AB current=5 box=3, H current=1 box=0, HR current=1 box=0, RBI current=1 box=0; Turner Hill: AB current=3 box=4, 2B current=1 box=0, HR current=0 box=1, RBI current=0 box=1; Christian Koss: H current=1 box=0; Osleivis Basabe: H current=1 box=2, 2B current=0 box=1, R current=0 box=1, SO current=1 box=0; Bryce Eldridge: AB current=5 box=4, HR current=0 box=1, BB current=1 box=0, SO current=1 box=2; Jung Hoo Lee: AB current=5 box=4, 2B current=1 box=0 | SF |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
BAL@OAKHIT
Predicted: BAL 6-5 (conf -0.10, lean)Actual: BAL 4-3Brier 0.204total bias +4
Root cause: Solid correct call
Evidence-backed diagnosis- published system edge 0.10 (lean); this is not the half-scale distance from 50%
- team-run absolute error 4; total bias +4
- pregame home-minus-away evidence: season -0.924, recent -0.043, pitcher +0.010, composite -0.678
- confidence pipeline reached 0.548 before later publication/recency controls; published home probability was 0.452
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs over-projected by 4 (11 vs 7)
- Brandon Young (BAL) allowed 3 fewer ER than projected (1 vs 4)
- Jacob Lopez (OAK) allowed 5 fewer ER than projected (1 vs 6)
| Team | Starter | Projected | Actual | ER err |
|---|
| BAL | Brandon Young | 4 ER / 16 outs | 1 ER / 21 outs | -3 |
| OAK | Jacob Lopez | 6 ER / 15 outs | 1 ER / 15 outs | -5 |
Decision internals: run factors home×0.998 / away×0.988; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.924 | positive favors home; negative favors away |
| Recent form | -0.043 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.01 | positive favors home; negative favors away |
| Composite | -0.678 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 5.49-5.36 | n/a | Base ML historical model prediction |
| variance_applied | 5-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 6-5 | 0.548 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 6-5 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 6-5 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 6-5 | n/a | Capped at max 12 realistic runs |
| simulation_result | 6-5 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 5.49-5.36 | 5-5 | -0.125 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-5 | 6-5 | 1 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 6-5 | 6-5 | -0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 6-5 | 6-5 | -0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 6-5 | 6-5 | -0 | Capped at max 12 realistic runs |
| simulation_result | 6-5 | 6-5 | -0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| BAL | Jackson Holliday | batter / 2B | better | predicted_RBI 2.64 |
| BAL | Carlos Narváez | batter / C | worse | predicted_BB 1.787 |
| OAK | Jeff McNeil | batter / 2B | worse | predicted_AB 1.757 |
| OAK | Max Muncy | batter / DH | better | predicted_RBI 1.621 |
| BAL | Christian Franklin | batter / CF | worse | predicted_AB 1.5 |
| OAK | Henry Bolte | batter / CF | better | predicted_AB 1.494 |
| BAL | Pete Alonso | batter / DH | better | predicted_RBI 1.305 |
| OAK | Alika Williams | batter / SS | worse | predicted_AB 1.291 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | BAL pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | BAL |
| info | CURRENT_PITCHER_DUPLICATION | OAK pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | OAK |
| info | CURRENT_FINALIZED_MATCH_BOX | BAL model-prediction batting totals differ from final simulation box score (AB: current=33, sim=32; H: current=9, sim=5; 2B: current=2, sim=0; HR: current=1, sim=2; RBI: current=5, sim=6; R: current=5, sim=6; BB: current=3, sim=4; SO: current=10, sim=6). | BAL |
| info | CURRENT_PLAYER_MATCH_BOX | BAL current finalized hitter lines are model predictions and differ from the simulation box score: Blaze Alexander: AB current=3 box=4, RBI current=0 box=1; Pete Alonso: AB current=5 box=3, RBI current=1 box=2, SO current=2 box=0; Coby Mayo: H current=1 box=0, RBI current=1 box=0, R current=1 box=0, BB current=0 box=1; Leody Taveras: AB current=3 box=4, RBI current=1 box=0, SO current=1 box=2; Christian Franklin: AB current=3 box=4, H current=1 box=0, BB current=1 box=0; Dylan Beavers: AB current=3 box=4, 2B current=1 box=0, RBI current=1 box=0, R current=0 box=1; Christian Encarnacion-Strand: AB current=5 box=4, H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0; Carlos Narváez: AB current=3 box=2, H current=1 box=0, R current=0 box=1, BB current=0 box=2 | BAL |
| info | CURRENT_FINALIZED_MATCH_BOX | OAK model-prediction batting totals differ from final simulation box score (AB: current=31, sim=38; H: current=8, sim=11; 2B: current=2, sim=3; HR: current=1, sim=0; RBI: current=3, sim=5; R: current=4, sim=5; BB: current=3, sim=1; SO: current=8, sim=5). | OAK |
| info | CURRENT_PLAYER_MATCH_BOX | OAK current finalized hitter lines are model predictions and differ from the simulation box score: Henry Bolte: AB current=4 box=5, RBI current=0 box=1, BB current=1 box=0, SO current=1 box=0; Tommy White: 2B current=0 box=1, RBI current=0 box=1; Max Muncy: AB current=3 box=4, H current=0 box=2, 2B current=0 box=1, RBI current=0 box=2; Carlos Cortes: H current=1 box=0, BB current=0 box=1; Jeff McNeil: AB current=3 box=5, H current=1 box=2, R current=0 box=1, SO current=0 box=1; Alika Williams: AB current=3 box=4, 2B current=1 box=0; Jonah Heim: 2B current=1 box=0, RBI current=1 box=0, BB current=1 box=0, SO current=1 box=0; Zack Gelof: AB current=4 box=5 | OAK |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
BOS@NYYHIT
Predicted: NYY 4-6 (conf +0.06, lean)Actual: NYY 0-1Brier 0.221total bias +9
Root cause: Right winner, wrong score
Evidence-backed diagnosis- published system edge 0.06 (lean); this is not the half-scale distance from 50%
- team-run absolute error 9; total bias +9
- pregame home-minus-away evidence: season +0.007, recent -0.218, pitcher +0.006, composite +0.322
- confidence pipeline reached 0.508 before later publication/recency controls; published home probability was 0.530
- score-volatility actions: no_change
- structural audit exposed 9 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs over-projected by 9 (10 vs 1)
- BOS offense over-projected by 4
- NYY offense over-projected by 5
- Patrick Sandoval (BOS) allowed 3 fewer ER than projected (1 vs 4)
| Team | Starter | Projected | Actual | ER err |
|---|
| BOS | Patrick Sandoval | 4 ER / 15 outs | 1 ER / 16 outs | -3 |
| NYY | Cam Schlittler | 2 ER / 18 outs | 0 ER / 17 outs | -2 |
Decision internals: run factors home×1.039 / away×1.033; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 0.007 | positive favors home; negative favors away |
| Recent form | -0.218 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.006 | positive favors home; negative favors away |
| Composite | 0.322 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.83-5.4 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-6 | 0.53 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 4-6 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-6 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-6 | n/a | Capped at max 12 realistic runs |
| simulation_result | 4-6 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.83-5.4 | 4-5 | -0.573 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 5-6 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-6 | 4-6 | 1 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-6 | 4-6 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-6 | 4-6 | 0 | Capped at max 12 realistic runs |
| simulation_result | 4-6 | 4-6 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| NYY | Paul Goldschmidt | batter / 1B | better | predicted_RBI 2.64 |
| BOS | Anthony Seigler | batter / 2B | worse | predicted_SO 2.378 |
| NYY | Heliot Ramos | batter / RF | better | predicted_H 2.122 |
| BOS | Wilyer Abreu | batter / RF | worse | predicted_SO 2.109 |
| BOS | Patrick Sandoval | pitcher / Starter | boosted | predicted_H 2 |
| NYY | Cam Schlittler | pitcher / Starter | better | predicted_ER -2 |
| BOS | Ceddanne Rafaela | batter / CF | better | predicted_H 1.87 |
| BOS | Adley Rutschman | batter / C | better | predicted_RBI 1.44 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | BOS pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | BOS |
| info | CURRENT_PITCHER_DUPLICATION | NYY pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | NYY |
| info | CURRENT_FINALIZED_MATCH_BOX | BOS model pitcher projections differ from final simulation pitching box score (H: current=8, sim=9). | BOS |
| info | CURRENT_FINALIZED_MATCH_BOX | BOS model-prediction batting totals differ from final simulation box score (AB: current=36, sim=35; 2B: current=2, sim=3; HR: current=1, sim=0; R: current=5, sim=4; BB: current=4, sim=6; SO: current=7, sim=10). | BOS |
| info | CURRENT_PLAYER_MATCH_BOX | BOS current finalized hitter lines are model predictions and differ from the simulation box score: Nick Sogard: AB current=4 box=3, H current=1 box=0, 2B current=1 box=0, R current=1 box=0; Adley Rutschman: AB current=5 box=4, 2B current=0 box=1, RBI current=1 box=2, R current=0 box=1; Ceddanne Rafaela: H current=1 box=3, 2B current=0 box=1, R current=0 box=1, SO current=1 box=0; Willson Contreras: H current=1 box=2, RBI current=1 box=0; Wilyer Abreu: R current=1 box=0, SO current=1 box=3; Caleb Durbin: H current=1 box=0, SO current=0 box=1; Mickey Gasper: H current=1 box=2, 2B current=0 box=1, HR current=1 box=0, RBI current=1 box=2; Anthony Seigler: AB current=3 box=4, H current=1 box=0, 2B current=1 box=0, SO current=1 box=3 | BOS |
| info | CURRENT_FINALIZED_MATCH_BOX | NYY model pitcher projections differ from final simulation pitching box score (BB: current=5, sim=6). | NYY |
| info | CURRENT_FINALIZED_MATCH_BOX | NYY model-prediction batting totals differ from final simulation box score (AB: current=33, sim=35; H: current=8, sim=9; 2B: current=1, sim=0; HR: current=1, sim=2; RBI: current=4, sim=6; R: current=4, sim=6; SO: current=9, sim=7; SF: current=0, sim=1). | NYY |
| info | CURRENT_PLAYER_MATCH_BOX | NYY current finalized hitter lines are model predictions and differ from the simulation box score: Heliot Ramos: H current=1 box=3, HR current=0 box=1, RBI current=0 box=1, R current=0 box=1; Cody Bellinger: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Ben Rice: AB current=4 box=5, H current=1 box=2, HR current=1 box=0, BB current=1 box=0; Paul Goldschmidt: AB current=4 box=5, HR current=0 box=1, RBI current=0 box=3, R current=0 box=1; José Caballero: AB current=3 box=4, SO current=1 box=0; Trent Grisham: AB current=4 box=3, H current=1 box=0, BB current=1 box=0, SO current=1 box=2; George Lombard Jr.: H current=1 box=0, R current=1 box=0; Jazz Chisholm Jr.: RBI current=1 box=0, R current=0 box=1, BB current=0 box=1 | NYY |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
CHW@MINMISS
Predicted: MIN 5-6 (conf +0.05, lean)Actual: CHW 7-4Brier 0.278total bias +0
Root cause: Small edge on the losing side
Evidence-backed diagnosis- published system edge 0.05 (lean); this is not the half-scale distance from 50%
- team-run absolute error 4; total bias +0
- pregame home-minus-away evidence: season -0.606, recent -0.274, pitcher +0.085, composite -0.035
- confidence pipeline reached 0.508 before later publication/recency controls; published home probability was 0.527
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
No large single-factor errors.
