MLB Prediction System

Metric Evolution & Audit

Every number that moved, the stage that moved it, the input it used, and the arithmetic that proves it

Slate 20261006Games 2Checks 314Schema metric_evolution_report_v1Built 2026-10-07 09:14Z

Verdict across the slate

314 invariants checked across 2 games: 2 OF 314 CHECKS FAILED

away staff totals a regulation workload: 30 outs recorded, 27 expected
home staff totals a regulation workload: 30 outs recorded, 27 expected

How to read this, and what a defect looks like

Every figure below is shown with the arithmetic that produced it and a verdict beside it, so a disagreement can be pointed at a specific line rather than at the system as a whole.

Stage chain. Each game lists every stage that touched the score, the score after it, and whether it changed the target. Most stages move nothing on most games; a chain that showed only movement would misrepresent it. A stage reporting a delta while changed_target is false computed an adjustment and did not apply it, which is normal and is stated.
Blend arithmetic. The KPI blend must satisfy blended = baseline + weight x (kpi_only - baseline) for all three targets. A FAIL here means the layer reported a number it did not compute.
Rounding. Published integers must be the rounding of the blended runs under the documented tie rule. A FAIL means the published score and the continuous line describe different games.
Log5 additivity. For every player metric, own skill plus opponent plus the Log5 remainder must equal the total against the league, exactly. A FAIL means the decomposition is not a decomposition.
Direction. A green delta helps the subject, a red one hurts him. Strikeouts invert for a hitter. If an elite arm's opponent shows green strikeouts, the orientation is wrong and the whole player section is suspect.
Measured versus published. The KPI layer moves team runs and the win probability. The per-player metric adjustments here are a measurement of the same catalog at player level, not what the pipeline published; the one club-wide ratio the pipeline actually applied to hitter lines is printed beside them.
Integer versus continuous. The integer player line is a simulation draw reconciled to the box score. The continuous expected_* beside it is the model's own view and is the rankable quantity. Judging the model by the integer is how a working model looks broken.

Slate overview

Game	Published	p(home)	Stages	Stages that moved it	KPI applied	Δ away runs	Δ home runs	Simulation draw chosen on	0-out starter
20261006_LAD@ATL	4 - 3	0.4948	8	2	yes	-0.5364	+0.2218	-	no
20261006_MIL@SD	4 - 5	0.5664	8	1	yes	-0.3389	-0.3442	-	no

KPI model in force. candidate 20261007T091423Z_47b9db81, 48 features built from 40 registered KPIs, blend weights {"home_runs": 0.38, "away_runs": 0.34, "home_win": 0.0}, frozen evaluation ends 2026-09-27. A target whose weight is zero contributes nothing by construction: the winner probability carries weight zero while its walk-forward verdict is inconclusive, so every win-probability movement below is expected to be exactly zero.

Per game

20261006_LAD@ATL — published 4-3, 157 checks, 2 failed

Artifact. 20261006_LAD@ATL__20261007T023903Z.json, phase post_start_rerun, generated 2026-10-07T02:39:03, first pitch 2026-10-06T22:00:00.
Published 4 - 3 with p(home) 0.494846. Simulation draw chosen on not recorded; ? of ? draws realized the target exactly.

Stage chain: every stage that touched the score

Stage	Score after	Δ away	Δ home	Cum. away	Cum. home	Changed target	Own adjustment and description
1. `historical_ml`	3.807428442993566 - 4.634357958257989	0	0	0	0	no	Base ML historical model prediction
2. `variance_applied`	4 - 5	+0.1926	+0.3656	+0.1926	+0.3656	yes	Deterministic baseline target (variance factor 1.0)
3. `enhanced_prediction`	4 - 3	0	-2.0000	+0.1926	-1.6344	yes	Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run, matchup_recency_guardrail)
4. `kpi_layer`	4 - 3	0	0	+0.1926	-1.6344	no	away -0.5364; home +0.2218 Savant KPI blend applied to the model target
5. `embedded_savant_run_environment`	4 - 3	0	0	+0.1926	-1.6344	no	away -0.0093; home -0.1827 Target score adjusted by temporal Savant lineup quality versus opposing starter quality
6. `fundamental_edge_target_review`	4 - 3	0	0	+0.1926	-1.6344	no	Refreshed score target and win probability after Savant/pitcher/team-strength adjustments
7. `capped_target`	4 - 3	0	0	+0.1926	-1.6344	no	Capped at max 12 realistic runs
8. `simulation_result`	4 - 3	-	-	-	-	no	Final inning-by-inning simulation result

A stage reporting its own adjustment while Changed target reads no computed a movement that a later reconciliation absorbed, or that rounded away at the integer score. That is recorded rather than hidden.

KPI team layer: the blend, with its arithmetic

Target	Baseline	KPI only	Weight	Blended	Computation	Verdict
`away_runs`	4.0000	3.3123	0.78	3.4636	4.0000 + 0.78 × (3.3123 − 4.0000) = 3.4636	PASS
`home_runs`	3.0000	3.5040	0.44	3.2218	3.0000 + 0.44 × (3.5040 − 3.0000) = 3.2218	PASS
`home_win`	0.4929	0.8254	0.00	0.4929	0.4929 + 0.00 × (0.8254 − 0.4929) = 0.4929	PASS

Published by the layer: away 4, home 3, p(home) 0.492867, continuous away 3.4636 / home 3.2218

Inputs and gates. candidate 20261005T125039Z_f5114dc0, value age 9 days, lineup coverage away 0.8830 / home 0.8889, movement cap 2.5 runs.
values_present ok values_fresh ok one_source_bundle ok lineup_coverage ok target_eligible ok movement_within_limit ok published_movement_within_limit ok

What drove the KPI forecast for `away_runs`

intercept 4.2573 + sum of contributions = 3.3123; prediction 3.3123; contributions outside the top listed 0.2556 runs

Feature	Raw	Standardised	Runs contributed	Which KPIs it is built from
`home_platoon_barrel`	0.6503	2.019	-0.7956	platoon term, selected by the opposing starter's throwing hand
`away_sp_command`	1.2273	2.183	-0.6964	starter group: pitcher_strikeout_rate, pitcher_walk_rate
`away_off_contact`	0.3998	1.574	+0.6025	offence group: batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`away_off_discipline`	0.3548	1.341	+0.4771	offence group: batter_whiff_rate_vs_fastball, batter_whiff_rate_vs_breaking, batter_whiff_rate_vs_offspeed, batter_strikeout_rate ...
`home_off_top3_contact`	1.2701	1.642	-0.4634	mean of the three best contact bats, a tail a club mean cannot express
`home_off_qualified_share`	0.8889	1.083	-0.4006	share of the lineup carrying a qualifying sample, not a skill measurement
`away_sp_whiff`	0.7754	1.372	+0.3512	starter group: pitcher_whiff_rate, pitcher_whiff_rate_fastball, pitcher_whiff_rate_breaking
`home_off_power`	0.5329	2.154	+0.3498	offence group: batter_home_run_rate, batter_bat_speed
`away_x_breaking`	0.2511	1.192	-0.3360	batter x pitcher interaction: pitcher_whiff_rate_breaking, batter_whiff_rate_vs_breaking
`home_sp_suppress`	0.5757	0.854	-0.2892	starter group: pitcher_barrel_rate_allowed, pitcher_hard_hit_rate_allowed, pitcher_contact_xwoba_allowed, pitcher_home_run_rate_allowed

What drove the KPI forecast for `home_runs`

intercept 4.2997 + sum of contributions = 3.5040; prediction 3.5040; contributions outside the top listed -0.4713 runs

Feature	Raw	Standardised	Runs contributed	Which KPIs it is built from
`home_off_power`	0.5329	2.154	+0.5722	offence group: batter_home_run_rate, batter_bat_speed
`away_off_ground`	0.5283	1.634	+0.3967	offence group: batter_ground_ball_rate_vs_slider, batter_ground_ball_rate_vs_fastball
`home_pen_minus_sp`	-0.2144	-1.823	-0.3904	composite feature
`away_x_suppress`	0.3247	1.324	-0.3180	batter x pitcher interaction: pitcher_barrel_rate_allowed, batter_barrel_rate_vs_fastball
`home_sp_command`	0.7570	1.209	-0.2936	starter group: pitcher_strikeout_rate, pitcher_walk_rate
`away_off_top3_contact`	1.1341	1.254	-0.2723	mean of the three best contact bats, a tail a club mean cannot express
`away_off_contact`	0.3998	1.574	+0.2554	offence group: batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`away_pen_minus_sp`	-0.2423	-1.782	-0.2550	composite feature
`home_platoon_barrel`	0.6503	2.019	-0.2478	platoon term, selected by the opposing starter's throwing hand
`home_off_worst3_discipline`	-0.3315	0.759	+0.2284	mean of the three weakest plate approaches

These are Ridge coefficients on standardised inputs, so each contribution is that feature's own push on the run forecast, in runs. They sum exactly to the prediction minus the intercept, which makes the list an accounting of the forecast rather than a ranking of vaguely important things.

Per-player metric movement

LAD against Evan Phillips (RHP)

The pipeline applied one club-wide ×0.8659 to every LAD hitter line, identical for the leadoff hitter and the ninth bat.
Opposing staff exposure. These hitters face the starter for 22.1517 of 36.938 projected batters (60.0%) and the 3 projected relievers for the remaining 40.0%. Every metric below is the exposure-weighted blend of both matchups; the starter-only figure is shown beside it. Real 2026 game logs put a starter at 21.75 batters faced on average (median 23, p10-p90 17-27) and a whole staff at 37.9: this starter is inside that band and this staff is inside its. The per-player columns are the measured alternative: each subject moved by his own KPI pairing. Exposure for a hitter is his projected plate appearances; for the starter it is batters faced.

Player	Pos	Metric	Published → KPI	Δ	League	Own rate	Opponent rate	Log5	Own skill	Opponent	Remainder
Mookie Betts	SS	Hitswhole staff	0.939 → 0.900	-0.040	0.2169	0.2268n=523 Good	0.2069n=628 Good	0.2165	+0.041	-0.041	-0.001 ok
		split	starter x0.9546 at 0.600 share (log5 0.2165 vs his 0.2069) + bullpen x0.9626 at 0.400 share (log5 0.2183 vs pen 0.2087) = x0.9578 \| starter alone would have been -0.043
		Strikeoutswhole staff	0.567 → 0.711	+0.144	0.2212	0.1168n=523 Very good	0.2947n=628 Very good	0.1629	-0.429	+0.302	-0.113 ok
		split	starter x1.3946 at 0.600 share (log5 0.1629 vs his 0.2947) + bullpen x1.0432 at 0.400 share (log5 0.1218 vs pen 0.2296) = x1.2539 \| starter alone would have been +0.224
		Walkswhole staff	0.281 → 0.211	-0.070	0.0889	0.0866n=523 Average	0.0573n=628 Very good	0.0557	-0.010	-0.130	+0.003 ok
		split	starter x0.6438 at 0.600 share (log5 0.0557 vs his 0.0573) + bullpen x0.9106 at 0.400 share (log5 0.0788 vs pen 0.0810) = x0.7506 \| starter alone would have been -0.100
		Home runswhole staff	0.173 → 0.138	-0.035	0.0304	0.0394n=523 Good	0.0219n=628 Very good	0.0285	+0.037	-0.035	-0.010 ok
		split	starter x0.7237 at 0.600 share (log5 0.0285 vs his 0.0219) + bullpen x0.9069 at 0.400 share (log5 0.0357 vs pen 0.0275) = x0.797 \| starter alone would have been -0.048
Neutralising Mookie Betts to the qualified-population mean moves the club's KPI run forecast by 0.3764 runs, and the weighted stage by 0.2936. Interactions make these non-additive across players, so they do not sum to the published movement.
Freddie Freeman	1B	Hitswhole staff	1.000 → 0.959	-0.042	0.2169	0.2394n=652 Very good	0.2069n=628 Good	0.2287	+0.094	-0.042	-0.003 ok
		split	starter x0.9553 at 0.600 share (log5 0.2287 vs his 0.2069) + bullpen x0.9632 at 0.400 share (log5 0.2306 vs pen 0.2087) = x0.9585 \| starter alone would have been -0.045
		Strikeoutswhole staff	0.889 → 1.093	+0.205	0.2212	0.1781n=652 Good	0.2947n=628 Very good	0.2417	-0.181	+0.308	-0.041 ok
		split	starter x1.3575 at 0.600 share (log5 0.2417 vs his 0.2947) + bullpen x1.0401 at 0.400 share (log5 0.1852 vs pen 0.2296) = x1.2304 \| starter alone would have been +0.318
		Walkswhole staff	0.275 → 0.207	-0.068	0.0889	0.0993n=652 Average	0.0573n=628 Very good	0.0643	+0.043	-0.132	-0.014 ok
		split	starter x0.647 at 0.600 share (log5 0.0643 vs his 0.0573) + bullpen x0.9117 at 0.400 share (log5 0.0906 vs pen 0.0810) = x0.753 \| starter alone would have been -0.097
		Home runswhole staff	0.129 → 0.103	-0.026	0.0304	0.0262n=652 Average	0.0219n=628 Very good	0.0189	-0.017	-0.035	+0.005 ok
		split	starter x0.721 at 0.600 share (log5 0.0189 vs his 0.0219) + bullpen x0.9057 at 0.400 share (log5 0.0238 vs pen 0.0275) = x0.7949 \| starter alone would have been -0.036
Neutralising Freddie Freeman to the qualified-population mean moves the club's KPI run forecast by 0.3824 runs, and the weighted stage by 0.2982. Interactions make these non-additive across players, so they do not sum to the published movement.
Teoscar Hernández	LF	Hitswhole staff	0.854 → 0.818	-0.036	0.2169	0.2292n=487 Good	0.2069n=628 Good	0.2188	+0.049	-0.040	-0.002 ok
		split	starter x0.9547 at 0.600 share (log5 0.2188 vs his 0.2069) + bullpen x0.9627 at 0.400 share (log5 0.2206 vs pen 0.2087) = x0.9579 \| starter alone would have been -0.039
		Strikeoutswhole staff	1.322 → 1.584	+0.262	0.2212	0.2684n=487 Below average	0.2947n=628 Very good	0.3506	+0.188	+0.293	+0.034 ok
		split	starter x1.3061 at 0.600 share (log5 0.3506 vs his 0.2947) + bullpen x1.0355 at 0.400 share (log5 0.2779 vs pen 0.2296) = x1.1978 \| starter alone would have been +0.405
		Walkswhole staff	0.259 → 0.194	-0.065	0.0889	0.0868n=487 Average	0.0573n=628 Very good	0.0559	-0.009	-0.126	+0.003 ok
		split	starter x0.6438 at 0.600 share (log5 0.0559 vs his 0.0573) + bullpen x0.9106 at 0.400 share (log5 0.0790 vs pen 0.0810) = x0.7506 \| starter alone would have been -0.092
		Home runswhole staff	0.139 → 0.111	-0.028	0.0304	0.0320n=487 Average	0.0219n=628 Very good	0.0231	+0.006	-0.034	-0.002 ok
		split	starter x0.7222 at 0.600 share (log5 0.0231 vs his 0.0219) + bullpen x0.9062 at 0.400 share (log5 0.0290 vs pen 0.0275) = x0.7959 \| starter alone would have been -0.039
Neutralising Teoscar Hernández to the qualified-population mean moves the club's KPI run forecast by 0.1722 runs, and the weighted stage by 0.1343. Interactions make these non-additive across players, so they do not sum to the published movement.
Will Smith	C	Hitswhole staff	0.776 → 0.743	-0.033	0.2169	0.2185n=276 Average	0.2069n=628 Good	0.2084	+0.006	-0.039	-0.000 ok
		split	starter x0.9541 at 0.600 share (log5 0.2084 vs his 0.2069) + bullpen x0.9622 at 0.400 share (log5 0.2102 vs pen 0.2087) = x0.9573 \| starter alone would have been -0.036
		Strikeoutswhole staff	0.857 → 1.057	+0.200	0.2212	0.1718n=276 Good	0.2947n=628 Very good	0.2338	-0.196	+0.291	-0.045 ok
		split	starter x1.3612 at 0.600 share (log5 0.2338 vs his 0.2947) + bullpen x1.0404 at 0.400 share (log5 0.1787 vs pen 0.2296) = x1.2328 \| starter alone would have been +0.310
		Walkswhole staff	0.319 → 0.241	-0.078	0.0889	0.1129n=276 Good	0.0573n=628 Very good	0.0734	+0.095	-0.125	-0.031 ok
		split	starter x0.6505 at 0.600 share (log5 0.0734 vs his 0.0573) + bullpen x0.913 at 0.400 share (log5 0.1030 vs pen 0.0810) = x0.7556 \| starter alone would have been -0.111
		Home runswhole staff	0.129 → 0.102	-0.026	0.0304	0.0334n=276 Average	0.0219n=628 Very good	0.0241	+0.012	-0.033	-0.003 ok
		split	starter x0.7225 at 0.600 share (log5 0.0241 vs his 0.0219) + bullpen x0.9064 at 0.400 share (log5 0.0302 vs pen 0.0275) = x0.7961 \| starter alone would have been -0.036
Neutralising Will Smith to the qualified-population mean moves the club's KPI run forecast by 0.3409 runs, and the weighted stage by 0.2659. Interactions make these non-additive across players, so they do not sum to the published movement.
Shohei Ohtani	DH	Hitswhole staff	0.879 → 0.841	-0.037	0.2169	0.2255n=613 Good	0.2069n=628 Good	0.2152	+0.035	-0.041	-0.001 ok
		split	starter x0.9545 at 0.600 share (log5 0.2152 vs his 0.2069) + bullpen x0.9625 at 0.400 share (log5 0.2170 vs pen 0.2087) = x0.9577 \| starter alone would have been -0.040
		Strikeoutswhole staff	1.270 → 1.532	+0.262	0.2212	0.2445n=613 Average	0.2947n=628 Very good	0.3226	+0.095	+0.301	+0.019 ok
		split	starter x1.3193 at 0.600 share (log5 0.3226 vs his 0.2947) + bullpen x1.0367 at 0.400 share (log5 0.2535 vs pen 0.2296) = x1.2062 \| starter alone would have been +0.406
		Walkswhole staff	0.329 → 0.249	-0.080	0.0889	0.1251n=613 Very good	0.0573n=628 Very good	0.0818	+0.148	-0.130	-0.048 ok
		split	starter x0.6536 at 0.600 share (log5 0.0818 vs his 0.0573) + bullpen x0.9141 at 0.400 share (log5 0.1144 vs pen 0.0810) = x0.7579 \| starter alone would have been -0.114
		Home runswhole staff	0.159 → 0.127	-0.032	0.0304	0.0434n=613 Very good	0.0219n=628 Very good	0.0314	+0.053	-0.035	-0.014 ok
		split	starter x0.7246 at 0.600 share (log5 0.0314 vs his 0.0219) + bullpen x0.9072 at 0.400 share (log5 0.0393 vs pen 0.0275) = x0.7977 \| starter alone would have been -0.044
Neutralising Shohei Ohtani to the qualified-population mean moves the club's KPI run forecast by -0.1031 runs, and the weighted stage by -0.0804. Interactions make these non-additive across players, so they do not sum to the published movement.
Miguel Rojas	2B	Hitswhole staff	0.704 → 0.674	-0.030	0.2169	0.2246n=252 Good	0.2069n=628 Good	0.2144	+0.025	-0.033	-0.001 ok
		split	starter x0.9545 at 0.600 share (log5 0.2144 vs his 0.2069) + bullpen x0.9625 at 0.400 share (log5 0.2162 vs pen 0.2087) = x0.9577 \| starter alone would have been -0.032
		Strikeoutswhole staff	0.491 → 0.614	+0.123	0.2212	0.1256n=252 Very good	0.2947n=628 Very good	0.1745	-0.312	+0.240	-0.080 ok
		split	starter x1.3891 at 0.600 share (log5 0.1745 vs his 0.2947) + bullpen x1.0428 at 0.400 share (log5 0.1310 vs pen 0.2296) = x1.2505 \| starter alone would have been +0.191
		Walkswhole staff	0.165 → 0.123	-0.042	0.0889	0.0717n=252 Below average	0.0573n=628 Very good	0.0459	-0.056	-0.103	+0.019 ok
		split	starter x0.6401 at 0.600 share (log5 0.0459 vs his 0.0573) + bullpen x0.9093 at 0.400 share (log5 0.0652 vs pen 0.0810) = x0.7479 \| starter alone would have been -0.059
		Home runswhole staff	0.076 → 0.060	-0.015	0.0304	0.0233n=252 Below average	0.0219n=628 Very good	0.0168	-0.023	-0.028	+0.006 ok
		split	starter x0.7204 at 0.600 share (log5 0.0168 vs his 0.0219) + bullpen x0.9055 at 0.400 share (log5 0.0211 vs pen 0.0275) = x0.7945 \| starter alone would have been -0.021
Neutralising Miguel Rojas to the qualified-population mean moves the club's KPI run forecast by 0.2808 runs, and the weighted stage by 0.2190. Interactions make these non-additive across players, so they do not sum to the published movement.
Andy Pages	CF	Hitswhole staff	0.780 → 0.747	-0.033	0.2169	0.2323n=569 Good	0.2069n=628 Good	0.2219	+0.057	-0.037	-0.002 ok
		split	starter x0.9549 at 0.600 share (log5 0.2219 vs his 0.2069) + bullpen x0.9628 at 0.400 share (log5 0.2237 vs pen 0.2087) = x0.9581 \| starter alone would have been -0.035
		Strikeoutswhole staff	0.841 → 1.028	+0.188	0.2212	0.1971n=569 Average	0.2947n=628 Very good	0.2654	-0.088	+0.270	-0.019 ok
		split	starter x1.3463 at 0.600 share (log5 0.2654 vs his 0.2947) + bullpen x1.0391 at 0.400 share (log5 0.2049 vs pen 0.2296) = x1.2233 \| starter alone would have been +0.291
		Walkswhole staff	0.247 → 0.185	-0.062	0.0889	0.0866n=569 Average	0.0573n=628 Very good	0.0558	-0.009	-0.116	+0.003 ok
		split	starter x0.6438 at 0.600 share (log5 0.0558 vs his 0.0573) + bullpen x0.9106 at 0.400 share (log5 0.0789 vs pen 0.0810) = x0.7506 \| starter alone would have been -0.088
		Home runswhole staff	0.137 → 0.109	-0.028	0.0304	0.0372n=569 Good	0.0219n=628 Very good	0.0269	+0.025	-0.031	-0.007 ok
		split	starter x0.7233 at 0.600 share (log5 0.0269 vs his 0.0219) + bullpen x0.9067 at 0.400 share (log5 0.0338 vs pen 0.0275) = x0.7967 \| starter alone would have been -0.038
Neutralising Andy Pages to the qualified-population mean moves the club's KPI run forecast by 0.2434 runs, and the weighted stage by 0.1898. Interactions make these non-additive across players, so they do not sum to the published movement.
Kyle Tucker	RF	Hitswhole staff	0.739 → 0.707	-0.032	0.2169	0.2076n=622 Below average	0.2069n=628 Good	0.1979	-0.035	-0.037	+0.001 ok
		split	starter x0.9535 at 0.600 share (log5 0.1979 vs his 0.2069) + bullpen x0.9617 at 0.400 share (log5 0.1996 vs pen 0.2087) = x0.9568 \| starter alone would have been -0.034
		Strikeoutswhole staff	0.887 → 1.094	+0.207	0.2212	0.1711n=622 Good	0.2947n=628 Very good	0.2330	-0.188	+0.276	-0.044 ok
		split	starter x1.3616 at 0.600 share (log5 0.2330 vs his 0.2947) + bullpen x1.0405 at 0.400 share (log5 0.1780 vs pen 0.2296) = x1.2331 \| starter alone would have been +0.321
		Walkswhole staff	0.275 → 0.208	-0.067	0.0889	0.1195n=622 Very good	0.0573n=628 Very good	0.0780	+0.115	-0.119	-0.037 ok
		split	starter x0.6522 at 0.600 share (log5 0.0780 vs his 0.0573) + bullpen x0.9136 at 0.400 share (log5 0.1092 vs pen 0.0810) = x0.7568 \| starter alone would have been -0.096
		Home runswhole staff	0.101 → 0.080	-0.021	0.0304	0.0282n=622 Average	0.0219n=628 Very good	0.0204	-0.008	-0.032	+0.002 ok
		split	starter x0.7214 at 0.600 share (log5 0.0204 vs his 0.0219) + bullpen x0.9059 at 0.400 share (log5 0.0256 vs pen 0.0275) = x0.7953 \| starter alone would have been -0.028
Neutralising Kyle Tucker to the qualified-population mean moves the club's KPI run forecast by 0.1829 runs, and the weighted stage by 0.1427. Interactions make these non-additive across players, so they do not sum to the published movement.
Enrique Hernández	3B	Hitswhole staff	0.676 → 0.647	-0.029	0.2169	0.2169n=0 averagebelow floor	0.2069n=628 Good	0.2069	0	-0.033	0 ok
		split	starter x0.954 at 0.600 share (log5 0.2069 vs his 0.2069) + bullpen x0.9621 at 0.400 share (log5 0.2087 vs pen 0.2087) = x0.9572 \| starter alone would have been -0.031
		Strikeoutswhole staff	1.078 → 1.309	+0.231	0.2212	0.2212n=0 averagebelow floor	0.2947n=628 Very good	0.2947	0	+0.242	0 ok
		split	starter x1.3325 at 0.600 share (log5 0.2947 vs his 0.2947) + bullpen x1.0379 at 0.400 share (log5 0.2296 vs pen 0.2296) = x1.2146 \| starter alone would have been +0.358
		Walkswhole staff	0.196 → 0.147	-0.049	0.0889	0.0889n=0 averagebelow floor	0.0573n=628 Very good	0.0573	0	-0.104	0 ok
		split	starter x0.6444 at 0.600 share (log5 0.0573 vs his 0.0573) + bullpen x0.9108 at 0.400 share (log5 0.0810 vs pen 0.0810) = x0.751 \| starter alone would have been -0.070
		Home runswhole staff	0.108 → 0.086	-0.022	0.0304	0.0304n=0 averagebelow floor	0.0219n=628 Very good	0.0219	0	-0.028	0 ok
		split	starter x0.7219 at 0.600 share (log5 0.0219 vs his 0.0219) + bullpen x0.9061 at 0.400 share (log5 0.0275 vs pen 0.0275) = x0.7956 \| starter alone would have been -0.030
Neutralising Enrique Hernández to the qualified-population mean moves the club's KPI run forecast by 0.0283 runs, and the weighted stage by 0.0220. Interactions make these non-additive across players, so they do not sum to the published movement.
STARTER Yoshinobu Yamamoto23.1926 BF	Starter	Hits	3.425 → 3.557	+0.132	0.2169	0.1839n=678 Very good	0.22498/9 qualified lineup mean	0.1910	-0.764	+0.185	-0.021 ok
		Strikeouts	6.455 → 6.206	-0.249	0.2212	0.2542n=678 Good	0.21238/9 qualified lineup mean	0.2444	+0.765	-0.206	-0.021 ok
		Walks	1.270 → 1.191	-0.078	0.0889	0.0670n=678 Very good	0.08368/9 qualified lineup mean	0.0629	-0.508	-0.124	+0.029 ok
		Home runs	0.799 → 0.923	+0.124	0.0304	0.0290n=678 Average	0.03518/9 qualified lineup mean	0.0335	-0.033	+0.110	-0.005 ok
RELIEVER Tanner Scott2.1768 BF	Reliever	Hits	0.446 → 0.463	+0.017	0.2169	0.2051n=251 Good	0.22498/9 qualified lineup mean	0.2127	-0.026	+0.017	-0.001 ok
		Strikeouts	0.655 → 0.632	-0.024	0.2212	0.3011n=251 Very good	0.21238/9 qualified lineup mean	0.2902	+0.174	-0.019	-0.004 ok
		Walks	0.155 → 0.145	-0.009	0.0889	0.0710n=251 Good	0.08368/9 qualified lineup mean	0.0666	-0.039	-0.012	+0.002 ok
		Home runs	0.055 → 0.063	+0.009	0.0304	0.0252n=251 Good	0.03518/9 qualified lineup mean	0.0291	-0.011	+0.010	-0.002 ok
RELIEVER Kris Bubic2.1102 BF	Reliever	Hits	0.421 → 0.437	+0.016	0.2169	0.1996n=226 Good	0.22498/9 qualified lineup mean	0.2071	-0.036	+0.017	-0.001 ok
		Strikeouts	0.508 → 0.488	-0.020	0.2212	0.2408n=226 Average	0.21238/9 qualified lineup mean	0.2314	+0.041	-0.019	-0.001 ok
		Walks	0.220 → 0.207	-0.013	0.0889	0.1043n=226 Below average	0.08368/9 qualified lineup mean	0.0981	+0.033	-0.011	-0.002 ok
		Home runs	0.055 → 0.064	+0.009	0.0304	0.0261n=226 Good	0.03518/9 qualified lineup mean	0.0302	-0.009	+0.010	-0.001 ok
RELIEVER Alex Vesia1.9069 BF	Middle Reliever (High Leverage)	Hits	0.387 → 0.402	+0.015	0.2169	0.1766n=237 Very good	0.22498/9 qualified lineup mean	0.1835	-0.077	+0.015	-0.002 ok
		Strikeouts	1.336 → 1.287	-0.049	0.2212	0.2892n=237 Very good	0.21238/9 qualified lineup mean	0.2786	+0.130	-0.017	-0.003 ok
		Walks	0.452 → 0.426	-0.026	0.0889	0.1166n=237 Poor	0.08368/9 qualified lineup mean	0.1098	+0.053	-0.010	-0.003 ok
		Home runs	0.052 → 0.060	+0.008	0.0304	0.0257n=237 Good	0.03518/9 qualified lineup mean	0.0297	-0.009	+0.009	-0.001 ok
RELIEVER Brock Stewart1.7123 BF	Middle Reliever	Hits	0.522 → 0.541	+0.019	0.2169	0.2169n=0 averagebelow floor	0.22498/9 qualified lineup mean	0.2249	0	+0.014	0 ok
		Strikeouts	0.965 → 0.926	-0.039	0.2212	0.2212n=0 averagebelow floor	0.21238/9 qualified lineup mean	0.2123	0	-0.015	0 ok
		Walks	0.331 → 0.311	-0.020	0.0889	0.0889n=0 averagebelow floor	0.08368/9 qualified lineup mean	0.0836	0	-0.009	0 ok
		Home runs	0.058 → 0.067	+0.009	0.0304	0.0304n=0 averagebelow floor	0.03518/9 qualified lineup mean	0.0351	0	+0.008	0 ok
RELIEVER Edwin Díaz1.6859 BF	Reliever	Hits	0.366 → 0.379	+0.013	0.2169	0.2169n=0 averagebelow floor	0.22498/9 qualified lineup mean	0.2249	0	+0.013	0 ok
		Strikeouts	0.373 → 0.358	-0.015	0.2212	0.2212n=0 averagebelow floor	0.21238/9 qualified lineup mean	0.2123	0	-0.015	0 ok
		Walks	0.150 → 0.141	-0.009	0.0889	0.0889n=0 averagebelow floor	0.08368/9 qualified lineup mean	0.0836	0	-0.009	0 ok
		Home runs	0.051 → 0.059	+0.008	0.0304	0.0304n=0 averagebelow floor	0.03518/9 qualified lineup mean	0.0351	0	+0.008	0 ok
RELIEVER Roki Sasaki1.6106 BF	Reliever	Hits	0.349 → 0.362	+0.013	0.2169	0.2166n=522 Average	0.22498/9 qualified lineup mean	0.2246	-0.000	+0.013	0 ok
		Strikeouts	0.376 → 0.361	-0.015	0.2212	0.2337n=522 Average	0.21238/9 qualified lineup mean	0.2245	+0.020	-0.014	-0.001 ok
		Walks	0.147 → 0.138	-0.009	0.0889	0.0913n=522 Average	0.08368/9 qualified lineup mean	0.0858	+0.004	-0.009	-0.000 ok
		Home runs	0.059 → 0.068	+0.009	0.0304	0.0367n=522 Poor	0.03518/9 qualified lineup mean	0.0424	+0.010	+0.008	+0.002 ok
RELIEVER Evan Phillips1.3495 BF	Middle Reliever	Hits	0.474 → 0.491	+0.018	0.2169	0.2169n=0 averagebelow floor	0.22498/9 qualified lineup mean	0.2249	0	+0.011	0 ok
		Strikeouts	0.835 → 0.801	-0.034	0.2212	0.2212n=0 averagebelow floor	0.21238/9 qualified lineup mean	0.2123	0	-0.012	0 ok
		Walks	0.170 → 0.160	-0.010	0.0889	0.0889n=0 averagebelow floor	0.08368/9 qualified lineup mean	0.0836	0	-0.007	0 ok
		Home runs	0.101 → 0.117	+0.016	0.0304	0.0304n=0 averagebelow floor	0.03518/9 qualified lineup mean	0.0351	0	+0.006	0 ok