| Team | Starter | Projected | Actual | ER err |
|---|
| CHW | Luis Castillo | 2 ER / 15 outs | 4 ER / 12 outs | 2 |
| MIN | Dean Kremer | 4 ER / 15 outs | 5 ER / 12 outs | 1 |
Decision internals: run factors home×1.066 / away×0.981; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.606 | positive favors home; negative favors away |
| Recent form | -0.274 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.085 | positive favors home; negative favors away |
| Composite | -0.035 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 4.17-4.46 | n/a | Base ML historical model prediction |
| variance_applied | 4-4 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-6 | 0.527 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by winner_starter_is_below_average, projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-6 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-6 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-6 | n/a | Capped at max 12 realistic runs |
| simulation_result | 5-6 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 4.17-4.46 | 4-4 | -0.286 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-4 | 5-6 | 1 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by winner_starter_is_below_average, projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-6 | 5-6 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-6 | 5-6 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-6 | 5-6 | 0 | Capped at max 12 realistic runs |
| simulation_result | 5-6 | 5-6 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| CHW | Munetaka Murakami | batter / 1B | better | predicted_RBI 2.408 |
| CHW | Luis Castillo | pitcher / Starter | better | predicted_SO 2 |
| CHW | José Urquidy | pitcher / Reliever | worse | predicted_H 2 |
| MIN | Josh Bell | batter / DH | boosted | predicted_H 1.99 |
| MIN | Trevor Larnach | batter / LF | better | predicted_2B 1.819 |
| MIN | Brooks Lee | batter / 3B | worse | predicted_AB 1.71 |
| MIN | Player-805805 | batter / CF | better | predicted_RBI 1.627 |
| MIN | Luke Keaschall | batter / RF | boosted | predicted_AB 1.556 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | CHW pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | CHW |
| info | CURRENT_PITCHER_DUPLICATION | MIN pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | MIN |
| info | CURRENT_FINALIZED_MATCH_BOX | CHW model-prediction batting totals differ from final simulation box score (AB: current=33, sim=36; H: current=9, sim=6; 2B: current=2, sim=3; HR: current=1, sim=2; R: current=6, sim=5; BB: current=4, sim=2; SO: current=9, sim=6). | CHW |
| info | CURRENT_PLAYER_MATCH_BOX | CHW current finalized hitter lines are model predictions and differ from the simulation box score: Sam Antonacci: AB current=3 box=4, H current=1 box=0, BB current=0 box=1; Munetaka Murakami: H current=1 box=3, HR current=0 box=2, RBI current=1 box=3, SO current=2 box=0; Braden Montgomery: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Miguel Vargas: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Andrew Benintendi: AB current=3 box=4, 2B current=0 box=1, R current=0 box=1, SO current=1 box=2; Colson Montgomery: H current=1 box=0, SO current=1 box=2; Chase Meidroth: 2B current=0 box=1, BB current=1 box=0, SO current=1 box=0; Drew Romo: AB current=3 box=4, H current=1 box=0, SO current=1 box=0 | CHW |
| info | CURRENT_FINALIZED_MATCH_BOX | MIN model-prediction batting totals differ from final simulation box score (AB: current=35, sim=43; H: current=8, sim=13; 2B: current=2, sim=5; HR: current=1, sim=2; RBI: current=4, sim=6; R: current=4, sim=6; BB: current=3, sim=5; SO: current=7, sim=10). | MIN |
| info | CURRENT_PLAYER_MATCH_BOX | MIN current finalized hitter lines are model predictions and differ from the simulation box score: Luke Keaschall: AB current=4 box=5; Kody Clemens: H current=1 box=2, HR current=0 box=1, R current=0 box=1, BB current=0 box=1; Brooks Lee: AB current=4 box=6, H current=1 box=2, RBI current=1 box=0; Royce Lewis: AB current=4 box=5, H current=1 box=0, SO current=1 box=2; Josh Bell: AB current=4 box=5, H current=1 box=3, 2B current=0 box=2, BB current=1 box=0; Player-805805: AB current=3 box=4, H current=0 box=2, HR current=0 box=1, RBI current=0 box=2; Ryan Jeffers: AB current=4 box=5, H current=1 box=0, 2B current=1 box=0, R current=1 box=0; Kaelen Culpepper: HR current=1 box=0, BB current=0 box=1 | MIN |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
CIN@CHCMISS
Predicted: CHC 5-7 (conf +0.12, moderate)Actual: CIN 10-8Brier 0.314total bias -6
Root cause: Run environment far higher than projected (under-dispersed)
Evidence-backed diagnosis- published system edge 0.12 (moderate); this is not the half-scale distance from 50%
- team-run absolute error 6; total bias -6
- pregame home-minus-away evidence: season +1.274, recent +0.546, pitcher +0.061, composite +1.332
- confidence pipeline reached 0.651 before later publication/recency controls; published home probability was 0.560
- score-volatility actions: supported_blowout_margin_expansion
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs under-projected by 6 (12 vs 18)
- CIN offense under-projected by 5
- committed to CHC at 0.12 confidence and it lost
| Team | Starter | Projected | Actual | ER err |
|---|
| CIN | Rhett Lowder | 2 ER / 15 outs | 3 ER / 15 outs | 1 |
| CHC | David Peterson | 4 ER / 14 outs | 2 ER / 15 outs | -2 |
Decision internals: run factors home×1.105 / away×1.044; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 1.274 | positive favors home; negative favors away |
| Recent form | 0.546 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.061 | positive favors home; negative favors away |
| Composite | 1.332 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 4.35-4.95 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-7 | 0.56 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by winner_starter_is_below_average, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-7 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-7 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-7 | n/a | Capped at max 12 realistic runs |
| simulation_result | 5-7 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 4.35-4.95 | 4-5 | 0.401 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 5-7 | 1 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by winner_starter_is_below_average, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-7 | 5-7 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-7 | 5-7 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-7 | 5-7 | 0 | Capped at max 12 realistic runs |
| simulation_result | 5-7 | 5-7 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| CIN | Sam Moll | pitcher / Reliever | boosted | predicted_ER 2 |
| CIN | Dane Myers | batter / CF | worse | predicted_RBI 1.73 |
| CHC | Ian Happ | batter / LF | worse | predicted_H 1.66 |
| CIN | Matt McLain | batter / 2B | better | predicted_AB 1.587 |
| CHC | Nico Hoerner | batter / SS | better | predicted_RBI 1.543 |
| CHC | Michael Conforto | batter / DH | worse | predicted_AB 1.5 |
| CHC | Seiya Suzuki | batter / RF | better | predicted_RBI 1.49 |
| CIN | Eugenio Suárez | batter / 3B | worse | predicted_AB 1.485 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | CHC pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | CHC |
| info | CURRENT_PITCHER_DUPLICATION | CIN pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | CIN |
| info | CURRENT_FINALIZED_MATCH_BOX | CHC model-prediction batting totals differ from final simulation box score (AB: current=35, sim=43; H: current=8, sim=13; 2B: current=2, sim=3; HR: current=1, sim=0; RBI: current=5, sim=7; R: current=5, sim=7; BB: current=4, sim=6; SO: current=8, sim=6). | CHC |
| info | CURRENT_PLAYER_MATCH_BOX | CHC current finalized hitter lines are model predictions and differ from the simulation box score: Seiya Suzuki: AB current=4 box=5, H current=1 box=2, RBI current=0 box=2, SO current=1 box=2; Alex Bregman: AB current=4 box=5, H current=1 box=2, 2B current=0 box=1, RBI current=1 box=0; Nico Hoerner: AB current=4 box=5, H current=1 box=2, 2B current=0 box=1, RBI current=0 box=2; Pete Crow-Armstrong: AB current=5 box=6, H current=1 box=2, HR current=1 box=0, BB current=1 box=0; Michael Busch: H current=1 box=0, 2B current=1 box=0, R current=1 box=0, BB current=1 box=2; Ian Happ: AB current=4 box=5, H current=0 box=2, BB current=1 box=0; Pedro Ramirez: AB current=3 box=4, H current=1 box=2, R current=0 box=1, BB current=0 box=1; Michael Conforto: AB current=3 box=4, H current=1 box=0, RBI current=1 box=0, BB current=0 box=1 | CHC |
| info | CURRENT_FINALIZED_MATCH_BOX | CIN model-prediction batting totals differ from final simulation box score (AB: current=32, sim=40; H: current=8, sim=10; 2B: current=1, sim=4; RBI: current=4, sim=5; R: current=4, sim=5). | CIN |
| info | CURRENT_PLAYER_MATCH_BOX | CIN current finalized hitter lines are model predictions and differ from the simulation box score: Tyler Stephenson: AB current=3 box=4, H current=1 box=0, BB current=0 box=1; Eugenio Suárez: AB current=4 box=5, H current=1 box=0, RBI current=1 box=0, BB current=1 box=0; JJ Bleday: AB current=4 box=5, HR current=1 box=0, RBI current=1 box=0, BB current=1 box=0; Elly De La Cruz: H current=1 box=3, BB current=1 box=0; Matt McLain: AB current=3 box=5, H current=0 box=2, HR current=0 box=1, RBI current=0 box=1; Héctor Rodríguez: AB current=3 box=4, 2B current=0 box=1, SO current=1 box=0; Dane Myers: AB current=3 box=5, H current=1 box=2, 2B current=0 box=1, RBI current=0 box=2; Sal Stewart: 2B current=0 box=1, BB current=0 box=1 | CIN |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
COL@ATLHIT
Predicted: ATL 3-8 (conf +0.70, strong)Actual: ATL 4-6Brier 0.023total bias +1
Root cause: Strong correct call
Evidence-backed diagnosis- published system edge 0.70 (strong); this is not the half-scale distance from 50%
- team-run absolute error 3; total bias +1
- pregame home-minus-away evidence: season +1.586, recent +0.842, pitcher +0.089, composite +2.179
- confidence pipeline reached 0.855 before later publication/recency controls; published home probability was 0.848
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
No large single-factor errors.