ATL against Dylan Dodd (LHP)

The pipeline applied one club-wide ×1.0739 to every ATL hitter line, identical for the leadoff hitter and the ninth bat.
Opposing staff exposure. These hitters face the starter for 23.1926 of 35.746 projected batters (64.9%) and the 3 projected relievers for the remaining 35.1%. Every metric below is the exposure-weighted blend of both matchups; the starter-only figure is shown beside it. Real 2026 game logs put a starter at 21.75 batters faced on average (median 23, p10-p90 17-27) and a whole staff at 37.9: this starter is inside that band and this staff is inside its. The per-player columns are the measured alternative: each subject moved by his own KPI pairing. Exposure for a hitter is his projected plate appearances; for the starter it is batters faced.

Player	Pos	Metric	Published → KPI	Δ	League	Own rate	Opponent rate	Log5	Own skill	Opponent	Remainder
Drake Baldwin	C	Hitswhole staff	0.990 → 0.877	-0.113	0.2169	0.2374n=598 Very good	0.1839n=678 Very good	0.2021	+0.090	-0.145	-0.010 ok
		split	starter x0.8515 at 0.649 share (log5 0.2021 vs his 0.1839) + bullpen x0.9501 at 0.351 share (log5 0.2255 vs pen 0.2058) = x0.8861 \| starter alone would have been -0.147
		Strikeoutswhole staff	0.986 → 1.129	+0.143	0.2212	0.2099n=598 Average	0.2542n=678 Good	0.2418	-0.050	+0.145	-0.005 ok
		split	starter x1.1517 at 0.649 share (log5 0.2418 vs his 0.2542) + bullpen x1.1337 at 0.351 share (log5 0.2380 vs pen 0.2503) = x1.1454 \| starter alone would have been +0.149
		Walkswhole staff	0.346 → 0.296	-0.050	0.0889	0.0915n=598 Average	0.0670n=678 Very good	0.0690	+0.011	-0.096	-0.003 ok
		split	starter x0.7544 at 0.649 share (log5 0.0690 vs his 0.0670) + bullpen x1.0447 at 0.351 share (log5 0.0956 vs pen 0.0929) = x0.8564 \| starter alone would have been -0.085
		Home runswhole staff	0.186 → 0.178	-0.009	0.0304	0.0371n=598 Good	0.0290n=678 Average	0.0354	+0.030	-0.006	-0.001 ok
		split	starter x0.9534 at 0.649 share (log5 0.0354 vs his 0.0290) + bullpen x0.9504 at 0.351 share (log5 0.0353 vs pen 0.0289) = x0.9523 \| starter alone would have been -0.009
Neutralising Drake Baldwin to the qualified-population mean moves the club's KPI run forecast by -0.0093 runs, and the weighted stage by -0.0041. Interactions make these non-additive across players, so they do not sum to the published movement.
Ronald Acuña Jr.	DH	Hitswhole staff	0.885 → 0.782	-0.103	0.2169	0.2212n=457 Average	0.1839n=678 Very good	0.1878	+0.018	-0.140	-0.002 ok
		split	starter x0.8488 at 0.649 share (log5 0.1878 vs his 0.1839) + bullpen x0.9491 at 0.351 share (log5 0.2099 vs pen 0.2058) = x0.884 \| starter alone would have been -0.134
		Strikeoutswhole staff	0.990 → 1.132	+0.142	0.2212	0.2190n=457 Average	0.2542n=678 Good	0.2518	-0.009	+0.140	-0.001 ok
		split	starter x1.1497 at 0.649 share (log5 0.2518 vs his 0.2542) + bullpen x1.132 at 0.351 share (log5 0.2480 vs pen 0.2503) = x1.1435 \| starter alone would have been +0.148
		Walkswhole staff	0.363 → 0.312	-0.051	0.0889	0.1104n=457 Good	0.0670n=678 Very good	0.0837	+0.091	-0.093	-0.020 ok
		split	starter x0.7583 at 0.649 share (log5 0.0837 vs his 0.0670) + bullpen x1.0437 at 0.351 share (log5 0.1152 vs pen 0.0929) = x0.8585 \| starter alone would have been -0.088
		Home runswhole staff	0.177 → 0.169	-0.008	0.0304	0.0389n=457 Good	0.0290n=678 Average	0.0371	+0.036	-0.006	-0.002 ok
		split	starter x0.9535 at 0.649 share (log5 0.0371 vs his 0.0290) + bullpen x0.9505 at 0.351 share (log5 0.0369 vs pen 0.0289) = x0.9524 \| starter alone would have been -0.008
Neutralising Ronald Acuña Jr. to the qualified-population mean moves the club's KPI run forecast by 0.1862 runs, and the weighted stage by 0.0819. Interactions make these non-additive across players, so they do not sum to the published movement.
Matt Olson	1B	Hitswhole staff	0.836 → 0.739	-0.097	0.2169	0.2232n=682 Average	0.1839n=678 Very good	0.1895	+0.026	-0.137	-0.003 ok
		split	starter x0.8491 at 0.649 share (log5 0.1895 vs his 0.1839) + bullpen x0.9492 at 0.351 share (log5 0.2119 vs pen 0.2058) = x0.8843 \| starter alone would have been -0.126
		Strikeoutswhole staff	1.080 → 1.230	+0.150	0.2212	0.2406n=682 Average	0.2542n=678 Good	0.2754	+0.080	+0.137	+0.008 ok
		split	starter x1.1449 at 0.649 share (log5 0.2754 vs his 0.2542) + bullpen x1.1279 at 0.351 share (log5 0.2713 vs pen 0.2503) = x1.1389 \| starter alone would have been +0.157
		Walkswhole staff	0.323 → 0.277	-0.046	0.0889	0.0919n=682 Average	0.0670n=678 Very good	0.0693	+0.012	-0.091	-0.003 ok
		split	starter x0.7545 at 0.649 share (log5 0.0693 vs his 0.0670) + bullpen x1.0447 at 0.351 share (log5 0.0960 vs pen 0.0929) = x0.8564 \| starter alone would have been -0.079
		Home runswhole staff	0.228 → 0.218	-0.011	0.0304	0.0518n=682 Very good	0.0290n=678 Average	0.0494	+0.089	-0.006	-0.004 ok
		split	starter x0.9541 at 0.649 share (log5 0.0494 vs his 0.0290) + bullpen x0.9511 at 0.351 share (log5 0.0493 vs pen 0.0289) = x0.953 \| starter alone would have been -0.011
Neutralising Matt Olson to the qualified-population mean moves the club's KPI run forecast by 0.3441 runs, and the weighted stage by 0.1514. Interactions make these non-additive across players, so they do not sum to the published movement.
Ozzie Albies	2B	Hitswhole staff	0.822 → 0.726	-0.096	0.2169	0.2154n=667 Average	0.1839n=678 Very good	0.1826	-0.006	-0.129	+0.001 ok
		split	starter x0.8478 at 0.649 share (log5 0.1826 vs his 0.1839) + bullpen x0.9487 at 0.351 share (log5 0.2043 vs pen 0.2058) = x0.8832 \| starter alone would have been -0.125
		Strikeoutswhole staff	0.697 → 0.806	+0.109	0.2212	0.1576n=667 Good	0.2542n=678 Good	0.1833	-0.249	+0.129	-0.028 ok
		split	starter x1.1634 at 0.649 share (log5 0.1833 vs his 0.2542) + bullpen x1.1438 at 0.351 share (log5 0.1802 vs pen 0.2503) = x1.1565 \| starter alone would have been +0.114
		Walkswhole staff	0.257 → 0.220	-0.037	0.0889	0.0707n=667 Below average	0.0670n=678 Very good	0.0530	-0.071	-0.086	+0.017 ok
		split	starter x0.7502 at 0.649 share (log5 0.0530 vs his 0.0670) + bullpen x1.0458 at 0.351 share (log5 0.0740 vs pen 0.0929) = x0.854 \| starter alone would have been -0.064
		Home runswhole staff	0.146 → 0.139	-0.007	0.0304	0.0322n=667 Average	0.0290n=678 Average	0.0307	+0.007	-0.006	-0.000 ok
		split	starter x0.9532 at 0.649 share (log5 0.0307 vs his 0.0290) + bullpen x0.9502 at 0.351 share (log5 0.0306 vs pen 0.0289) = x0.9521 \| starter alone would have been -0.007
Neutralising Ozzie Albies to the qualified-population mean moves the club's KPI run forecast by 0.6243 runs, and the weighted stage by 0.2747. Interactions make these non-additive across players, so they do not sum to the published movement.
Michael Harris II	CF	Hitswhole staff	1.031 → 0.915	-0.115	0.2169	0.2533n=608 Very good	0.1839n=678 Very good	0.2163	+0.146	-0.132	-0.016 ok
		split	starter x0.8541 at 0.649 share (log5 0.2163 vs his 0.1839) + bullpen x0.951 at 0.351 share (log5 0.2409 vs pen 0.2058) = x0.8881 \| starter alone would have been -0.150
		Strikeoutswhole staff	0.877 → 1.008	+0.131	0.2212	0.1915n=608 Average	0.2542n=678 Good	0.2214	-0.119	+0.133	-0.013 ok
		split	starter x1.1558 at 0.649 share (log5 0.2214 vs his 0.2542) + bullpen x1.1372 at 0.351 share (log5 0.2178 vs pen 0.2503) = x1.1493 \| starter alone would have been +0.137
		Walkswhole staff	0.151 → 0.129	-0.022	0.0889	0.0462n=608 Poor	0.0670n=678 Very good	0.0344	-0.172	-0.088	+0.041 ok
		split	starter x0.7453 at 0.649 share (log5 0.0344 vs his 0.0670) + bullpen x1.047 at 0.351 share (log5 0.0484 vs pen 0.0929) = x0.8513 \| starter alone would have been -0.038
		Home runswhole staff	0.179 → 0.171	-0.009	0.0304	0.0379n=608 Good	0.0290n=678 Average	0.0361	+0.030	-0.006	-0.001 ok
		split	starter x0.9535 at 0.649 share (log5 0.0361 vs his 0.0290) + bullpen x0.9504 at 0.351 share (log5 0.0360 vs pen 0.0289) = x0.9524 \| starter alone would have been -0.008
Neutralising Michael Harris II to the qualified-population mean moves the club's KPI run forecast by 0.0680 runs, and the weighted stage by 0.0299. Interactions make these non-additive across players, so they do not sum to the published movement.
Mauricio Dubón	LF	Hitswhole staff	0.897 → 0.795	-0.102	0.2169	0.2382n=603 Very good	0.1839n=678 Very good	0.2028	+0.082	-0.126	-0.009 ok
		split	starter x0.8516 at 0.649 share (log5 0.2028 vs his 0.1839) + bullpen x0.9501 at 0.351 share (log5 0.2263 vs pen 0.2058) = x0.8862 \| starter alone would have been -0.133
		Strikeoutswhole staff	0.569 → 0.661	+0.093	0.2212	0.1270n=603 Very good	0.2542n=678 Good	0.1486	-0.360	+0.126	-0.043 ok
		split	starter x1.1703 at 0.649 share (log5 0.1486 vs his 0.2542) + bullpen x1.1498 at 0.351 share (log5 0.1460 vs pen 0.2503) = x1.1631 \| starter alone would have been +0.097
		Walkswhole staff	0.200 → 0.171	-0.029	0.0889	0.0631n=603 Below average	0.0670n=678 Very good	0.0473	-0.099	-0.084	+0.023 ok
		split	starter x0.7486 at 0.649 share (log5 0.0473 vs his 0.0670) + bullpen x1.0461 at 0.351 share (log5 0.0660 vs pen 0.0929) = x0.8531 \| starter alone would have been -0.050