| Team | Starter | Projected | Actual | ER err |
|---|
| COL | Tomoyuki Sugano | 8 ER / 16 outs | 6 ER / 12 outs | -2 |
| ATL | Grant Holmes | 3 ER / 15 outs | 3 ER / 15 outs | 0 |
Decision internals: run factors home×1.077 / away×0.988; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 1.586 | positive favors home; negative favors away |
| Recent form | 0.842 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.089 | positive favors home; negative favors away |
| Composite | 2.179 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 4.11-5.61 | n/a | Base ML historical model prediction |
| variance_applied | 4-6 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 3-7 | 0.848 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 3-8 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 3-8 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 3-8 | n/a | Capped at max 12 realistic runs |
| simulation_result | 3-8 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 4.11-5.61 | 4-6 | 0.505 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-6 | 3-7 | 2 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 3-7 | 3-8 | 1 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 3-8 | 3-8 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 3-8 | 3-8 | 0 | Capped at max 12 realistic runs |
| simulation_result | 3-8 | 3-8 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| ATL | Ronald Acuña Jr. | batter / RF | better | predicted_RBI 6.55 |
| COL | Tomoyuki Sugano | pitcher / Starter | neutral | predicted_SO 2 |
| ATL | Drake Baldwin | batter / DH | better | predicted_AB -1.873 |
| COL | Zac Veen | batter / RF | worse | predicted_BB 1.8 |
| ATL | Sean Murphy | batter / C | worse | predicted_SO 1.714 |
| COL | Jake McCarthy | batter / LF | worse | predicted_AB -1.59 |
| ATL | Ozzie Albies | batter / 2B | worse | predicted_AB -1.44 |
| COL | Cole Carrigg | batter / CF | worse | predicted_RBI 1.419 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | ATL pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | ATL |
| info | CURRENT_PITCHER_DUPLICATION | COL pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | COL |
| info | CURRENT_FINALIZED_MATCH_BOX | ATL model-prediction batting totals differ from final simulation box score (AB: current=34, sim=31; H: current=8, sim=7; 2B: current=2, sim=1; HR: current=1, sim=2; RBI: current=4, sim=8; R: current=4, sim=8; BB: current=3, sim=4). | ATL |
| info | CURRENT_PLAYER_MATCH_BOX | ATL current finalized hitter lines are model predictions and differ from the simulation box score: Drake Baldwin: AB current=4 box=2, 2B current=0 box=1, BB current=1 box=2, SO current=1 box=0; Ozzie Albies: AB current=4 box=3, H current=1 box=0, BB current=0 box=1, SO current=1 box=2; Mauricio Dubón: 2B current=1 box=0, RBI current=1 box=0, R current=0 box=1, SO current=0 box=1; Ronald Acuña Jr.: AB current=5 box=4, H current=1 box=2, HR current=0 box=2, RBI current=0 box=7; Matt Olson: AB current=4 box=3, H current=1 box=2, HR current=1 box=0, RBI current=1 box=0; Michael Harris II: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=0 box=1; Mike Yastrzemski: AB current=3 box=4, H current=1 box=0, R current=0 box=1, SO current=1 box=0; Sean Murphy: AB current=3 box=4, H current=1 box=0, SO current=1 box=3 | ATL |
| info | CURRENT_FINALIZED_MATCH_BOX | COL model-prediction batting totals differ from final simulation box score (AB: current=35, sim=32; H: current=8, sim=5; 2B: current=2, sim=1; 3B: current=0, sim=1; RBI: current=4, sim=3; R: current=4, sim=3; BB: current=3, sim=2; SO: current=10, sim=6). | COL |
| info | CURRENT_PLAYER_MATCH_BOX | COL current finalized hitter lines are model predictions and differ from the simulation box score: Cole Carrigg: 2B current=1 box=0, HR current=0 box=1, RBI current=1 box=2, BB current=1 box=0; Willi Castro: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=0; Mickey Moniak: 3B current=0 box=1, R current=0 box=1; Jake McCarthy: AB current=6 box=4, 2B current=1 box=0, SO current=1 box=0; Zac Veen: AB current=3 box=1, H current=1 box=0, RBI current=1 box=0, BB current=0 box=2; Connor Norby: AB current=3 box=4, H current=1 box=0, RBI current=1 box=0, SO current=1 box=0; Hunter Goodman: HR current=1 box=0, R current=1 box=0, SO current=2 box=1; TJ Rumfield: 2B current=0 box=1, BB current=1 box=0 | COL |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
HOU@NYMMISS
Predicted: NYM 3-4 (conf +0.10, lean)Actual: HOU 3-1Brier 0.301total bias +3
Root cause: Several errors of similar size
Evidence-backed diagnosis- published system edge 0.10 (lean); this is not the half-scale distance from 50%
- team-run absolute error 3; total bias +3
- pregame home-minus-away evidence: season -0.038, recent +0.148, pitcher +0.033, composite +0.269
- confidence pipeline reached 0.530 before later publication/recency controls; published home probability was 0.549
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs over-projected by 3 (7 vs 4)
- NYM offense over-projected by 3
| Team | Starter | Projected | Actual | ER err |
|---|
| HOU | Hunter Brown | 3 ER / 16 outs | 1 ER / 20 outs | -2 |
| NYM | Christian Scott | 2 ER / 17 outs | 3 ER / 16 outs | 1 |
Decision internals: run factors home×0.967 / away×0.933; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.038 | positive favors home; negative favors away |
| Recent form | 0.148 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.033 | positive favors home; negative favors away |
| Composite | 0.269 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.91-4.64 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 3-4 | 0.549 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 3-4 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 3-4 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 3-4 | n/a | Capped at max 12 realistic runs |
| simulation_result | 3-4 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.91-4.64 | 4-5 | 0.271 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 3-4 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 3-4 | 3-4 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 3-4 | 3-4 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 3-4 | 3-4 | 0 | Capped at max 12 realistic runs |
| simulation_result | 3-4 | 3-4 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| NYM | Bo Bichette | batter / 3B | worse | predicted_RBI 2.62 |
| NYM | Luis Torrens | batter / C | worse | predicted_SO 1.375 |
| HOU | Yainer Diaz | batter / C | worse | predicted_AB -1.25 |
| HOU | Yordan Alvarez | batter / DH | worse | predicted_H -1.22 |
| HOU | Daulton Varsho | batter / CF | worse | predicted_SO 1.188 |
| NYM | Marcus Semien | batter / 2B | boosted | predicted_H 1.18 |
| NYM | Luis Robert Jr. | batter / LF | better | predicted_H 1.173 |
| NYM | Francisco Lindor | batter / SS | worse | predicted_SO 1.16 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | HOU pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | HOU |
| info | CURRENT_PITCHER_DUPLICATION | NYM pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | NYM |
| info | CURRENT_FINALIZED_MATCH_BOX | HOU model-prediction batting totals differ from final simulation box score (AB: current=36, sim=32; H: current=8, sim=5; 2B: current=1, sim=2; HR: current=1, sim=0; RBI: current=4, sim=3; R: current=4, sim=3; BB: current=3, sim=0; SO: current=8, sim=11). | HOU |