		Home runswhole staff	0.099 → 0.094	-0.005	0.0304	0.0231n=603 Below average	0.0290n=678 Average	0.0220	-0.028	-0.005	+0.001 ok
		split	starter x0.9528 at 0.649 share (log5 0.0220 vs his 0.0290) + bullpen x0.9497 at 0.351 share (log5 0.0219 vs pen 0.0289) = x0.9517 \| starter alone would have been -0.005
Neutralising Mauricio Dubón to the qualified-population mean moves the club's KPI run forecast by 0.0602 runs, and the weighted stage by 0.0265. Interactions make these non-additive across players, so they do not sum to the published movement.
Lane Thomas	RF	Hitswhole staff	0.703 → 0.621	-0.083	0.2169	0.2079n=370 Below average	0.1839n=678 Very good	0.1760	-0.032	-0.116	+0.004 ok
		split	starter x0.8466 at 0.649 share (log5 0.1760 vs his 0.1839) + bullpen x0.9482 at 0.351 share (log5 0.1972 vs pen 0.2058) = x0.8823 \| starter alone would have been -0.108
		Strikeoutswhole staff	0.953 → 1.085	+0.132	0.2212	0.2455n=370 Average	0.2542n=678 Good	0.2808	+0.085	+0.116	+0.008 ok
		split	starter x1.1439 at 0.649 share (log5 0.2808 vs his 0.2542) + bullpen x1.1269 at 0.351 share (log5 0.2766 vs pen 0.2503) = x1.1379 \| starter alone would have been +0.137
		Walkswhole staff	0.292 → 0.251	-0.041	0.0889	0.1096n=370 Good	0.0670n=678 Very good	0.0831	+0.073	-0.077	-0.016 ok
		split	starter x0.7581 at 0.649 share (log5 0.0831 vs his 0.0670) + bullpen x1.0438 at 0.351 share (log5 0.1144 vs pen 0.0929) = x0.8584 \| starter alone would have been -0.071
		Home runswhole staff	0.117 → 0.112	-0.006	0.0304	0.0300n=370 Average	0.0290n=678 Average	0.0286	-0.001	-0.005	+0.000 ok
		split	starter x0.9531 at 0.649 share (log5 0.0286 vs his 0.0290) + bullpen x0.9501 at 0.351 share (log5 0.0285 vs pen 0.0289) = x0.952 \| starter alone would have been -0.005
Neutralising Lane Thomas to the qualified-population mean moves the club's KPI run forecast by 0.1430 runs, and the weighted stage by 0.0629. Interactions make these non-additive across players, so they do not sum to the published movement.
Austin Riley	3B	Hitswhole staff	0.711 → 0.627	-0.084	0.2169	0.2034n=614 Below average	0.1839n=678 Very good	0.1721	-0.048	-0.117	+0.006 ok
		split	starter x0.8459 at 0.649 share (log5 0.1721 vs his 0.1839) + bullpen x0.948 at 0.351 share (log5 0.1928 vs pen 0.2058) = x0.8818 \| starter alone would have been -0.110
		Strikeoutswhole staff	1.209 → 1.360	+0.151	0.2212	0.3090n=614 Poor	0.2542n=678 Good	0.3492	+0.313	+0.118	+0.026 ok
		split	starter x1.1302 at 0.649 share (log5 0.3492 vs his 0.2542) + bullpen x1.115 at 0.351 share (log5 0.3445 vs pen 0.2503) = x1.1249 \| starter alone would have been +0.157
		Walkswhole staff	0.233 → 0.199	-0.034	0.0889	0.0799n=614 Average	0.0670n=678 Very good	0.0601	-0.032	-0.078	+0.007 ok
		split	starter x0.752 at 0.649 share (log5 0.0601 vs his 0.0670) + bullpen x1.0453 at 0.351 share (log5 0.0835 vs pen 0.0929) = x0.855 \| starter alone would have been -0.058
		Home runswhole staff	0.121 → 0.115	-0.006	0.0304	0.0308n=614 Average	0.0290n=678 Average	0.0293	+0.001	-0.005	-0.000 ok
		split	starter x0.9531 at 0.649 share (log5 0.0293 vs his 0.0290) + bullpen x0.9501 at 0.351 share (log5 0.0292 vs pen 0.0289) = x0.952 \| starter alone would have been -0.006
Neutralising Austin Riley to the qualified-population mean moves the club's KPI run forecast by 0.1063 runs, and the weighted stage by 0.0468. Interactions make these non-additive across players, so they do not sum to the published movement.
Ha-Seong Kim	SS	Hitswhole staff	0.601 → 0.531	-0.070	0.2169	0.2169n=0 averagebelow floor	0.1839n=678 Very good	0.1839	0	-0.102	0 ok
		split	starter x0.8481 at 0.649 share (log5 0.1839 vs his 0.1839) + bullpen x0.9488 at 0.351 share (log5 0.2058 vs pen 0.2058) = x0.8835 \| starter alone would have been -0.091
		Strikeoutswhole staff	0.866 → 0.990	+0.124	0.2212	0.2212n=0 averagebelow floor	0.2542n=678 Good	0.2542	0	+0.102	0 ok
		split	starter x1.1492 at 0.649 share (log5 0.2542 vs his 0.2542) + bullpen x1.1315 at 0.351 share (log5 0.2503 vs pen 0.2503) = x1.143 \| starter alone would have been +0.129
		Walkswhole staff	0.257 → 0.220	-0.037	0.0889	0.0889n=0 averagebelow floor	0.0670n=678 Very good	0.0670	0	-0.068	0 ok
		split	starter x0.7539 at 0.649 share (log5 0.0670 vs his 0.0670) + bullpen x1.0448 at 0.351 share (log5 0.0929 vs pen 0.0929) = x0.8561 \| starter alone would have been -0.063
		Home runswhole staff	0.076 → 0.072	-0.004	0.0304	0.0304n=0 averagebelow floor	0.0290n=678 Average	0.0290	0	-0.004	0 ok
		split	starter x0.9531 at 0.649 share (log5 0.0290 vs his 0.0290) + bullpen x0.9501 at 0.351 share (log5 0.0289 vs pen 0.0289) = x0.952 \| starter alone would have been -0.004
Neutralising Ha-Seong Kim to the qualified-population mean moves the club's KPI run forecast by -0.0090 runs, and the weighted stage by -0.0039. Interactions make these non-additive across players, so they do not sum to the published movement.
STARTER Chris Sale22.1517 BF	Starter	Hits	4.088 → 4.240	+0.152	0.2169	0.2069n=628 Good	0.22488/9 qualified lineup mean	0.2146	-0.221	+0.176	-0.006 ok
		Strikeouts	6.870 → 5.947	-0.923	0.2212	0.2947n=628 Very good	0.18888/9 qualified lineup mean	0.2551	+1.629	-0.717	-0.160 ok
		Walks	0.856 → 0.948	+0.093	0.0889	0.0573n=628 Very good	0.09828/9 qualified lineup mean	0.0635	-0.701	+0.205	-0.068 ok
		Home runs	0.347 → 0.376	+0.029	0.0304	0.0219n=628 Very good	0.03298/9 qualified lineup mean	0.0237	-0.187	+0.055	-0.015 ok
RELIEVER Grant Holmes2.5669 BF	Reliever	Hits	0.555 → 0.575	+0.020	0.2169	0.2162n=607 Average	0.22488/9 qualified lineup mean	0.2241	-0.002	+0.020	0 ok
		Strikeouts	0.477 → 0.405	-0.072	0.2212	0.1859n=607 Below average	0.18888/9 qualified lineup mean	0.1576	-0.091	-0.083	+0.011 ok
		Walks	0.235 → 0.260	+0.024	0.0889	0.0916n=607 Average	0.09828/9 qualified lineup mean	0.1011	+0.007	+0.024	+0.001 ok
		Home runs	0.087 → 0.094	+0.007	0.0304	0.0338n=607 Below average	0.03298/9 qualified lineup mean	0.0365	+0.009	+0.006	+0.001 ok
RELIEVER Victor Mederos2.3407 BF	Reliever	Hits	0.481 → 0.498	+0.018	0.2169	0.2054n=148 Good	0.22488/9 qualified lineup mean	0.2130	-0.027	+0.019	-0.001 ok
		Strikeouts	0.531 → 0.454	-0.077	0.2212	0.2267n=148 Average	0.18888/9 qualified lineup mean	0.1937	+0.013	-0.076	-0.001 ok
		Walks	0.215 → 0.237	+0.022	0.0889	0.0917n=148 Average	0.09828/9 qualified lineup mean	0.1012	+0.006	+0.022	+0.001 ok
		Home runs	0.057 → 0.062	+0.005	0.0304	0.0245n=148 Very good	0.03298/9 qualified lineup mean	0.0265	-0.014	+0.006	-0.001 ok
RELIEVER Dylan Lee2.2682 BF	Middle Reliever (High Leverage)	Hits	0.744 → 0.772	+0.028	0.2169	0.2007n=253 Good	0.22488/9 qualified lineup mean	0.2082	-0.037	+0.018	-0.001 ok
		Strikeouts	1.044 → 0.901	-0.143	0.2212	0.2783n=253 Very good	0.18888/9 qualified lineup mean	0.2402	+0.130	-0.073	-0.013 ok
		Walks	0.247 → 0.273	+0.026	0.0889	0.0755n=253 Good	0.09828/9 qualified lineup mean	0.0835	-0.030	+0.021	-0.003 ok
		Home runs	0.149 → 0.161	+0.012	0.0304	0.0237n=253 Very good	0.03298/9 qualified lineup mean	0.0256	-0.015	+0.006	-0.001 ok
RELIEVER Robert Suarez2.252 BF	Middle Reliever (High Leverage)	Hits	0.817 → 0.847	+0.030	0.2169	0.2070n=130 Good	0.22488/9 qualified lineup mean	0.2147	-0.022	+0.018	-0.001 ok
		Strikeouts	0.599 → 0.510	-0.089	0.2212	0.2081n=130 Average	0.18888/9 qualified lineup mean	0.1772	-0.029	-0.073	+0.003 ok
		Walks	0.179 → 0.198	+0.019	0.0889	0.0696n=130 Good	0.09828/9 qualified lineup mean	0.0770	-0.043	+0.021	-0.004 ok
		Home runs	0.022 → 0.024	+0.002	0.0304	0.0253n=130 Good	0.03298/9 qualified lineup mean	0.0273	-0.011	+0.006	-0.001 ok
RELIEVER Brent Suter1.9666 BF	Reliever	Hits	0.439 → 0.455	+0.016	0.2169	0.2232n=377 Average	0.22488/9 qualified lineup mean	0.2313	+0.012	+0.016	+0.000 ok
		Strikeouts	0.412 → 0.351	-0.061	0.2212	0.2094n=377 Average	0.18888/9 qualified lineup mean	0.1784	-0.023	-0.064	+0.003 ok
		Walks	0.147 → 0.163	+0.016	0.0889	0.0748n=377 Good	0.09828/9 qualified lineup mean	0.0828	-0.028	+0.018	-0.003 ok
		Home runs	0.054 → 0.058	+0.004	0.0304	0.0274n=377 Average	0.03298/9 qualified lineup mean	0.0296	-0.006	+0.005	-0.001 ok
RELIEVER Raisel Iglesias1.8054 BF	Reliever	Hits	0.383 → 0.397	+0.014	0.2169	0.2121n=233 Average	0.22488/9 qualified lineup mean	0.2199	-0.009	+0.014	-0.000 ok
		Strikeouts	0.460 → 0.395	-0.065	0.2212	0.2546n=233 Good	0.18888/9 qualified lineup mean	0.2187	+0.060	-0.059	-0.006 ok
		Walks	0.143 → 0.159	+0.015	0.0889	0.0794n=233 Average	0.09828/9 qualified lineup mean	0.0878	-0.017	+0.017	-0.002 ok
		Home runs	0.052 → 0.056	+0.004	0.0304	0.0288n=233 Average	0.03298/9 qualified lineup mean	0.0312	-0.003	+0.004	-0.000 ok
RELIEVER Dylan Dodd1.5851 BF	Middle Reliever	Hits	0.678 → 0.704	+0.026	0.2169	0.1933n=186 Very good	0.22488/9 qualified lineup mean	0.2006	-0.037	+0.013	-0.001 ok
		Strikeouts	1.022 → 0.879	-0.143	0.2212	0.2620n=186 Good	0.18888/9 qualified lineup mean	0.2254	+0.065	-0.051	-0.007 ok
		Walks	0.223 → 0.246	+0.023	0.0889	0.0815n=186 Average	0.09828/9 qualified lineup mean	0.0901	-0.012	+0.015	-0.001 ok
		Home runs	0.154 → 0.167	+0.013	0.0304	0.0294n=186 Average	0.03298/9 qualified lineup mean	0.0318	-0.002	+0.004	-0.000 ok

Published pitching, integer draw beside the continuous model line

Pitcher	Role	BF	IP	expected IP	H int/cont	SO int/cont	BB int/cont	ER int/cont	Rate source
Yoshinobu Yamamoto	Starter	23.2	6.0	6.166	6 / 3.425	5 / 6.455	1 / 1.270	3 / 1.522	all_players_stats_recent_season_blend
Brock Stewart	Middle Reliever	3.9	1.0	1.014	1 / 0.522	0 / 0.965	1 / 0.331	0 / 0.246	all_players_stats_recent_season_blend
Evan Phillips	Middle Reliever	2.7	0.2	0.676	0 / 0.474	1 / 0.835	0 / 0.170	0 / 0.231	all_players_stats_recent_season_blend
Alex Vesia	Middle Reliever (High Leverage)	3.9	2.1	1.014	0 / 0.387	2 / 1.336	1 / 0.452	0 / 0.218	all_players_stats_recent_season_blend
Chris Sale	Starter	22.2	5.2	5.736	5 / 4.088	4 / 6.870	1 / 0.856	2 / 1.522	all_players_stats_recent_season_blend
Dylan Dodd	Middle Reliever	3.9	1.0	0.993	1 / 0.678	1 / 1.022	0 / 0.223	1 / 0.369	all_players_stats_recent_season_blend
Robert Suarez	Middle Reliever (High Leverage)	4.0	1.0	0.993	2 / 0.817	1 / 0.599	0 / 0.179	0 / 0.177	all_players_stats_recent_season_blend
Dylan Lee	Middle Reliever (High Leverage)	4.0	2.1	0.993	1 / 0.744	2 / 1.044	1 / 0.247	0 / 0.362	all_players_stats_recent_season_blend

Audit: 2 CHECK(S) FAILED

Invariant	Verdict	Evidence
blend arithmetic away_runs	PASS	4.0000 + 0.78 x (3.3123 - 4.0000) = 3.4636, recorded 3.4636
blend arithmetic home_runs	PASS	3.0000 + 0.44 x (3.5040 - 3.0000) = 3.2218, recorded 3.2218
blend arithmetic home_win	PASS	0.4929 + 0.00 x (0.8254 - 0.4929) = 0.4929, recorded 0.4929
published rounding faithful to the blend	PASS	blended 3.4636-3.2218 rounds to 4-3, layer published 4-3
published probability on the score's side	PASS	4-3 with p(home) 0.492867
Log5 additivity136 checks	PASS	136 of 136 pass
away club fields exactly one starter	PASS	1 starters in a staff of 4
away staff totals a regulation workload	FAIL	30 outs recorded, 27 expected
away starter carries a continuous line	PASS	expected_IP 6.166
away starter batters faced inside the real band	PASS	Yoshinobu Yamamoto projected 23.2 batters faced; real starters run p01-p99 4-30, median 23, mean 21.75
away reliever batters faced inside the real band3 checks	PASS	3 of 3 pass
home club fields exactly one starter	PASS	1 starters in a staff of 4
home staff totals a regulation workload	FAIL	30 outs recorded, 27 expected
home starter carries a continuous line	PASS	expected_IP 5.736
home starter batters faced inside the real band	PASS	Chris Sale projected 22.2 batters faced; real starters run p01-p99 4-30, median 23, mean 21.75
home reliever batters faced inside the real band3 checks	PASS	3 of 3 pass
away hitters satisfy every coherence rule	PASS	9 hitters checked; H<=AB, HR<=H, 2B+3B+HR<=H, SO<=AB, continuous line present
home hitters satisfy every coherence rule	PASS	9 hitters checked; H<=AB, HR<=H, 2B+3B+HR<=H, SO<=AB, continuous line present

20261006_MIL@SD — published 4-5, 157 checks, all pass

Artifact. 20261006_MIL@SD__20261007T023916Z.json, phase post_start_rerun, generated 2026-10-07T02:39:16, first pitch 2026-10-07T01:30:00.
Published 4 - 5 with p(home) 0.566430. Simulation draw chosen on not recorded; ? of ? draws realized the target exactly.

Stage chain: every stage that touched the score

Stage	Score after	Δ away	Δ home	Cum. away	Cum. home	Changed target	Own adjustment and description
1. `historical_ml`	3.5313164191848228 - 4.700193372781561	0	0	0	0	no	Base ML historical model prediction
2. `variance_applied`	4 - 5	+0.4687	+0.2998	+0.4687	+0.2998	yes	Deterministic baseline target (variance factor 1.0)
3. `enhanced_prediction`	4 - 5	0	0	+0.4687	+0.2998	no	Enhanced ML with recent form, H2H, low-score analysis (confidence shrunk by projected_margin_is_one_run)
4. `kpi_layer`	4 - 5	0	0	+0.4687	+0.2998	no	away -0.3389; home -0.3442 Savant KPI blend applied to the model target
5. `embedded_savant_run_environment`	4 - 5	0	0	+0.4687	+0.2998	no	away +0.1075; home +0.1114 Target score adjusted by temporal Savant lineup quality versus opposing starter quality
6. `fundamental_edge_target_review`	4 - 5	0	0	+0.4687	+0.2998	no	Refreshed score target and win probability after Savant/pitcher/team-strength adjustments
7. `capped_target`	4 - 5	0	0	+0.4687	+0.2998	no	Capped at max 12 realistic runs
8. `simulation_result`	4 - 5	-	-	-	-	no	Final inning-by-inning simulation result

KPI team layer: the blend, with its arithmetic

Target	Baseline	KPI only	Weight	Blended	Computation	Verdict
`away_runs`	4.0000	3.5655	0.78	3.6611	4.0000 + 0.78 × (3.5655 − 4.0000) = 3.6611	PASS
`home_runs`	5.0000	4.2176	0.44	4.6558	5.0000 + 0.44 × (4.2176 − 5.0000) = 4.6558	PASS
`home_win`	0.5782	0.6792	0.00	0.5782	0.5782 + 0.00 × (0.6792 − 0.5782) = 0.5782	PASS

Published by the layer: away 4, home 5, p(home) 0.578153, continuous away 3.6611 / home 4.6558

Inputs and gates. candidate 20261005T125039Z_f5114dc0, value age 9 days, lineup coverage away 1.0000 / home 0.6842, movement cap 2.5 runs.
values_present ok values_fresh ok one_source_bundle ok lineup_coverage ok target_eligible ok movement_within_limit ok published_movement_within_limit ok

What drove the KPI forecast for `away_runs`

intercept 4.2573 + sum of contributions = 3.5655; prediction 3.5655; contributions outside the top listed -0.0666 runs

Feature	Raw	Standardised	Runs contributed	Which KPIs it is built from
`home_off_power_spread`	0.5852	-0.761	-0.4963	offence group (dispersion across the lineup, not its mean): batter_home_run_rate, batter_bat_speed
`away_off_contact`	0.3328	1.283	+0.4912	offence group: batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`away_off_qualified_share`	1.0000	2.120	+0.4497	share of the lineup carrying a qualifying sample, not a skill measurement
`away_pen_share`	0.5781	2.680	-0.4249	composite feature
`away_off_contact_spread`	0.9025	1.040	-0.3596	offence group (dispersion across the lineup, not its mean): batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`away_sp_command`	0.7166	1.113	-0.3550	starter group: pitcher_strikeout_rate, pitcher_walk_rate
`home_off_discipline_spread`	0.2163	-2.308	+0.3484	offence group (dispersion across the lineup, not its mean): batter_whiff_rate_vs_fastball, batter_whiff_rate_vs_breaking, batter_whiff_rate_vs_offspeed, batter_strikeout_rate ...
`home_off_top3_contact`	1.0774	1.040	-0.2936	mean of the three best contact bats, a tail a club mean cannot express
`home_off_qualified_share`	0.6842	-0.784	+0.2898	share of the lineup carrying a qualifying sample, not a skill measurement
`home_sp_velocity`	0.9849	1.338	-0.2748	starter group: pitcher_release_speed

What drove the KPI forecast for `home_runs`

intercept 4.2997 + sum of contributions = 4.2176; prediction 4.2176; contributions outside the top listed -0.4904 runs

Feature	Raw	Standardised	Runs contributed	Which KPIs it is built from
`away_off_ground`	-0.7382	-2.577	-0.6259	offence group: batter_ground_ball_rate_vs_slider, batter_ground_ball_rate_vs_fastball
`away_pen_share`	0.5781	2.680	+0.4215	composite feature
`home_sp_velocity`	0.9849	1.338	+0.4170	starter group: pitcher_release_speed
`away_off_top3_contact`	1.3386	1.919	-0.4168	mean of the three best contact bats, a tail a club mean cannot express
`home_off_worst3_discipline`	-0.1770	1.364	+0.4107	mean of the three weakest plate approaches
`away_off_contact_spread`	0.9025	1.040	-0.3375	offence group (dispersion across the lineup, not its mean): batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`home_off_power`	0.2755	1.020	+0.2711	offence group: batter_home_run_rate, batter_bat_speed
`away_pen_minus_sp`	0.5723	1.864	+0.2667	composite feature
`away_off_contact`	0.3328	1.283	+0.2082	offence group: batter_barrel_rate_vs_fastball, batter_barrel_rate_vs_breaking, batter_barrel_rate_vs_offspeed, batter_hard_hit_rate_vs_fastball ...
`home_off_ground`	-0.2120	-0.831	-0.2067	offence group: batter_ground_ball_rate_vs_slider, batter_ground_ball_rate_vs_fastball

Per-player metric movement

MIL against Chad Patrick (RHP)

The pipeline applied one club-wide ×0.9153 to every MIL hitter line, identical for the leadoff hitter and the ninth bat.
Opposing staff exposure. These hitters face the starter for 15.895 of 37.625 projected batters (42.2%) and the 4 projected relievers for the remaining 57.8%. Every metric below is the exposure-weighted blend of both matchups; the starter-only figure is shown beside it. Real 2026 game logs put a starter at 21.75 batters faced on average (median 23, p10-p90 17-27) and a whole staff at 37.9: this starter is OUTSIDE that band and this staff is inside its. The per-player columns are the measured alternative: each subject moved by his own KPI pairing. Exposure for a hitter is his projected plate appearances; for the starter it is batters faced.