| info | CURRENT_PLAYER_MATCH_BOX | HOU current finalized hitter lines are model predictions and differ from the simulation box score: Isaac Paredes: 2B current=0 box=1, R current=0 box=1, BB current=1 box=0, SO current=0 box=1; Daulton Varsho: 2B current=0 box=1, SO current=1 box=2; Jose Altuve: AB current=5 box=4, H current=1 box=0, R current=1 box=0, BB current=1 box=0; Yordan Alvarez: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Taylor Trammell: H current=1 box=0, SO current=1 box=2; Nelson Velázquez: AB current=3 box=2, SO current=1 box=0, HBP current=0 box=1; Jeremy Peña: AB current=5 box=4, H current=1 box=2, HR current=1 box=0; Yainer Diaz: AB current=4 box=3, H current=1 box=0 | HOU |
| info | CURRENT_FINALIZED_MATCH_BOX | NYM model-prediction batting totals differ from final simulation box score (H: current=9, sim=8; 2B: current=1, sim=2; HR: current=1, sim=0; BB: current=3, sim=2; SO: current=8, sim=9; SF: current=0, sim=1). | NYM |
| info | CURRENT_PLAYER_MATCH_BOX | NYM current finalized hitter lines are model predictions and differ from the simulation box score: A.J. Ewing: AB current=4 box=3, H current=1 box=0, SO current=1 box=2; Carson Benge: 2B current=0 box=1, RBI current=1 box=0, R current=1 box=0; Luis Robert Jr.: H current=1 box=2, HR current=1 box=0, RBI current=1 box=0, BB current=1 box=0; Jared Young: H current=1 box=0, 2B current=1 box=0, BB current=0 box=1, SO current=0 box=1; Brett Baty: AB current=3 box=4, SO current=1 box=0; Francisco Lindor: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=2; Bo Bichette: RBI current=1 box=3, SO current=1 box=0; Marcus Semien: H current=1 box=2, 2B current=0 box=1, R current=0 box=1 | NYM |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
KC@CLEMISS
Predicted: CLE 4-5 (conf +0.15, moderate)Actual: KC 3-0Brier 0.329total bias +6
Root cause: Run environment far lower than projected (over-dispersed)
Evidence-backed diagnosis- published system edge 0.15 (moderate); this is not the half-scale distance from 50%
- team-run absolute error 6; total bias +6
- pregame home-minus-away evidence: season +0.345, recent -0.178, pitcher +0.079, composite +0.498
- confidence pipeline reached 0.589 before later publication/recency controls; published home probability was 0.574
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs over-projected by 6 (9 vs 3)
- CLE offense over-projected by 5
- committed to CLE at 0.15 confidence and it lost
| Team | Starter | Projected | Actual | ER err |
|---|
| KC | Michael Wacha | 2 ER / 18 outs | 0 ER / 24 outs | -2 |
| CLE | Tanner Bibee | 2 ER / 18 outs | 1 ER / 18 outs | -1 |
Decision internals: run factors home×1.064 / away×0.985; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 0.345 | positive favors home; negative favors away |
| Recent form | -0.178 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.079 | positive favors home; negative favors away |
| Composite | 0.498 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.76-4.98 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 0.574 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | n/a | Capped at max 12 realistic runs |
| simulation_result | 4-5 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.76-4.98 | 4-5 | -0.22 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 4-5 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | 4-5 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | 4-5 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | 4-5 | 0 | Capped at max 12 realistic runs |
| simulation_result | 4-5 | 4-5 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| CLE | Jo Adell | batter / RF | better | predicted_RBI 3.18 |
| KC | Bobby Witt Jr. | batter / SS | worse | predicted_AB -1.78 |
| CLE | Patrick Bailey | batter / C | worse | predicted_BB 1.704 |
| KC | Jac Caglianone | batter / RF | worse | predicted_RBI -1.69 |
| KC | Josh Rojas | batter / 3B | better | predicted_RBI 1.68 |
| KC | Carter Jensen | batter / C | better | predicted_RBI 1.5 |
| CLE | José Ramírez | batter / 3B | worse | predicted_AB -1.4 |
| CLE | Chase DeLauter | batter / DH | better | predicted_SO 1.389 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | CLE pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | CLE |
| info | CURRENT_PITCHER_DUPLICATION | KC pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | KC |
| info | CURRENT_FINALIZED_MATCH_BOX | CLE model-prediction batting totals differ from final simulation box score (AB: current=35, sim=31; H: current=8, sim=7; 2B: current=1, sim=3; RBI: current=4, sim=5; BB: current=3, sim=4; SO: current=6, sim=8). | CLE |
| info | CURRENT_PLAYER_MATCH_BOX | CLE current finalized hitter lines are model predictions and differ from the simulation box score: Chase DeLauter: AB current=4 box=3, 2B current=0 box=1, RBI current=1 box=0, SO current=1 box=2; Nathaniel Lowe: AB current=3 box=4, H current=1 box=0, SO current=1 box=2; Jo Adell: AB current=5 box=4, H current=1 box=2, 2B current=0 box=1, RBI current=1 box=4; José Ramírez: AB current=4 box=3, 2B current=1 box=0, RBI current=1 box=0, BB current=0 box=1; Steven Kwan: AB current=5 box=4, H current=2 box=1, BB current=1 box=0; Travis Bazzana: 2B current=0 box=1, RBI current=1 box=0, BB current=1 box=0; Patrick Bailey: AB current=3 box=2, BB current=0 box=2; Petey Halpin: AB current=3 box=4, SO current=1 box=0 | CLE |
| info | CURRENT_FINALIZED_MATCH_BOX | KC model-prediction batting totals differ from final simulation box score (AB: current=36, sim=34; H: current=9, sim=7; 2B: current=2, sim=3; RBI: current=5, sim=4; SO: current=7, sim=10). | KC |
| info | CURRENT_PLAYER_MATCH_BOX | KC current finalized hitter lines are model predictions and differ from the simulation box score: Carter Jensen: AB current=4 box=5, HR current=0 box=1, RBI current=1 box=2, BB current=1 box=0; Vinnie Pasquantino: 2B current=0 box=1, RBI current=1 box=0, R current=1 box=0, BB current=1 box=0; Nick Loftin: R current=0 box=1, BB current=0 box=1, SO current=1 box=0; Isaac Collins: AB current=4 box=3, H current=1 box=0, BB current=0 box=1; Salvador Perez: AB current=5 box=4, H current=1 box=0, RBI current=1 box=0, SO current=1 box=2; Bobby Witt Jr.: AB current=5 box=3, H current=1 box=0, HR current=1 box=0, RBI current=1 box=0; Jac Caglianone: AB current=5 box=4, 2B current=1 box=0, RBI current=1 box=0, SO current=1 box=2; Josh Rojas: AB current=3 box=4, H current=1 box=2, 2B current=0 box=1, RBI current=0 box=2 | KC |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
LAD@DETHIT
Predicted: LAD 4-1 (conf -0.33, strong)Actual: LAD 2-1Brier 0.112total bias +2
Root cause: Strong correct call
Evidence-backed diagnosis- published system edge 0.33 (strong); this is not the half-scale distance from 50%
- team-run absolute error 2; total bias +2
- pregame home-minus-away evidence: season -0.761, recent -0.736, pitcher -0.034, composite -1.218
- confidence pipeline reached 0.661 before later publication/recency controls; published home probability was 0.335
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
No large single-factor errors.