Player	Pos	Metric	Published → KPI	Δ	League	Own rate	Opponent rate	Log5	Own skill	Opponent	Remainder
Jackson Chourio	LF	Hitswhole staff	0.971 → 0.918	-0.053	0.2169	0.2457n=567 Very good	0.2091n=140 Average	0.2372	+0.128	-0.034	-0.003 ok
		split	starter x0.9656 at 0.422 share (log5 0.2372 vs his 0.2091) + bullpen x0.9302 at 0.578 share (log5 0.2285 vs pen 0.2012) = x0.9452 \| starter alone would have been -0.033
		Strikeoutswhole staff	1.169 → 1.406	+0.237	0.2212	0.2351n=567 Average	0.2713n=140 Good	0.2872	+0.062	+0.222	+0.009 ok
		split	starter x1.2215 at 0.422 share (log5 0.2872 vs his 0.2713) + bullpen x1.1898 at 0.578 share (log5 0.2798 vs pen 0.2641) = x1.2032 \| starter alone would have been +0.259
		Walkswhole staff	0.342 → 0.316	-0.025	0.0889	0.0839n=567 Average	0.0740n=140 Good	0.0698	-0.022	-0.066	+0.003 ok
		split	starter x0.8311 at 0.422 share (log5 0.0698 vs his 0.0740) + bullpen x0.9952 at 0.578 share (log5 0.0835 vs pen 0.0885) = x0.9259 \| starter alone would have been -0.058
		Home runswhole staff	0.169 → 0.148	-0.021	0.0304	0.0397n=567 Good	0.0265n=140 Good	0.0348	+0.041	-0.017	-0.005 ok
		split	starter x0.8746 at 0.422 share (log5 0.0348 vs his 0.0265) + bullpen x0.8782 at 0.578 share (log5 0.0349 vs pen 0.0267) = x0.8767 \| starter alone would have been -0.021
Neutralising Jackson Chourio to the qualified-population mean moves the club's KPI run forecast by 0.2792 runs, and the weighted stage by 0.2178. Interactions make these non-additive across players, so they do not sum to the published movement.
Jake Bauers	1B	Hitswhole staff	0.786 → 0.741	-0.045	0.2169	0.2179n=567 Average	0.2091n=140 Average	0.2101	+0.004	-0.033	-0.000 ok
		split	starter x0.9644 at 0.422 share (log5 0.2101 vs his 0.2091) + bullpen x0.9278 at 0.578 share (log5 0.2022 vs pen 0.2012) = x0.9433 \| starter alone would have been -0.028
		Strikeoutswhole staff	1.181 → 1.408	+0.228	0.2212	0.2676n=567 Below average	0.2713n=140 Good	0.3239	+0.197	+0.212	+0.026 ok
		split	starter x1.2101 at 0.422 share (log5 0.3239 vs his 0.2713) + bullpen x1.1802 at 0.578 share (log5 0.3159 vs pen 0.2641) = x1.1928 \| starter alone would have been +0.248
		Walkswhole staff	0.572 → 0.532	-0.040	0.0889	0.1510n=567 Very good	0.0740n=140 Good	0.1271	+0.263	-0.063	-0.038 ok
		split	starter x0.8415 at 0.422 share (log5 0.1271 vs his 0.0740) + bullpen x0.9956 at 0.578 share (log5 0.1503 vs pen 0.0885) = x0.9305 \| starter alone would have been -0.091
		Home runswhole staff	0.150 → 0.132	-0.018	0.0304	0.0397n=567 Good	0.0265n=140 Good	0.0348	+0.040	-0.016	-0.005 ok
		split	starter x0.8746 at 0.422 share (log5 0.0348 vs his 0.0265) + bullpen x0.8782 at 0.578 share (log5 0.0349 vs pen 0.0267) = x0.8767 \| starter alone would have been -0.019
Neutralising Jake Bauers to the qualified-population mean moves the club's KPI run forecast by 0.2847 runs, and the weighted stage by 0.2221. Interactions make these non-additive across players, so they do not sum to the published movement.
William Contreras	C	Hitswhole staff	0.849 → 0.802	-0.048	0.2169	0.2303n=630 Good	0.2091n=140 Average	0.2222	+0.056	-0.032	-0.001 ok
		split	starter x0.9649 at 0.422 share (log5 0.2222 vs his 0.2091) + bullpen x0.9288 at 0.578 share (log5 0.2139 vs pen 0.2012) = x0.9441 \| starter alone would have been -0.030
		Strikeoutswhole staff	0.812 → 0.999	+0.187	0.2212	0.1529n=630 Good	0.2713n=140 Good	0.1913	-0.286	+0.209	-0.049 ok
		split	starter x1.2513 at 0.422 share (log5 0.1913 vs his 0.2713) + bullpen x1.2145 at 0.578 share (log5 0.1857 vs pen 0.2641) = x1.23 \| starter alone would have been +0.204
		Walkswhole staff	0.363 → 0.336	-0.027	0.0889	0.0942n=630 Average	0.0740n=140 Good	0.0785	+0.022	-0.062	-0.003 ok
		split	starter x0.8327 at 0.422 share (log5 0.0785 vs his 0.0740) + bullpen x0.9953 at 0.578 share (log5 0.0938 vs pen 0.0885) = x0.9266 \| starter alone would have been -0.061
		Home runswhole staff	0.107 → 0.094	-0.013	0.0304	0.0291n=630 Average	0.0265n=140 Good	0.0254	-0.005	-0.016	+0.001 ok
		split	starter x0.8734 at 0.422 share (log5 0.0254 vs his 0.0265) + bullpen x0.877 at 0.578 share (log5 0.0255 vs pen 0.0267) = x0.8755 \| starter alone would have been -0.014
Neutralising William Contreras to the qualified-population mean moves the club's KPI run forecast by 0.3181 runs, and the weighted stage by 0.2481. Interactions make these non-additive across players, so they do not sum to the published movement.
Brice Turang	2B	Hitswhole staff	0.797 → 0.752	-0.045	0.2169	0.2264n=666 Good	0.2091n=140 Average	0.2184	+0.040	-0.032	-0.001 ok
		split	starter x0.9647 at 0.422 share (log5 0.2184 vs his 0.2091) + bullpen x0.9285 at 0.578 share (log5 0.2102 vs pen 0.2012) = x0.9438 \| starter alone would have been -0.028
		Strikeoutswhole staff	1.118 → 1.341	+0.223	0.2212	0.2473n=666 Below average	0.2713n=140 Good	0.3011	+0.108	+0.208	+0.015 ok
		split	starter x1.2172 at 0.422 share (log5 0.3011 vs his 0.2713) + bullpen x1.1862 at 0.578 share (log5 0.2934 vs pen 0.2641) = x1.1993 \| starter alone would have been +0.243
		Walkswhole staff	0.492 → 0.457	-0.035	0.0889	0.1243n=666 Very good	0.0740n=140 Good	0.1041	+0.147	-0.062	-0.022 ok
		split	starter x0.8374 at 0.422 share (log5 0.1041 vs his 0.0740) + bullpen x0.9954 at 0.578 share (log5 0.1237 vs pen 0.0885) = x0.9286 \| starter alone would have been -0.080
		Home runswhole staff	0.113 → 0.099	-0.014	0.0304	0.0312n=666 Average	0.0265n=140 Good	0.0273	+0.003	-0.016	-0.000 ok
		split	starter x0.8737 at 0.422 share (log5 0.0273 vs his 0.0265) + bullpen x0.8772 at 0.578 share (log5 0.0274 vs pen 0.0267) = x0.8757 \| starter alone would have been -0.014
Neutralising Brice Turang to the qualified-population mean moves the club's KPI run forecast by 0.4175 runs, and the weighted stage by 0.3256. Interactions make these non-additive across players, so they do not sum to the published movement.
Christian Yelich	DH	Hitswhole staff	0.799 → 0.753	-0.046	0.2169	0.2108n=487 Average	0.2091n=140 Average	0.2032	-0.025	-0.031	+0.001 ok
		split	starter x0.964 at 0.422 share (log5 0.2032 vs his 0.2091) + bullpen x0.9272 at 0.578 share (log5 0.1955 vs pen 0.2012) = x0.9427 \| starter alone would have been -0.029
		Strikeoutswhole staff	1.116 → 1.331	+0.215	0.2212	0.2684n=487 Below average	0.2713n=140 Good	0.3247	+0.192	+0.203	+0.025 ok
		split	starter x1.2099 at 0.422 share (log5 0.3247 vs his 0.2713) + bullpen x1.18 at 0.578 share (log5 0.3167 vs pen 0.2641) = x1.1926 \| starter alone would have been +0.234
		Walkswhole staff	0.395 → 0.366	-0.029	0.0889	0.0983n=487 Average	0.0740n=140 Good	0.0819	+0.038	-0.061	-0.006 ok
		split	starter x0.8333 at 0.422 share (log5 0.0819 vs his 0.0740) + bullpen x0.9953 at 0.578 share (log5 0.0979 vs pen 0.0885) = x0.9269 \| starter alone would have been -0.066
		Home runswhole staff	0.116 → 0.101	-0.015	0.0304	0.0267n=487 Average	0.0265n=140 Good	0.0233	-0.015	-0.016	+0.002 ok
		split	starter x0.8731 at 0.422 share (log5 0.0233 vs his 0.0265) + bullpen x0.8767 at 0.578 share (log5 0.0234 vs pen 0.0267) = x0.8752 \| starter alone would have been -0.015
Neutralising Christian Yelich to the qualified-population mean moves the club's KPI run forecast by 0.1820 runs, and the weighted stage by 0.1420. Interactions make these non-additive across players, so they do not sum to the published movement.
Garrett Mitchell	CF	Hitswhole staff	0.761 → 0.718	-0.043	0.2169	0.2215n=519 Average	0.2091n=140 Average	0.2136	+0.018	-0.030	-0.001 ok
		split	starter x0.9645 at 0.422 share (log5 0.2136 vs his 0.2091) + bullpen x0.9281 at 0.578 share (log5 0.2056 vs pen 0.2012) = x0.9435 \| starter alone would have been -0.027
		Strikeoutswhole staff	1.296 → 1.531	+0.234	0.2212	0.3061n=519 Poor	0.2713n=140 Good	0.3664	+0.330	+0.195	+0.040 ok
		split	starter x1.1969 at 0.422 share (log5 0.3664 vs his 0.2713) + bullpen x1.1692 at 0.578 share (log5 0.3579 vs pen 0.2641) = x1.1809 \| starter alone would have been +0.255
		Walkswhole staff	0.395 → 0.366	-0.029	0.0889	0.1091n=519 Good	0.0740n=140 Good	0.0911	+0.078	-0.058	-0.012 ok
		split	starter x0.835 at 0.422 share (log5 0.0911 vs his 0.0740) + bullpen x0.9953 at 0.578 share (log5 0.1086 vs pen 0.0885) = x0.9276 \| starter alone would have been -0.065
		Home runswhole staff	0.083 → 0.073	-0.010	0.0304	0.0243n=519 Below average	0.0265n=140 Good	0.0212	-0.024	-0.015	+0.003 ok
		split	starter x0.8729 at 0.422 share (log5 0.0212 vs his 0.0265) + bullpen x0.8765 at 0.578 share (log5 0.0213 vs pen 0.0267) = x0.875 \| starter alone would have been -0.011
Neutralising Garrett Mitchell to the qualified-population mean moves the club's KPI run forecast by -0.0305 runs, and the weighted stage by -0.0238. Interactions make these non-additive across players, so they do not sum to the published movement.
Cooper Pratt	SS	Hitswhole staff	0.751 → 0.709	-0.042	0.2169	0.2255n=258 Good	0.2091n=140 Average	0.2176	+0.032	-0.029	-0.001 ok
		split	starter x0.9647 at 0.422 share (log5 0.2176 vs his 0.2091) + bullpen x0.9284 at 0.578 share (log5 0.2094 vs pen 0.2012) = x0.9437 \| starter alone would have been -0.026
		Strikeoutswhole staff	0.799 → 0.974	+0.175	0.2212	0.1851n=258 Good	0.2713n=140 Good	0.2294	-0.134	+0.186	-0.021 ok
		split	starter x1.2395 at 0.422 share (log5 0.2294 vs his 0.2713) + bullpen x1.2047 at 0.578 share (log5 0.2230 vs pen 0.2641) = x1.2194 \| starter alone would have been +0.191
		Walkswhole staff	0.311 → 0.288	-0.023	0.0889	0.0917n=258 Average	0.0740n=140 Good	0.0763	+0.010	-0.056	-0.002 ok
		split	starter x0.8323 at 0.422 share (log5 0.0763 vs his 0.0740) + bullpen x0.9953 at 0.578 share (log5 0.0913 vs pen 0.0885) = x0.9264 \| starter alone would have been -0.052
		Home runswhole staff	0.072 → 0.063	-0.009	0.0304	0.0211n=258 Below average	0.0265n=140 Good	0.0184	-0.034	-0.014	+0.004 ok
		split	starter x0.8725 at 0.422 share (log5 0.0184 vs his 0.0265) + bullpen x0.8761 at 0.578 share (log5 0.0185 vs pen 0.0267) = x0.8746 \| starter alone would have been -0.009
Neutralising Cooper Pratt to the qualified-population mean moves the club's KPI run forecast by 0.1306 runs, and the weighted stage by 0.1018. Interactions make these non-additive across players, so they do not sum to the published movement.
Sal Frelick	RF	Hitswhole staff	0.723 → 0.682	-0.042	0.2169	0.2106n=374 Below average	0.2091n=140 Average	0.2030	-0.023	-0.028	+0.001 ok
		split	starter x0.964 at 0.422 share (log5 0.2030 vs his 0.2091) + bullpen x0.9271 at 0.578 share (log5 0.1952 vs pen 0.2012) = x0.9427 \| starter alone would have been -0.026
		Strikeoutswhole staff	0.484 → 0.603	+0.118	0.2212	0.1106n=374 Very good	0.2713n=140 Good	0.1402	-0.401	+0.182	-0.074 ok
		split	starter x1.2672 at 0.422 share (log5 0.1402 vs his 0.2713) + bullpen x1.2277 at 0.578 share (log5 0.1358 vs pen 0.2641) = x1.2444 \| starter alone would have been +0.129
		Walkswhole staff	0.301 → 0.278	-0.022	0.0889	0.0843n=374 Average	0.0740n=140 Good	0.0701	-0.017	-0.054	+0.003 ok
		split	starter x0.8312 at 0.422 share (log5 0.0701 vs his 0.0740) + bullpen x0.9952 at 0.578 share (log5 0.0839 vs pen 0.0885) = x0.9259 \| starter alone would have been -0.051
		Home runswhole staff	0.063 → 0.055	-0.008	0.0304	0.0173n=374 Poor	0.0265n=140 Good	0.0151	-0.048	-0.014	+0.006 ok
		split	starter x0.8721 at 0.422 share (log5 0.0151 vs his 0.0265) + bullpen x0.8757 at 0.578 share (log5 0.0151 vs pen 0.0267) = x0.8742 \| starter alone would have been -0.008
Neutralising Sal Frelick to the qualified-population mean moves the club's KPI run forecast by 0.1252 runs, and the weighted stage by 0.0977. Interactions make these non-additive across players, so they do not sum to the published movement.
Joey Ortiz	3B	Hitswhole staff	0.753 → 0.711	-0.042	0.2169	0.2271n=403 Good	0.2091n=140 Average	0.2191	+0.038	-0.029	-0.001 ok
		split	starter x0.9648 at 0.422 share (log5 0.2191 vs his 0.2091) + bullpen x0.9285 at 0.578 share (log5 0.2108 vs pen 0.2012) = x0.9438 \| starter alone would have been -0.026
		Strikeoutswhole staff	0.875 → 1.062	+0.187	0.2212	0.2010n=403 Average	0.2713n=140 Good	0.2479	-0.075	+0.184	-0.011 ok
		split	starter x1.2337 at 0.422 share (log5 0.2479 vs his 0.2713) + bullpen x1.1999 at 0.578 share (log5 0.2411 vs pen 0.2641) = x1.2142 \| starter alone would have been +0.204
		Walkswhole staff	0.290 → 0.269	-0.021	0.0889	0.0873n=403 Average	0.0740n=140 Good	0.0726	-0.006	-0.055	+0.001 ok
		split	starter x0.8316 at 0.422 share (log5 0.0726 vs his 0.0740) + bullpen x0.9952 at 0.578 share (log5 0.0869 vs pen 0.0885) = x0.9261 \| starter alone would have been -0.049
		Home runswhole staff	0.087 → 0.076	-0.011	0.0304	0.0255n=403 Below average	0.0265n=140 Good	0.0223	-0.018	-0.014	+0.002 ok
		split	starter x0.873 at 0.422 share (log5 0.0223 vs his 0.0265) + bullpen x0.8766 at 0.578 share (log5 0.0224 vs pen 0.0267) = x0.8751 \| starter alone would have been -0.011
Neutralising Joey Ortiz to the qualified-population mean moves the club's KPI run forecast by 0.0909 runs, and the weighted stage by 0.0709. Interactions make these non-additive across players, so they do not sum to the published movement.
STARTER Dustin May17.0326 BF	Starter	Hits	3.791 → 3.842	+0.051	0.2169	0.2249n=621 Below average	0.21986/9 qualified lineup mean	0.2279	+0.136	+0.051	+0.001 ok
		Strikeouts	4.456 → 4.316	-0.140	0.2212	0.2307n=621 Average	0.21426/9 qualified lineup mean	0.2234	+0.161	-0.120	-0.004 ok
		Walks	1.224 → 1.250	+0.027	0.0889	0.0822n=621 Average	0.09086/9 qualified lineup mean	0.0840	-0.114	+0.033	-0.002 ok
		Home runs	0.627 → 0.663	+0.036	0.0304	0.0258n=621 Good	0.03216/9 qualified lineup mean	0.0273	-0.078	+0.029	-0.004 ok
RELIEVER DL Hall4.3644 BF	Reliever	Hits	0.759 → 0.770	+0.011	0.2169	0.1738n=234 Very good	0.21986/9 qualified lineup mean	0.1763	-0.188	+0.013	-0.002 ok
		Strikeouts	1.108 → 1.074	-0.034	0.2212	0.2538n=234 Good	0.21426/9 qualified lineup mean	0.2461	+0.142	-0.031	-0.003 ok
		Walks	0.535 → 0.546	+0.011	0.0889	0.1226n=234 Poor	0.09086/9 qualified lineup mean	0.1252	+0.147	+0.008	+0.003 ok
		Home runs	0.119 → 0.126	+0.007	0.0304	0.0273n=234 Average	0.03216/9 qualified lineup mean	0.0288	-0.013	+0.007	-0.001 ok
RELIEVER JoJo Romero3.9668 BF	Long Reliever	Hits	0.723 → 0.733	+0.010	0.2169	0.2159n=233 Average	0.21986/9 qualified lineup mean	0.2188	-0.004	+0.012	0 ok
		Strikeouts	1.045 → 1.012	-0.033	0.2212	0.2259n=233 Average	0.21426/9 qualified lineup mean	0.2187	+0.018	-0.028	-0.000 ok
		Walks	0.280 → 0.286	+0.006	0.0889	0.0794n=233 Average	0.09086/9 qualified lineup mean	0.0811	-0.038	+0.008	-0.001 ok
		Home runs	0.088 → 0.093	+0.005	0.0304	0.0303n=233 Average	0.03216/9 qualified lineup mean	0.0320	-0.001	+0.007	0 ok
RELIEVER Antonio Senzatela3.7088 BF	Middle Reliever (High Leverage)	Hits	1.050 → 1.064	+0.014	0.2169	0.2274n=303 Below average	0.21986/9 qualified lineup mean	0.2305	+0.039	+0.011	+0.000 ok
		Strikeouts	0.737 → 0.713	-0.024	0.2212	0.1977n=303 Below average	0.21426/9 qualified lineup mean	0.1912	-0.087	-0.026	+0.002 ok
		Walks	0.399 → 0.408	+0.009	0.0889	0.0847n=303 Average	0.09086/9 qualified lineup mean	0.0865	-0.016	+0.007	-0.000 ok
		Home runs	0.058 → 0.061	+0.003	0.0304	0.0261n=303 Good	0.03216/9 qualified lineup mean	0.0276	-0.016	+0.006	-0.001 ok
RELIEVER Trevor Megill2.691 BF	Reliever	Hits	0.521 → 0.528	+0.007	0.2169	0.1936n=242 Very good	0.21986/9 qualified lineup mean	0.1963	-0.063	+0.008	-0.001 ok
		Strikeouts	0.725 → 0.703	-0.022	0.2212	0.2692n=242 Good	0.21426/9 qualified lineup mean	0.2612	+0.129	-0.019	-0.003 ok
		Walks	0.195 → 0.200	+0.004	0.0889	0.0726n=242 Good	0.09086/9 qualified lineup mean	0.0742	-0.044	+0.005	-0.001 ok
		Home runs	0.057 → 0.060	+0.003	0.0304	0.0212n=242 Very good	0.03216/9 qualified lineup mean	0.0224	-0.025	+0.005	-0.001 ok
RELIEVER Aaron Ashby2.5017 BF	Reliever	Hits	0.506 → 0.513	+0.007	0.2169	0.2024n=372 Good	0.21986/9 qualified lineup mean	0.2052	-0.036	+0.007	-0.000 ok
		Strikeouts	0.718 → 0.697	-0.021	0.2212	0.2869n=372 Very good	0.21426/9 qualified lineup mean	0.2785	+0.164	-0.018	-0.003 ok
		Walks	0.260 → 0.265	+0.005	0.0889	0.1039n=372 Below average	0.09086/9 qualified lineup mean	0.1061	+0.037	+0.005	+0.001 ok
		Home runs	0.063 → 0.067	+0.004	0.0304	0.0251n=372 Good	0.03216/9 qualified lineup mean	0.0266	-0.013	+0.004	-0.001 ok
RELIEVER Abner Uribe1.9272 BF	Reliever	Hits	0.389 → 0.394	+0.005	0.2169	0.2018n=230 Good	0.21986/9 qualified lineup mean	0.2047	-0.029	+0.006	-0.000 ok
		Strikeouts	0.489 → 0.474	-0.015	0.2212	0.2538n=230 Good	0.21426/9 qualified lineup mean	0.2461	+0.063	-0.013	-0.001 ok
		Walks	0.159 → 0.163	+0.004	0.0889	0.0826n=230 Average	0.09086/9 qualified lineup mean	0.0844	-0.012	+0.004	-0.000 ok
		Home runs	0.047 → 0.050	+0.003	0.0304	0.0245n=230 Very good	0.03216/9 qualified lineup mean	0.0259	-0.011	+0.003	-0.001 ok
RELIEVER Chad Patrick1.6525 BF	Reliever	Hits	0.329 → 0.334	+0.005	0.2169	0.1991n=438 Good	0.21986/9 qualified lineup mean	0.2019	-0.029	+0.005	-0.000 ok
		Strikeouts	0.353 → 0.342	-0.011	0.2212	0.2137n=438 Average	0.21426/9 qualified lineup mean	0.2069	-0.012	-0.012	+0.000 ok
		Walks	0.172 → 0.176	+0.004	0.0889	0.1041n=438 Below average	0.09086/9 qualified lineup mean	0.1063	+0.025	+0.003	+0.001 ok
		Home runs	0.046 → 0.049	+0.003	0.0304	0.0278n=438 Average	0.03216/9 qualified lineup mean	0.0294	-0.004	+0.003	-0.000 ok