| Team | Starter | Projected | Actual | ER err |
|---|
| LAD | Tarik Skubal | 1 ER / 18 outs | 1 ER / 18 outs | 0 |
| DET | Drew Anderson | 1 ER / 13 outs | 0 ER / 14 outs | -1 |
Decision internals: run factors home×0.982 / away×1.016; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.761 | positive favors home; negative favors away |
| Recent form | -0.736 | positive favors home; negative favors away |
| Starting-pitcher matchup | -0.034 | positive favors home; negative favors away |
| Composite | -1.218 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 4.17-4.08 | n/a | Base ML historical model prediction |
| variance_applied | 4-4 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-2 | 0.665 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 4-1 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-1 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-1 | n/a | Capped at max 12 realistic runs |
| simulation_result | 4-1 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 4.17-4.08 | 4-4 | -0.089 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-4 | 4-2 | 2 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 4-2 | 4-1 | 1 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-1 | 4-1 | -0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-1 | 4-1 | -0 | Capped at max 12 realistic runs |
| simulation_result | 4-1 | 4-1 | -0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| LAD | Shohei Ohtani | batter / DH | worse | predicted_SO 1.927 |
| DET | Dillon Dingler | batter / C | worse | predicted_AB -1.85 |
| LAD | Mookie Betts | batter / SS | worse | predicted_H 1.76 |
| LAD | Alek Thomas | batter / CF | worse | predicted_H 1.69 |
| LAD | Hunter Feduccia | batter / C | boosted | predicted_AB 1.465 |
| DET | Spencer Torkelson | batter / 1B | worse | predicted_AB -1.42 |
| DET | Hao-Yu Lee | batter / 3B | worse | predicted_SO 1.317 |
| LAD | Freddie Freeman | batter / 1B | better | predicted_SO 1.31 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | DET pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | DET |
| info | CURRENT_PITCHER_DUPLICATION | LAD pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | LAD |
| info | CURRENT_FINALIZED_MATCH_BOX | DET model-prediction batting totals differ from final simulation box score (AB: current=32, sim=30; H: current=8, sim=3; 2B: current=2, sim=1; RBI: current=4, sim=1; R: current=4, sim=1; BB: current=3, sim=2; SO: current=8, sim=4). | DET |
| info | CURRENT_PLAYER_MATCH_BOX | DET current finalized hitter lines are model predictions and differ from the simulation box score: Hao-Yu Lee: AB current=3 box=4, H current=1 box=0, SO current=1 box=2; Eduardo Valencia: AB current=3 box=4, SO current=1 box=0; Ben Malgeri: H current=1 box=0, RBI current=1 box=0, BB current=1 box=0; Dillon Dingler: AB current=5 box=3, H current=1 box=0, RBI current=1 box=0, BB current=0 box=1; Gleyber Torres: 2B current=0 box=1, R current=1 box=0, BB current=1 box=0, SO current=1 box=0; Javier Báez: H current=1 box=0, 2B current=1 box=0, R current=1 box=0, SO current=1 box=0; Spencer Torkelson: AB current=3 box=2, BB current=0 box=1, SO current=1 box=0; Kevin McGonigle: H current=1 box=0, 2B current=1 box=0, R current=1 box=0, BB current=1 box=0 | DET |
| info | CURRENT_FINALIZED_MATCH_BOX | LAD model-prediction batting totals differ from final simulation box score (AB: current=35, sim=37; H: current=8, sim=11; 2B: current=2, sim=4; HR: current=1, sim=0; R: current=5, sim=4; BB: current=3, sim=1; SO: current=8, sim=15; SF: current=0, sim=1). | LAD |
| info | CURRENT_PLAYER_MATCH_BOX | LAD current finalized hitter lines are model predictions and differ from the simulation box score: Shohei Ohtani: AB current=4 box=5, HR current=1 box=0, BB current=1 box=0, SO current=1 box=3; Mookie Betts: H current=1 box=3, RBI current=1 box=0, SO current=0 box=1; Freddie Freeman: 2B current=1 box=0, RBI current=0 box=1, BB current=1 box=0, SO current=1 box=2; Max Muncy: 2B current=0 box=1, RBI current=1 box=0, R current=1 box=0, BB current=0 box=1; Kyle Tucker: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=2; Tommy Edman: AB current=4 box=3, SF current=0 box=1; Alek Thomas: AB current=3 box=4, H current=0 box=2, 2B current=0 box=1, SO current=1 box=2; Teoscar Hernández: SO current=1 box=2 | LAD |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
MIA@WASMISS
Predicted: MIA 5-4 (conf -0.11, moderate)Actual: WAS 2-9Brier 0.308total bias -2
Root cause: Starting-pitcher projection was wrong
Evidence-backed diagnosis- published system edge 0.11 (moderate); this is not the half-scale distance from 50%
- team-run absolute error 8; total bias -2
- pregame home-minus-away evidence: season -0.201, recent -0.702, pitcher -0.018, composite -0.505
- confidence pipeline reached 0.566 before later publication/recency controls; published home probability was 0.445
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- MIA offense over-projected by 3
- WAS offense under-projected by 5
- Eury Pérez (MIA) allowed 4 more ER than projected (5 vs 1)
- committed to MIA at 0.11 confidence and it lost
| Team | Starter | Projected | Actual | ER err |
|---|
| MIA | Eury Pérez | 1 ER / 17 outs | 5 ER / 9 outs | 4 |
| WAS | Jackson Kent | 3 ER / 11 outs | 1 ER / 21 outs | -2 |
Decision internals: run factors home×0.975 / away×0.993; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.201 | positive favors home; negative favors away |
| Recent form | -0.702 | positive favors home; negative favors away |
| Starting-pitcher matchup | -0.018 | positive favors home; negative favors away |
| Composite | -0.505 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 6.17-3.08 | n/a | Base ML historical model prediction |
| variance_applied | 6-3 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-4 | 0.555 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-4 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-4 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-4 | n/a | Capped at max 12 realistic runs |
| simulation_result | 5-4 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 6.17-3.08 | 6-3 | -0.089 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 6-3 | 5-4 | -2 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-4 | 5-4 | -0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-4 | 5-4 | -0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-4 | 5-4 | -0 | Capped at max 12 realistic runs |
| simulation_result | 5-4 | 5-4 | -0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| WAS | CJ Abrams | batter / SS | better | predicted_RBI 3.37 |
| MIA | Eury Pérez | pitcher / Starter | neutral | predicted_SO 2 |
| MIA | Otto Lopez | batter / SS | worse | predicted_AB -1.95 |
| WAS | Andrés Chaparro | batter / DH | better | predicted_AB 1.5 |
| WAS | Jackson Kent | pitcher / Starter | worse | predicted_IP -1.467 |
| WAS | Jacob Young | batter / CF | worse | predicted_H 1.24 |
| WAS | Dylan Crews | batter / RF | worse | predicted_H -1.09 |
| WAS | Brady House | batter / 3B | worse | predicted_SO -1.086 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | MIA pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | MIA |
| info | CURRENT_PITCHER_DUPLICATION | WAS pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | WAS |
| info | CURRENT_FINALIZED_MATCH_BOX | MIA model-prediction batting totals differ from final simulation box score (H: current=8, sim=6; 3B: current=0, sim=2; HR: current=1, sim=0; RBI: current=4, sim=5; BB: current=3, sim=0; SO: current=8, sim=5; SF: current=0, sim=1). | MIA |
| info | CURRENT_PLAYER_MATCH_BOX | MIA current finalized hitter lines are model predictions and differ from the simulation box score: Agustín Ramírez: AB current=3 box=4, 3B current=0 box=1, RBI current=0 box=1, BB current=1 box=0; Javier Sanoja: AB current=3 box=4, 2B current=0 box=1, RBI current=1 box=0; Otto Lopez: AB current=5 box=3, H current=1 box=0, 2B current=1 box=0, SO current=1 box=0; Xavier Edwards: 3B current=0 box=1, RBI current=0 box=1, BB current=1 box=0, SO current=1 box=0; Esteury Ruiz: SO current=1 box=0; Kyle Stowers: BB current=1 box=0; Heriberto Hernández: HR current=1 box=0; Joe Mack: H current=1 box=0, SO current=1 box=0 | MIA |
| info | CURRENT_FINALIZED_MATCH_BOX | WAS model-prediction batting totals differ from final simulation box score (AB: current=32, sim=34; H: current=8, sim=7; 2B: current=2, sim=0; 3B: current=0, sim=1; RBI: current=5, sim=4; SO: current=8, sim=6). | WAS |
| info | CURRENT_PLAYER_MATCH_BOX | WAS current finalized hitter lines are model predictions and differ from the simulation box score: Abimelec Ortiz: AB current=3 box=4, H current=1 box=0, SO current=1 box=0; Daylen Lile: H current=1 box=2, HR current=1 box=0, RBI current=1 box=0; CJ Abrams: AB current=4 box=5, HR current=0 box=1, RBI current=1 box=4, SO current=1 box=3; Jorbit Vivas: H current=1 box=0; Andrés Chaparro: AB current=3 box=4, H current=1 box=2, 2B current=1 box=0, RBI current=1 box=0; Dylan Crews: AB current=5 box=4, H current=1 box=0, RBI current=1 box=0, R current=1 box=0; Brady House: AB current=4 box=3, H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0; Jacob Young: AB current=3 box=4, H current=1 box=2, 3B current=0 box=1, R current=0 box=1 | WAS |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
PHI@LAAMISS
Predicted: LAA 4-5 (conf +0.13, moderate)Actual: PHI 5-3Brier 0.32total bias +1
Root cause: Backed the wrong side with real conviction
Evidence-backed diagnosis- published system edge 0.13 (moderate); this is not the half-scale distance from 50%
- team-run absolute error 3; total bias +1
- pregame home-minus-away evidence: season -0.866, recent -0.660, pitcher +0.104, composite -0.232
- confidence pipeline reached 0.570 before later publication/recency controls; published home probability was 0.566
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- committed to LAA at 0.13 confidence and it lost
| Team | Starter | Projected | Actual | ER err |