SD against Joe Musgrove (RHP)

The pipeline applied one club-wide ×0.9312 to every SD hitter line, identical for the leadoff hitter and the ninth bat.
Opposing staff exposure. These hitters face the starter for 17.0326 of 37.845 projected batters (45.0%) and the 2 projected relievers for the remaining 55.0%. Every metric below is the exposure-weighted blend of both matchups; the starter-only figure is shown beside it. Real 2026 game logs put a starter at 21.75 batters faced on average (median 23, p10-p90 17-27) and a whole staff at 37.9: this starter is inside that band and this staff is inside its. The per-player columns are the measured alternative: each subject moved by his own KPI pairing. Exposure for a hitter is his projected plate appearances; for the starter it is batters faced.

Player	Pos	Metric	Published → KPI	Δ	League	Own rate	Opponent rate	Log5	Own skill	Opponent	Remainder
Fernando Tatis Jr.	RF	Hitswhole staff	0.993 → 0.972	-0.021	0.2169	0.2416n=704 Very good	0.2249n=621 Below average	0.2502	+0.109	+0.035	+0.003 ok
		split	starter x1.0357 at 0.450 share (log5 0.2502 vs his 0.2249) + bullpen x0.9333 at 0.550 share (log5 0.2255 vs pen 0.2020) = x0.9794 \| starter alone would have been +0.035
		Strikeoutswhole staff	0.893 → 0.958	+0.065	0.2212	0.1870n=704 Good	0.2307n=621 Average	0.1954	-0.151	+0.042	-0.005 ok
		split	starter x1.0447 at 0.450 share (log5 0.1954 vs his 0.2307) + bullpen x1.095 at 0.550 share (log5 0.2048 vs pen 0.2413) = x1.0724 \| starter alone would have been +0.040
		Walkswhole staff	0.382 → 0.380	-0.002	0.0889	0.0906n=704 Average	0.0822n=621 Average	0.0838	+0.007	-0.030	-0.001 ok
		split	starter x0.9247 at 0.450 share (log5 0.0838 vs his 0.0822) + bullpen x1.0539 at 0.550 share (log5 0.0955 vs pen 0.0937) = x0.9957 \| starter alone would have been -0.029
		Home runswhole staff	0.141 → 0.122	-0.020	0.0304	0.0341n=704 Average	0.0258n=621 Good	0.0290	+0.016	-0.020	-0.002 ok
		split	starter x0.8504 at 0.450 share (log5 0.0290 vs his 0.0258) + bullpen x0.8684 at 0.550 share (log5 0.0296 vs pen 0.0264) = x0.8603 \| starter alone would have been -0.021
Neutralising Fernando Tatis Jr. to the qualified-population mean moves the club's KPI run forecast by 0.0920 runs, and the weighted stage by 0.0405. Interactions make these non-additive across players, so they do not sum to the published movement.
Jackson Merrill	CF	Hitswhole staff	0.892 → 0.874	-0.019	0.2169	0.2277n=664 Good	0.2249n=621 Below average	0.2360	+0.047	+0.034	+0.001 ok
		split	starter x1.0364 at 0.450 share (log5 0.2360 vs his 0.2249) + bullpen x0.9322 at 0.550 share (log5 0.2123 vs pen 0.2020) = x0.9791 \| starter alone would have been +0.032
		Strikeoutswhole staff	1.111 → 1.186	+0.075	0.2212	0.2396n=664 Average	0.2307n=621 Average	0.2496	+0.079	+0.041	+0.002 ok
		split	starter x1.0417 at 0.450 share (log5 0.2496 vs his 0.2307) + bullpen x1.0883 at 0.550 share (log5 0.2608 vs pen 0.2413) = x1.0673 \| starter alone would have been +0.046
		Walkswhole staff	0.305 → 0.304	-0.001	0.0889	0.0659n=664 Below average	0.0822n=621 Average	0.0608	-0.099	-0.029	+0.007 ok
		split	starter x0.9228 at 0.450 share (log5 0.0608 vs his 0.0822) + bullpen x1.0555 at 0.550 share (log5 0.0695 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.024
		Home runswhole staff	0.142 → 0.122	-0.020	0.0304	0.0367n=664 Good	0.0258n=621 Good	0.0312	+0.027	-0.020	-0.004 ok
		split	starter x0.8507 at 0.450 share (log5 0.0312 vs his 0.0258) + bullpen x0.8687 at 0.550 share (log5 0.0318 vs pen 0.0264) = x0.8606 \| starter alone would have been -0.021
Neutralising Jackson Merrill to the qualified-population mean moves the club's KPI run forecast by 0.3190 runs, and the weighted stage by 0.1403. Interactions make these non-additive across players, so they do not sum to the published movement.
Manny Machado	3B	Hitswhole staff	0.818 → 0.800	-0.018	0.2169	0.2000n=677 Poor	0.2249n=621 Below average	0.2076	-0.072	+0.034	-0.002 ok
		split	starter x1.0377 at 0.450 share (log5 0.2076 vs his 0.2249) + bullpen x0.9299 at 0.550 share (log5 0.1860 vs pen 0.2020) = x0.9784 \| starter alone would have been +0.031
		Strikeoutswhole staff	0.906 → 0.969	+0.063	0.2212	0.2133n=677 Average	0.2307n=621 Average	0.2225	-0.034	+0.040	-0.001 ok
		split	starter x1.0432 at 0.450 share (log5 0.2225 vs his 0.2307) + bullpen x1.0917 at 0.550 share (log5 0.2328 vs pen 0.2413) = x1.0699 \| starter alone would have been +0.039
		Walkswhole staff	0.455 → 0.453	-0.002	0.0889	0.1088n=677 Good	0.0822n=621 Average	0.1007	+0.084	-0.029	-0.006 ok
		split	starter x0.9261 at 0.450 share (log5 0.1007 vs his 0.0822) + bullpen x1.0528 at 0.550 share (log5 0.1145 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.034
		Home runswhole staff	0.161 → 0.139	-0.022	0.0304	0.0404n=677 Good	0.0258n=621 Good	0.0344	+0.043	-0.019	-0.006 ok
		split	starter x0.8512 at 0.450 share (log5 0.0344 vs his 0.0258) + bullpen x0.8691 at 0.550 share (log5 0.0351 vs pen 0.0264) = x0.861 \| starter alone would have been -0.024
Neutralising Manny Machado to the qualified-population mean moves the club's KPI run forecast by 0.1326 runs, and the weighted stage by 0.0583. Interactions make these non-additive across players, so they do not sum to the published movement.
Ty France	1B	Hitswhole staff	0.987 → 0.967	-0.020	0.2169	0.2398n=500 Very good	0.2249n=621 Below average	0.2484	+0.095	+0.033	+0.002 ok
		split	starter x1.0358 at 0.450 share (log5 0.2484 vs his 0.2249) + bullpen x0.9332 at 0.550 share (log5 0.2238 vs pen 0.2020) = x0.9794 \| starter alone would have been +0.035
		Strikeoutswhole staff	0.963 → 1.030	+0.067	0.2212	0.2165n=500 Average	0.2307n=621 Average	0.2258	-0.020	+0.039	-0.001 ok
		split	starter x1.043 at 0.450 share (log5 0.2258 vs his 0.2307) + bullpen x1.0913 at 0.550 share (log5 0.2362 vs pen 0.2413) = x1.0696 \| starter alone would have been +0.041
		Walkswhole staff	0.310 → 0.309	-0.001	0.0889	0.0769n=500 Below average	0.0822n=621 Average	0.0710	-0.050	-0.028	+0.004 ok
		split	starter x0.9237 at 0.450 share (log5 0.0710 vs his 0.0822) + bullpen x1.0548 at 0.550 share (log5 0.0811 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.024
		Home runswhole staff	0.159 → 0.137	-0.022	0.0304	0.0380n=500 Good	0.0258n=621 Good	0.0323	+0.032	-0.019	-0.005 ok
		split	starter x0.8509 at 0.450 share (log5 0.0323 vs his 0.0258) + bullpen x0.8688 at 0.550 share (log5 0.0330 vs pen 0.0264) = x0.8607 \| starter alone would have been -0.024
Neutralising Ty France to the qualified-population mean moves the club's KPI run forecast by 0.0936 runs, and the weighted stage by 0.0412. Interactions make these non-additive across players, so they do not sum to the published movement.
Ethan Salas	C	Hitswhole staff	0.742 → 0.726	-0.016	0.2169	0.2169n=0 averagebelow floor	0.2249n=621 Below average	0.2249	0	+0.028	0 ok
		split	starter x1.0369 at 0.450 share (log5 0.2249 vs his 0.2249) + bullpen x0.9313 at 0.550 share (log5 0.2020 vs pen 0.2020) = x0.9788 \| starter alone would have been +0.027
		Strikeoutswhole staff	0.841 → 0.899	+0.058	0.2212	0.2212n=0 averagebelow floor	0.2307n=621 Average	0.2307	0	+0.033	0 ok
		split	starter x1.0427 at 0.450 share (log5 0.2307 vs his 0.2307) + bullpen x1.0907 at 0.550 share (log5 0.2413 vs pen 0.2413) = x1.0691 \| starter alone would have been +0.036
		Walkswhole staff	0.262 → 0.261	-0.001	0.0889	0.0889n=0 averagebelow floor	0.0822n=621 Average	0.0822	0	-0.023	0 ok
		split	starter x0.9246 at 0.450 share (log5 0.0822 vs his 0.0822) + bullpen x1.054 at 0.550 share (log5 0.0937 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.020
		Home runswhole staff	0.110 → 0.095	-0.015	0.0304	0.0304n=0 averagebelow floor	0.0258n=621 Good	0.0258	0	-0.016	0 ok
		split	starter x0.8499 at 0.450 share (log5 0.0258 vs his 0.0258) + bullpen x0.8679 at 0.550 share (log5 0.0264 vs pen 0.0264) = x0.8598 \| starter alone would have been -0.017
Neutralising Ethan Salas to the qualified-population mean moves the club's KPI run forecast by 0.0000 runs, and the weighted stage by 0.0000. Interactions make these non-additive across players, so they do not sum to the published movement.
Xander Bogaerts	SS	Hitswhole staff	0.742 → 0.726	-0.016	0.2169	0.2059n=606 Below average	0.2249n=621 Below average	0.2137	-0.043	+0.031	-0.001 ok
		split	starter x1.0374 at 0.450 share (log5 0.2137 vs his 0.2249) + bullpen x0.9304 at 0.550 share (log5 0.1916 vs pen 0.2020) = x0.9786 \| starter alone would have been +0.028
		Strikeoutswhole staff	0.724 → 0.777	+0.053	0.2212	0.1829n=606 Good	0.2307n=621 Average	0.1912	-0.149	+0.037	-0.005 ok
		split	starter x1.0449 at 0.450 share (log5 0.1912 vs his 0.2307) + bullpen x1.0955 at 0.550 share (log5 0.2004 vs pen 0.2413) = x1.0727 \| starter alone would have been +0.033
		Walkswhole staff	0.448 → 0.446	-0.002	0.0889	0.1208n=606 Very good	0.0822n=621 Average	0.1120	+0.124	-0.026	-0.008 ok
		split	starter x0.927 at 0.450 share (log5 0.1120 vs his 0.0822) + bullpen x1.0521 at 0.550 share (log5 0.1271 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.033
		Home runswhole staff	0.098 → 0.084	-0.014	0.0304	0.0242n=606 Below average	0.0258n=621 Good	0.0205	-0.024	-0.018	+0.004 ok
		split	starter x0.8491 at 0.450 share (log5 0.0205 vs his 0.0258) + bullpen x0.8672 at 0.550 share (log5 0.0210 vs pen 0.0264) = x0.8591 \| starter alone would have been -0.015
Neutralising Xander Bogaerts to the qualified-population mean moves the club's KPI run forecast by -0.0141 runs, and the weighted stage by -0.0062. Interactions make these non-additive across players, so they do not sum to the published movement.