|---|
| PHI | Andrew Painter | 3 ER / 15 outs | 3 ER / 18 outs | 0 |
| LAA | Reid Detmers | 3 ER / 17 outs | 1 ER / 18 outs | -2 |
Decision internals: run factors home×1.065 / away×0.961; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.866 | positive favors home; negative favors away |
| Recent form | -0.66 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.104 | positive favors home; negative favors away |
| Composite | -0.232 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.85-5.27 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 0.566 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | n/a | Capped at max 12 realistic runs |
| simulation_result | 4-5 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.85-5.27 | 4-5 | -0.412 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 4-5 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | 4-5 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | 4-5 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | 4-5 | 0 | Capped at max 12 realistic runs |
| simulation_result | 4-5 | 4-5 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| PHI | J.T. Realmuto | batter / C | better | predicted_RBI 2.52 |
| PHI | Kyle Schwarber | batter / DH | worse | predicted_SO 2.484 |
| LAA | Denzer Guzman | batter / 3B | better | predicted_H 2.36 |
| LAA | Tyler Heineman | batter / C | better | predicted_AB 2.352 |
| LAA | Vaughn Grissom | batter / 1B | worse | predicted_SO 2.273 |
| PHI | José Alvarado | pitcher / Reliever | boosted | predicted_IP 2.1 |
| LAA | Josh Lowe | batter / RF | better | predicted_AB 2.035 |
| LAA | Wade Meckler | batter / LF | worse | predicted_AB 1.846 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | LAA pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | LAA |
| info | CURRENT_PITCHER_DUPLICATION | PHI pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | PHI |
| info | CURRENT_FINALIZED_MATCH_BOX | LAA model-prediction batting totals differ from final simulation box score (AB: current=32, sim=43; H: current=7, sim=10; 2B: current=1, sim=3; RBI: current=4, sim=5; R: current=3, sim=5; BB: current=3, sim=1; SO: current=9, sim=13; HBP: current=0, sim=1). | LAA |
| info | CURRENT_PLAYER_MATCH_BOX | LAA current finalized hitter lines are model predictions and differ from the simulation box score: Vaughn Grissom: AB current=4 box=5, H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0; Wade Meckler: AB current=3 box=5, SO current=1 box=2; Mike Trout: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=2; Josh Lowe: AB current=3 box=5, RBI current=0 box=1, R current=0 box=1, SO current=1 box=0; Moisés Ballesteros: AB current=4 box=5, H current=1 box=0, RBI current=1 box=0, SO current=1 box=2; Christian Moore: AB current=3 box=4, H current=0 box=1, R current=0 box=1, SO current=1 box=3; Zach Neto: AB current=4 box=5, H current=1 box=2, 2B current=0 box=1, HR current=1 box=0; Tyler Heineman: AB current=3 box=5, H current=0 box=2, RBI current=0 box=1, R current=0 box=1 | LAA |
| info | CURRENT_FINALIZED_MATCH_BOX | PHI model-prediction batting totals differ from final simulation box score (AB: current=35, sim=39; H: current=9, sim=6; 2B: current=2, sim=0; HR: current=1, sim=2; RBI: current=5, sim=4; R: current=5, sim=4; SO: current=8, sim=13). | PHI |
| info | CURRENT_PLAYER_MATCH_BOX | PHI current finalized hitter lines are model predictions and differ from the simulation box score: Bryce Harper: AB current=4 box=5, 2B current=1 box=0, RBI current=1 box=0, BB current=1 box=0; Alec Bohm: AB current=4 box=5, H current=1 box=0, RBI current=1 box=0, SO current=1 box=2; Bryan De La Cruz: AB current=3 box=4, H current=1 box=0, RBI current=1 box=0, BB current=0 box=1; Bryson Stott: AB current=3 box=4, R current=1 box=0; Kyle Schwarber: H current=1 box=0, HR current=1 box=0, RBI current=1 box=0, R current=1 box=0; Luis Arraez: AB current=5 box=4, H current=1 box=2, 2B current=1 box=0, BB current=0 box=1; J.T. Realmuto: HR current=0 box=1, RBI current=1 box=3, R current=0 box=1, BB current=0 box=1; Trea Turner: HR current=0 box=1, RBI current=0 box=1, BB current=1 box=0, SO current=1 box=0 | PHI |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
PIT@STLMISS
Predicted: PIT 5-3 (conf -0.20, strong)Actual: STL 1-4Brier 0.36total bias +3
Root cause: Backed the wrong side with real conviction
Evidence-backed diagnosis- published system edge 0.20 (strong); this is not the half-scale distance from 50%
- team-run absolute error 5; total bias +3
- pregame home-minus-away evidence: season -0.292, recent -0.298, pitcher -0.091, composite -0.795
- confidence pipeline reached 0.655 before later publication/recency controls; published home probability was 0.400
- score-volatility actions: supported_blowout_margin_expansion
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs over-projected by 3 (8 vs 5)
- PIT offense over-projected by 4
- committed to PIT at 0.20 confidence and it lost
| Team | Starter | Projected | Actual | ER err |
|---|
| PIT | Jared Jones | 2 ER / 15 outs | 3 ER / 14 outs | 1 |
| STL | Quinn Mathews | 2 ER / 13 outs | 1 ER / 17 outs | -1 |
Decision internals: run factors home×0.958 / away×1.049; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.292 | positive favors home; negative favors away |
| Recent form | -0.298 | positive favors home; negative favors away |
| Starting-pitcher matchup | -0.091 | positive favors home; negative favors away |
| Composite | -0.795 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 4.68-3.39 | n/a | Base ML historical model prediction |
| variance_applied | 5-3 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-3 | 0.6 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-3 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-3 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-3 | n/a | Capped at max 12 realistic runs |
| simulation_result | 5-3 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 4.68-3.39 | 5-3 | 0.718 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 5-3 | 5-3 | -0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 5-3 | 5-3 | -0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 5-3 | 5-3 | -0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 5-3 | 5-3 | -0 | Capped at max 12 realistic runs |
| simulation_result | 5-3 | 5-3 | -0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| PIT | Oneil Cruz | batter / CF | worse | predicted_AB -2.63 |
| PIT | Esmerlyn Valdez | batter / RF | better | predicted_RBI 2.433 |
| PIT | Jared Triolo | batter / SS | worse | predicted_AB 1.955 |
| STL | JJ Wetherholt | batter / SS | worse | predicted_SO 1.266 |
| PIT | Rafael Flores Jr. | batter / 1B | neutral | predicted_H 1.227 |
| STL | Nolan Gorman | batter / 3B | worse | predicted_SO -1.203 |
| STL | Alec Burleson | batter / 1B | worse | predicted_RBI 1.16 |
| PIT | Nick Gonzales | batter / 2B | worse | predicted_H -1.148 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | PIT pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | PIT |
| info | CURRENT_PITCHER_DUPLICATION | STL pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | STL |
| info | CURRENT_FINALIZED_MATCH_BOX | PIT model-prediction batting totals differ from final simulation box score (AB: current=32, sim=35; H: current=8, sim=9; 2B: current=1, sim=2; HR: current=1, sim=2; RBI: current=4, sim=5; R: current=4, sim=5; SO: current=9, sim=10; SF: current=0, sim=1). | PIT |
| info | CURRENT_PLAYER_MATCH_BOX | PIT current finalized hitter lines are model predictions and differ from the simulation box score: Esmerlyn Valdez: AB current=3 box=5, H current=1 box=2, 2B current=0 box=1, HR current=0 box=1; Bryan Reynolds: 2B current=1 box=0, RBI current=1 box=0; Nick Gonzales: H current=1 box=0, R current=1 box=0; Rafael Flores Jr.: AB current=3 box=4, H current=1 box=2, HR current=0 box=1, R current=0 box=1; Jake Mangum: H current=1 box=2, R current=0 box=1, SO current=1 box=0; Oneil Cruz: AB current=5 box=2, H current=1 box=0, HR current=1 box=0, RBI current=1 box=0; Nick Yorke: AB current=3 box=4, 2B current=0 box=1; Henry Davis: H current=0 box=1, RBI current=0 box=1, R current=0 box=1, SF current=0 box=1 | PIT |
| info | CURRENT_FINALIZED_MATCH_BOX | STL model-prediction batting totals differ from final simulation box score (AB: current=33, sim=30; H: current=8, sim=3; HR: current=1, sim=2; RBI: current=4, sim=3; R: current=5, sim=3; BB: current=3, sim=0; SO: current=8, sim=7; HBP: current=0, sim=1). | STL |
| info | CURRENT_PLAYER_MATCH_BOX | STL current finalized hitter lines are model predictions and differ from the simulation box score: JJ Wetherholt: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=2; José Fermín: H current=1 box=0; Iván Herrera: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=0; Joshua Baez: H current=1 box=0, HR current=1 box=0, RBI current=1 box=0, R current=1 box=0; Jordan Walker: AB current=5 box=4, 2B current=1 box=0, HR current=0 box=1; Alec Burleson: HR current=0 box=1, RBI current=1 box=2; Nolan Gorman: H current=0 box=1, 2B current=0 box=1, R current=0 box=1, BB current=1 box=0; Nathan Church: AB current=4 box=3, H current=1 box=0, RBI current=1 box=0 | STL |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
SD@TBHIT
Predicted: TB 4-5 (conf +0.11, moderate)Actual: TB 4-9Brier 0.196total bias -4
Root cause: Solid correct call
Evidence-backed diagnosis- published system edge 0.11 (moderate); this is not the half-scale distance from 50%
- team-run absolute error 4; total bias -4
- pregame home-minus-away evidence: season +0.062, recent -0.400, pitcher +0.060, composite +0.246
- confidence pipeline reached 0.549 before later publication/recency controls; published home probability was 0.557
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs under-projected by 4 (9 vs 13)
- TB offense under-projected by 4
| Team | Starter | Projected | Actual | ER err |
|---|
| SD | Casey Mize | 2 ER / 15 outs | 2 ER / 15 outs | 0 |
| TB | Shane McClanahan | 2 ER / 16 outs | 3 ER / 17 outs | 1 |
Decision internals: run factors home×1.053 / away×0.993; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 0.062 | positive favors home; negative favors away |
| Recent form | -0.4 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.06 | positive favors home; negative favors away |
| Composite | 0.246 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.88-4.61 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 0.557 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | n/a | Capped at max 12 realistic runs |
| simulation_result | 4-5 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.88-4.61 | 4-5 | 0.262 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 4-5 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run) |
| embedded_savant_run_environment | 4-5 | 4-5 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 4-5 | 4-5 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 4-5 | 4-5 | 0 | Capped at max 12 realistic runs |
| simulation_result | 4-5 | 4-5 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| TB | Jorge Mateo | batter / 2B | worse | predicted_2B 1.877 |
| TB | Yandy Díaz | batter / DH | worse | predicted_RBI 1.49 |
| SD | Manny Machado | batter / 3B | worse | predicted_AB -1.41 |
| TB | Junior Caminero | batter / 3B | boosted | predicted_RBI 1.34 |
| SD | Fernando Tatis Jr. | batter / RF | better | predicted_RBI 1.26 |
| TB | Cedric Mullins | batter / CF | worse | predicted_SO 1.223 |
| SD | Ty France | batter / 1B | worse | predicted_H 1.16 |
| TB | Liam Hicks | batter / C | worse | predicted_H -1.14 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | SD pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | SD |
| info | CURRENT_PITCHER_DUPLICATION | TB pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | TB |
| info | CURRENT_FINALIZED_MATCH_BOX | SD model-prediction batting totals differ from final simulation box score (AB: current=34, sim=35; 2B: current=1, sim=2; BB: current=3, sim=1). | SD |
| info | CURRENT_PLAYER_MATCH_BOX | SD current finalized hitter lines are model predictions and differ from the simulation box score: Manny Machado: AB current=4 box=3, H current=1 box=0, R current=1 box=0; Jackson Merrill: AB current=5 box=4, 2B current=0 box=1; Luis Rengifo: R current=0 box=1; Austin Hays: AB current=3 box=4; Xander Bogaerts: AB current=3 box=4, BB current=1 box=0; Freddy Fermin: AB current=3 box=4, H current=1 box=0; Jake Cronenworth: 2B current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=0; Ty France: AB current=3 box=4, H current=1 box=2, 2B current=0 box=1, RBI current=1 box=0 | SD |
| info | CURRENT_FINALIZED_MATCH_BOX | TB model-prediction batting totals differ from final simulation box score (AB: current=35, sim=34; H: current=9, sim=10; 2B: current=1, sim=5; 3B: current=0, sim=1; RBI: current=4, sim=5; R: current=4, sim=5; BB: current=3, sim=2; SO: current=6, sim=8). | TB |
| info | CURRENT_PLAYER_MATCH_BOX | TB current finalized hitter lines are model predictions and differ from the simulation box score: Jonathan Aranda: AB current=4 box=3, H current=1 box=0, RBI current=1 box=0, BB current=0 box=1; Liam Hicks: H current=1 box=0, RBI current=1 box=0, R current=1 box=0, BB current=1 box=0; Yandy Díaz: AB current=5 box=4, H current=1 box=2, 2B current=0 box=1, HR current=1 box=0; Junior Caminero: H current=1 box=2, 2B current=0 box=1, HR current=0 box=1, RBI current=1 box=2; Jorge Mateo: AB current=3 box=4, H current=1 box=2, 2B current=0 box=2; Chandler Simpson: AB current=4 box=3, R current=0 box=1, BB current=0 box=1, SO current=0 box=1; Jonny DeLuca: SO current=1 box=0; Cedric Mullins: BB current=1 box=0, SO current=1 box=2 | TB |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
SEA@TORHIT
Predicted: TOR 2-4 (conf +0.06, lean)Actual: TOR 1-8Brier 0.221total bias -3
Root cause: Right winner, wrong score
Evidence-backed diagnosis- published system edge 0.06 (lean); this is not the half-scale distance from 50%
- team-run absolute error 5; total bias -3
- pregame home-minus-away evidence: season -0.002, recent -0.029, pitcher +0.003, composite +0.359
- confidence pipeline reached 0.580 before later publication/recency controls; published home probability was 0.530
- score-volatility actions: supported_blowout_margin_expansion
- structural audit exposed 9 non-pass finding(s); see the per-game audit table
Quantified contributing factors
- total runs under-projected by 3 (6 vs 9)
- TOR offense under-projected by 4
- Emerson Hancock (SEA) allowed 4 more ER than projected (5 vs 1)
| Team | Starter | Projected | Actual | ER err |
|---|
| SEA | Emerson Hancock | 1 ER / 17 outs | 5 ER / 15 outs | 4 |
| TOR | Dylan Cease | 1 ER / 18 outs | 1 ER / 18 outs | 0 |
Decision internals: run factors home×0.928 / away×0.924; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | -0.002 | positive favors home; negative favors away |
| Recent form | -0.029 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.003 | positive favors home; negative favors away |
| Composite | 0.359 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.72-5.02 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 3-4 | 0.53 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 2-4 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 2-4 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 2-4 | n/a | Capped at max 12 realistic runs |
| simulation_result | 2-4 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.72-5.02 | 4-5 | -0.298 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 3-4 | 0 | Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by matchup_recency_guardrail) |
| embedded_savant_run_environment | 3-4 | 2-4 | 1 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 2-4 | 2-4 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 2-4 | 2-4 | 0 | Capped at max 12 realistic runs |
| simulation_result | 2-4 | 2-4 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| TOR | Dylan Cease | pitcher / Starter | better | predicted_SO 2 |
| TOR | Ernie Clement | batter / 2B | better | predicted_RBI 1.71 |
| TOR | Brett Bateman | batter / CF | worse | predicted_H 1.55 |
| TOR | Vladimir Guerrero Jr. | batter / 1B | worse | predicted_SO 1.445 |
| SEA | Josh Naylor | batter / 1B | worse | predicted_AB -1.42 |
| SEA | Cal Raleigh | batter / C | better | predicted_SO -1.323 |
| TOR | Andrés Giménez | batter / SS | worse | predicted_SO 1.319 |
| SEA | Julio Rodríguez | batter / CF | worse | predicted_AB -1.26 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| error | BATTING_REALISM | simulation:SEA HR=3 exceeds RBI=2; every HR creates at least one RBI unless special scoring data is absent. | SEA |
| info | CURRENT_PITCHER_DUPLICATION | SEA pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | SEA |
| info | CURRENT_PITCHER_DUPLICATION | TOR pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | TOR |
| info | CURRENT_FINALIZED_MATCH_BOX | SEA model-prediction batting totals differ from final simulation box score (AB: current=37, sim=33; H: current=8, sim=6; 2B: current=1, sim=0; HR: current=1, sim=3; RBI: current=4, sim=2; R: current=5, sim=2; BB: current=4, sim=1). | SEA |
| info | CURRENT_PLAYER_MATCH_BOX | SEA current finalized hitter lines are model predictions and differ from the simulation box score: Randy Arozarena: H current=1 box=0, RBI current=1 box=0, R current=1 box=0, BB current=1 box=0; Dominic Canzone: 2B current=1 box=0, RBI current=1 box=0, SO current=1 box=0; Julio Rodríguez: AB current=5 box=4, H current=1 box=0, RBI current=1 box=0, R current=1 box=0; Josh Naylor: AB current=4 box=3, H current=1 box=0, BB current=0 box=1, SO current=1 box=0; Taylor Ward: HR current=0 box=1, RBI current=0 box=1, BB current=1 box=0; J.P. Crawford: AB current=4 box=3, BB current=1 box=0, SO current=1 box=2; Brock Rodden: SO current=1 box=2; Cal Raleigh: RBI current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=1 box=0 | SEA |
| info | CURRENT_FINALIZED_MATCH_BOX | TOR model pitcher projections differ from final simulation pitching box score (HR: current=2, sim=3). | TOR |
| info | CURRENT_FINALIZED_MATCH_BOX | TOR model-prediction batting totals differ from final simulation box score (AB: current=34, sim=32; H: current=8, sim=7; 2B: current=2, sim=3; RBI: current=3, sim=4; R: current=3, sim=4; BB: current=2, sim=1; SO: current=7, sim=9). | TOR |
| info | CURRENT_PLAYER_MATCH_BOX | TOR current finalized hitter lines are model predictions and differ from the simulation box score: Brett Bateman: AB current=5 box=4, H current=1 box=3, RBI current=0 box=1, SO current=1 box=0; Jesús Sánchez: H current=1 box=0, 2B current=1 box=0, BB current=0 box=1, SO current=1 box=0; Vladimir Guerrero Jr.: H current=1 box=0, R current=1 box=0, BB current=1 box=0, SO current=0 box=2; Nathan Lukes: H current=1 box=0, RBI current=1 box=0; George Springer: H current=1 box=0, RBI current=1 box=0; Kazuma Okamoto: H current=1 box=2, 2B current=0 box=1, BB current=1 box=0; Ernie Clement: AB current=4 box=3, RBI current=0 box=2, R current=0 box=1; Brandon Valenzuela: H current=0 box=1, 2B current=0 box=1, R current=0 box=1, SO current=1 box=2 | TOR |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
TEX@MILHIT
Predicted: MIL 2-6 (conf +0.70, strong)Actual: MIL 1-6Brier 0.023total bias +1
Root cause: Strong correct call
Evidence-backed diagnosis- published system edge 0.70 (strong); this is not the half-scale distance from 50%
- team-run absolute error 1; total bias +1
- pregame home-minus-away evidence: season +1.485, recent +0.840, pitcher +0.047, composite +1.851
- confidence pipeline reached 0.855 before later publication/recency controls; published home probability was 0.848
- score-volatility actions: no_change
- structural audit exposed 7 non-pass finding(s); see the per-game audit table
Quantified contributing factors
No large single-factor errors.
| Team | Starter | Projected | Actual | ER err |
|---|
| TEX | Cody Bradford | 2 ER / 15 outs | 3 ER / 17 outs | 1 |
| MIL | Logan Henderson | 2 ER / 18 outs | 1 ER / 21 outs | -1 |
Decision internals: run factors home×1.001 / away×0.954; savant lineup adjustment applied
Full model trace, player evidence, and audit findings
Fundamental evidence entering the pick
| Signal | Home-away edge | Meaning |
|---|
| Season quality | 1.485 | positive favors home; negative favors away |
| Recent form | 0.84 | positive favors home; negative favors away |
| Starting-pitcher matchup | 0.047 | positive favors home; negative favors away |
| Composite | 1.851 | positive favors home; negative favors away |
Score and confidence evolution
| Stage | Away-home runs | Confidence | What changed |
|---|
| historical_ml | 3.71-5.18 | n/a | Base ML historical model prediction |
| variance_applied | 4-5 | n/a | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 2-6 | 0.848 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 2-6 | n/a | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 2-6 | n/a | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 2-6 | n/a | Capped at max 12 realistic runs |
| simulation_result | 2-6 | n/a | Final inning-by-inning simulation result |
Official postgame stage attribution
| Stage | Before | After | Toward final pick | What changed |
|---|
| variance_applied | 3.71-5.18 | 4-5 | -0.466 | Deterministic baseline target (variance factor 1.0) |
| enhanced_prediction | 4-5 | 2-6 | 3 | Enhanced ML with recent form, H2H, low-score analysis |
| embedded_savant_run_environment | 2-6 | 2-6 | 0 | Target score adjusted by temporal Savant lineup quality versus opposing starter quality |
| fundamental_edge_target_review | 2-6 | 2-6 | 0 | Refreshed score target and win probability after Savant/pitcher/team-strength adjustments |
| capped_target | 2-6 | 2-6 | 0 | Capped at max 12 realistic runs |
| simulation_result | 2-6 | 2-6 | 0 | Final inning-by-inning simulation result |
Largest player-level adjustments from the ultra-detailed trace
These are pregame model adjustments, not claims about actual postgame player error.
| Team | Player | Role | Net effect | Largest change |
|---|
| MIL | David Hamilton | batter / 3B | better | predicted_RBI 3.63 |
| MIL | Logan Henderson | pitcher / Starter | better | predicted_SO 2 |
| TEX | Wyatt Langford | batter / LF | worse | predicted_AB -1.866 |
| TEX | Brandon Nimmo | batter / RF | worse | predicted_AB -1.76 |
| MIL | Christian Yelich | batter / DH | worse | predicted_AB -1.73 |
| TEX | Corey Seager | batter / SS | worse | predicted_AB -1.26 |
| MIL | Jackson Chourio | batter / LF | worse | predicted_AB -1.2 |
| MIL | Garrett Mitchell | batter / CF | worse | predicted_H -1.19 |
Structural audit findings
| Severity | Rule | Finding | Subject |
|---|
| info | CURRENT_PITCHER_DUPLICATION | MIL pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | MIL |
| info | CURRENT_PITCHER_DUPLICATION | TEX pitcher list appears in both predicted_performance.pitchers and player_predictions_by_position.pitchers; IDs match and were de-duplicated in this audit. | TEX |
| info | CURRENT_FINALIZED_MATCH_BOX | MIL model-prediction batting totals differ from final simulation box score (AB: current=36, sim=31; H: current=9, sim=7; 2B: current=2, sim=1; HR: current=1, sim=4; RBI: current=5, sim=6; R: current=5, sim=6; BB: current=4, sim=5; SO: current=8, sim=7). | MIL |
| info | CURRENT_PLAYER_MATCH_BOX | MIL current finalized hitter lines are model predictions and differ from the simulation box score: Jackson Chourio: AB current=5 box=4, 2B current=0 box=1, HR current=1 box=0, RBI current=1 box=0; Brice Turang: AB current=5 box=4, HR current=0 box=1; William Contreras: AB current=5 box=4, HR current=0 box=1, BB current=1 box=0; Joey Ortiz: AB current=3 box=2, H current=1 box=0, BB current=0 box=1, SO current=1 box=0; Christian Yelich: AB current=5 box=3, 2B current=1 box=0, RBI current=1 box=0; Garrett Mitchell: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0; Sal Frelick: AB current=3 box=4, H current=1 box=0, SO current=0 box=1; David Hamilton: H current=1 box=2, HR current=0 box=2, RBI current=0 box=4, R current=0 box=1 | MIL |
| info | CURRENT_FINALIZED_MATCH_BOX | TEX model-prediction batting totals differ from final simulation box score (AB: current=33, sim=30; H: current=7, sim=3; 2B: current=2, sim=1; RBI: current=4, sim=2; R: current=5, sim=2; SO: current=9, sim=6). | TEX |
| info | CURRENT_PLAYER_MATCH_BOX | TEX current finalized hitter lines are model predictions and differ from the simulation box score: Wyatt Langford: AB current=4 box=2, H current=1 box=0, R current=1 box=0, BB current=1 box=2; Ezequiel Duran: H current=1 box=0, 2B current=1 box=0, RBI current=1 box=0, R current=1 box=0; Corey Seager: AB current=4 box=3, H current=1 box=0, RBI current=1 box=0, R current=1 box=0; Brandon Nimmo: AB current=5 box=3, H current=1 box=0, 2B current=1 box=0, BB current=0 box=1; Joc Pederson: AB current=3 box=4, H current=0 box=1, SO current=1 box=0; Cody Freeman: HR current=0 box=1, RBI current=0 box=1, R current=0 box=1, SO current=1 box=0; Evan Carter: AB current=3 box=4, 2B current=0 box=1, SO current=1 box=0; Danny Jansen: SO current=1 box=2 | TEX |
| info | RECONCILIATION_PROPAGATION | No score_target_reconciliation step was used for this game. | |
Cross-game lever tension
Each candidate fix, and exactly which games it would help versus 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 · 5 helped vs 3 hurt
Would help- CIN@CHC (miss): lost a 0.12-confidence pick; wider tails lower overconfidence
- KC@CLE (miss): lost a 0.15-confidence pick; wider tails lower overconfidence
- MIA@WAS (miss): lost a 0.11-confidence pick; wider tails lower overconfidence
- PHI@LAA (miss): lost a 0.13-confidence pick; wider tails lower overconfidence
- PIT@STL (miss): lost a 0.20-confidence pick; wider tails lower overconfidence
Would hurt / risk- COL@ATL (hit): strong correct call at 0.70; widening softens a good edge
- LAD@DET (hit): strong correct call at 0.33; widening softens a good edge
- TEX@MIL (hit): strong correct call at 0.70; widening softens a good edge
player.pitcher.v4 form weights / player.pitcher_prior_games
Targets: starter_er_mae, starter_outs_mae · 5 helped vs 4 hurt
Would help- ARI@SF (miss): starter ER off by 5
- BAL@OAK (hit): starter ER off by 5
- BOS@NYY (hit): starter ER off by 3
- MIA@WAS (miss): starter ER off by 4
- SEA@TOR (hit): starter ER off by 4
Would hurt / risk- LAD@DET (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
- PIT@STL (miss): starter projection was accurate (ER off 1); reweight risks disturbing it
- SD@TB (hit): starter projection was accurate (ER off 1); reweight risks disturbing it
- TEX@MIL (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 · 10 helped vs 2 hurt
Would help- ARI@SF (miss): team runs off by 5
- BAL@OAK (hit): team runs off by 4
- BOS@NYY (hit): team runs off by 9
- CHW@MIN (miss): team runs off by 4
- CIN@CHC (miss): team runs off by 6
- KC@CLE (miss): team runs off by 6
- MIA@WAS (miss): team runs off by 8
- PIT@STL (miss): team runs off by 5
- SD@TB (hit): team runs off by 4
- SEA@TOR (hit): team runs off by 5
Would hurt / risk- LAD@DET (hit): team runs already close (off 2); reweight risks disturbing it
- TEX@MIL (hit): team runs already close (off 1); reweight risks disturbing it
Day over day: did it improve or worsen
Compared against packet 20260827. Last recorded change: v4.8.2 (2026-08-28) Enforce measurable deep-report coverage and publish adaptive status through Nginx.
| Metric | Prior | Current | Δ | Result |
|---|
| brier_score | 0.2422 | 0.2418 | -0.000404 | improved |
| accuracy | 0.598 | 0.5833 | -0.0147 | regressed |
| team_run_mae | 2.4363 | 2.4306 | -0.0057 | improved |
| total_run_mae | 3.7745 | 3.713 | -0.0615 | improved |
| calibration_error | 0.108 | 0.1004 | -0.0076 | improved |
| starter_er_mae | 1.8154 | 1.7854 | -0.03 | improved |
| starter_outs_mae | 3.2154 | 3.1073 | -0.1081 | improved |
The adjustment the system made in response
Status: applied to production · APPLIED TO PRODUCTION. The champion changed from default-v1 to candidate-31a39524087cf556 at 2026-08-29T06:02:24.658142+00:00. These values are now live. The previous champion is retained and the rollback monitor is armed: the next settled slates measure whether this improved or worsened predictions, and it reverts automatically on a measured regression.
Hypothesis: Reducing confidence.moderate_cap from 0.6 to 0.58 may improve overall calibration, as suggested by the errors in observed hit rates in moderate confidence buckets.
| Parameter | Change | Expected effect | Risk | Evidence |
|---|
| confidence.moderate_cap | 0.6 → 0.58 | calibration_error decrease | low risk | calibration.bucket.0.55-0.60, calibration.bucket.0.60-0.65, calibration.bucket.0.65-0.70 |
Luna deep post-mortem (full)