Dustin Harris	LF	Hitswhole staff	0.858 → 0.840	-0.018	0.2169	0.2169n=0 averagebelow floor	0.2249n=621 Below average	0.2249	0	+0.032	0 ok
		split	starter x1.0369 at 0.450 share (log5 0.2249 vs his 0.2249) + bullpen x0.9313 at 0.550 share (log5 0.2020 vs pen 0.2020) = x0.9788 \| starter alone would have been +0.032
		Strikeoutswhole staff	0.629 → 0.673	+0.043	0.2212	0.2212n=0 averagebelow floor	0.2307n=621 Average	0.2307	0	+0.037	0 ok
		split	starter x1.0427 at 0.450 share (log5 0.2307 vs his 0.2307) + bullpen x1.0907 at 0.550 share (log5 0.2413 vs pen 0.2413) = x1.0691 \| starter alone would have been +0.027
		Walkswhole staff	0.310 → 0.309	-0.001	0.0889	0.0889n=0 averagebelow floor	0.0822n=621 Average	0.0822	0	-0.027	0 ok
		split	starter x0.9246 at 0.450 share (log5 0.0822 vs his 0.0822) + bullpen x1.054 at 0.550 share (log5 0.0937 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.023
		Home runswhole staff	0.072 → 0.062	-0.010	0.0304	0.0304n=0 averagebelow floor	0.0258n=621 Good	0.0258	0	-0.018	0 ok
		split	starter x0.8499 at 0.450 share (log5 0.0258 vs his 0.0258) + bullpen x0.8679 at 0.550 share (log5 0.0264 vs pen 0.0264) = x0.8598 \| starter alone would have been -0.011
Neutralising Dustin Harris to the qualified-population mean moves the club's KPI run forecast by 0.0197 runs, and the weighted stage by 0.0087. Interactions make these non-additive across players, so they do not sum to the published movement.
Samad Taylor	DH	Hitswhole staff	0.788 → 0.772	-0.017	0.2169	0.2169n=0 averagebelow floor	0.2249n=621 Below average	0.2249	0	+0.029	0 ok
		split	starter x1.0369 at 0.450 share (log5 0.2249 vs his 0.2249) + bullpen x0.9313 at 0.550 share (log5 0.2020 vs pen 0.2020) = x0.9788 \| starter alone would have been +0.029
		Strikeoutswhole staff	0.960 → 1.026	+0.066	0.2212	0.2212n=0 averagebelow floor	0.2307n=621 Average	0.2307	0	+0.034	0 ok
		split	starter x1.0427 at 0.450 share (log5 0.2307 vs his 0.2307) + bullpen x1.0907 at 0.550 share (log5 0.2413 vs pen 0.2413) = x1.0691 \| starter alone would have been +0.041
		Walkswhole staff	0.337 → 0.336	-0.001	0.0889	0.0889n=0 averagebelow floor	0.0822n=621 Average	0.0822	0	-0.024	0 ok
		split	starter x0.9246 at 0.450 share (log5 0.0822 vs his 0.0822) + bullpen x1.054 at 0.550 share (log5 0.0937 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.025
		Home runswhole staff	0.076 → 0.065	-0.011	0.0304	0.0304n=0 averagebelow floor	0.0258n=621 Good	0.0258	0	-0.016	0 ok
		split	starter x0.8499 at 0.450 share (log5 0.0258 vs his 0.0258) + bullpen x0.8679 at 0.550 share (log5 0.0264 vs pen 0.0264) = x0.8598 \| starter alone would have been -0.011
Neutralising Samad Taylor to the qualified-population mean moves the club's KPI run forecast by -0.0095 runs, and the weighted stage by -0.0042. Interactions make these non-additive across players, so they do not sum to the published movement.
Jake Cronenworth	2B	Hitswhole staff	0.728 → 0.712	-0.016	0.2169	0.2086n=425 Below average	0.2249n=621 Below average	0.2164	-0.030	+0.029	-0.001 ok
		split	starter x1.0373 at 0.450 share (log5 0.2164 vs his 0.2249) + bullpen x0.9306 at 0.550 share (log5 0.1941 vs pen 0.2020) = x0.9786 \| starter alone would have been +0.027
		Strikeoutswhole staff	0.831 → 0.887	+0.057	0.2212	0.2275n=425 Average	0.2307n=621 Average	0.2372	+0.023	+0.035	+0.001 ok
		split	starter x1.0424 at 0.450 share (log5 0.2372 vs his 0.2307) + bullpen x1.0899 at 0.550 share (log5 0.2480 vs pen 0.2413) = x1.0685 \| starter alone would have been +0.035
		Walkswhole staff	0.338 → 0.336	-0.001	0.0889	0.0893n=425 Average	0.0822n=621 Average	0.0826	+0.001	-0.025	-0.000 ok
		split	starter x0.9246 at 0.450 share (log5 0.0826 vs his 0.0822) + bullpen x1.054 at 0.550 share (log5 0.0941 vs pen 0.0937) = x0.9958 \| starter alone would have been -0.025
		Home runswhole staff	0.082 → 0.070	-0.011	0.0304	0.0218n=425 Below average	0.0258n=621 Good	0.0185	-0.032	-0.017	+0.005 ok
		split	starter x0.8487 at 0.450 share (log5 0.0185 vs his 0.0258) + bullpen x0.8669 at 0.550 share (log5 0.0189 vs pen 0.0264) = x0.8587 \| starter alone would have been -0.012
Neutralising Jake Cronenworth to the qualified-population mean moves the club's KPI run forecast by 0.1040 runs, and the weighted stage by 0.0457. Interactions make these non-additive across players, so they do not sum to the published movement.
STARTER Nick Pivetta15.895 BF	Starter	Hits	3.266 → 3.379	+0.113	0.2169	0.2091n=140 Average	0.22439/9 qualified lineup mean	0.2164	-0.123	+0.118	-0.003 ok
		Strikeouts	4.260 → 4.256	-0.004	0.2212	0.2713n=140 Good	0.22109/9 qualified lineup mean	0.2711	+0.796	-0.003	-0.001 ok
		Walks	0.905 → 1.053	+0.148	0.0889	0.0740n=140 Good	0.10329/9 qualified lineup mean	0.0861	-0.238	+0.227	-0.035 ok
		Home runs	0.297 → 0.281	-0.016	0.0304	0.0265n=140 Good	0.02879/9 qualified lineup mean	0.0251	-0.061	-0.026	+0.003 ok
RELIEVER Jason Adam5.2739 BF	Setup Man	Hits	0.752 → 0.778	+0.026	0.2169	0.2150n=142 Average	0.22439/9 qualified lineup mean	0.2225	-0.010	+0.039	-0.000 ok
		Strikeouts	0.507 → 0.507	-0.001	0.2212	0.1789n=142 Below average	0.22109/9 qualified lineup mean	0.1787	-0.223	-0.001	+0.000 ok
		Walks	0.385 → 0.448	+0.063	0.0889	0.0802n=142 Average	0.10329/9 qualified lineup mean	0.0932	-0.046	+0.075	-0.007 ok
		Home runs	0.111 → 0.105	-0.006	0.0304	0.0316n=142 Average	0.02879/9 qualified lineup mean	0.0298	+0.006	-0.009	-0.000 ok
RELIEVER Adrian Morejon4.5198 BF	Middle Reliever (High Leverage)	Hits	1.069 → 1.107	+0.038	0.2169	0.1993n=352 Good	0.22439/9 qualified lineup mean	0.2063	-0.079	+0.034	-0.002 ok
		Strikeouts	1.436 → 1.435	-0.001	0.2212	0.2886n=352 Very good	0.22109/9 qualified lineup mean	0.2884	+0.305	-0.001	-0.000 ok
		Walks	0.419 → 0.487	+0.068	0.0889	0.0824n=352 Average	0.10329/9 qualified lineup mean	0.0958	-0.029	+0.065	-0.004 ok
		Home runs	0.108 → 0.102	-0.006	0.0304	0.0220n=352 Very good	0.02879/9 qualified lineup mean	0.0208	-0.038	-0.007	+0.002 ok
RELIEVER Bradgley Rodriguez4.1678 BF	Middle Reliever (High Leverage)	Hits	0.784 → 0.811	+0.027	0.2169	0.2063n=301 Good	0.22439/9 qualified lineup mean	0.2135	-0.044	+0.031	-0.001 ok
		Strikeouts	1.468 → 1.467	-0.001	0.2212	0.2590n=301 Good	0.22109/9 qualified lineup mean	0.2588	+0.158	-0.001	-0.000 ok
		Walks	0.497 → 0.578	+0.081	0.0889	0.0742n=301 Good	0.10329/9 qualified lineup mean	0.0863	-0.061	+0.060	-0.009 ok
		Home runs	0.052 → 0.049	-0.003	0.0304	0.0262n=301 Good	0.02879/9 qualified lineup mean	0.0247	-0.018	-0.007	+0.001 ok
RELIEVER Mason Miller3.0292 BF	Closer	Hits	0.679 → 0.704	+0.025	0.2169	0.1780n=272 Very good	0.22439/9 qualified lineup mean	0.1845	-0.118	+0.023	-0.003 ok
		Strikeouts	1.781 → 1.780	-0.001	0.2212	0.4166n=272 Very good	0.22109/9 qualified lineup mean	0.4164	+0.592	-0.001	-0.000 ok
		Walks	0.584 → 0.677	+0.093	0.0889	0.0931n=272 Average	0.10329/9 qualified lineup mean	0.1080	+0.013	+0.043	+0.002 ok
		Home runs	0.032 → 0.030	-0.002	0.0304	0.0231n=272 Very good	0.02879/9 qualified lineup mean	0.0218	-0.022	-0.005	+0.001 ok
RELIEVER Yuki Matsui2.738 BF	Reliever	Hits	0.498 → 0.516	+0.018	0.2169	0.1819n=260 Very good	0.22439/9 qualified lineup mean	0.1884	-0.096	+0.020	-0.003 ok
		Strikeouts	0.690 → 0.690	-0.001	0.2212	0.2520n=260 Good	0.22109/9 qualified lineup mean	0.2518	+0.084	-0.001	-0.000 ok
		Walks	0.328 → 0.379	+0.051	0.0889	0.1198n=260 Poor	0.10329/9 qualified lineup mean	0.1383	+0.084	+0.039	+0.011 ok
		Home runs	0.080 → 0.075	-0.004	0.0304	0.0291n=260 Average	0.02879/9 qualified lineup mean	0.0275	-0.004	-0.005	+0.000 ok
RELIEVER David Morgan1.2864 BF	Reliever	Hits	0.284 → 0.294	+0.010	0.2169	0.2210n=162 Average	0.22439/9 qualified lineup mean	0.2285	+0.005	+0.010	+0.000 ok
		Strikeouts	0.301 → 0.301	-0.000	0.2212	0.2343n=162 Average	0.22109/9 qualified lineup mean	0.2341	+0.017	-0.000	0 ok
		Walks	0.145 → 0.168	+0.023	0.0889	0.1127n=162 Below average	0.10329/9 qualified lineup mean	0.1303	+0.031	+0.018	+0.004 ok
		Home runs	0.033 → 0.031	-0.002	0.0304	0.0256n=162 Good	0.02879/9 qualified lineup mean	0.0242	-0.006	-0.002	+0.000 ok
RELIEVER Joe Musgrove0.7193 BF	Reliever	Hits	0.156 → 0.161	+0.005	0.2169	0.2169n=0 averagebelow floor	0.22439/9 qualified lineup mean	0.2243	0	+0.005	0 ok
		Strikeouts	0.159 → 0.159	-0.000	0.2212	0.2212n=0 averagebelow floor	0.22109/9 qualified lineup mean	0.2210	0	-0.000	0 ok
		Walks	0.064 → 0.074	+0.010	0.0889	0.0889n=0 averagebelow floor	0.10329/9 qualified lineup mean	0.1032	0	+0.010	0 ok
		Home runs	0.022 → 0.021	-0.001	0.0304	0.0304n=0 averagebelow floor	0.02879/9 qualified lineup mean	0.0287	0	-0.001	0 ok

Published pitching, integer draw beside the continuous model line

Pitcher	Role	BF	IP	expected IP	H int/cont	SO int/cont	BB int/cont	ER int/cont	Rate source
Dustin May	Starter	17.0	4.0	4.006	8 / 3.791	4 / 4.456	2 / 1.224	5 / 2.010	all_players_stats_recent_season_blend
JoJo Romero	Long Reliever	4.0	1.0	1.013	1 / 0.723	1 / 1.045	0 / 0.280	0 / 0.265	all_players_stats_recent_season_blend
Antonio Senzatela	Middle Reliever (High Leverage)	4.5	3.0	1.013	1 / 1.050	2 / 0.737	0 / 0.399	0 / 0.370	all_players_stats_recent_season_blend
Nick Pivetta	Starter	15.9	4.0	3.908	4 / 3.266	2 / 4.260	1 / 0.905	3 / 0.982	all_players_stats_recent_season_blend
Bradgley Rodriguez	Middle Reliever (High Leverage)	5.2	1.1	1.318	2 / 0.784	1 / 1.468	1 / 0.497	0 / 0.184	all_players_stats_recent_season_blend
Adrian Morejon	Middle Reliever (High Leverage)	5.4	1.1	1.318	1 / 1.069	1 / 1.436	0 / 0.419	1 / 0.412	all_players_stats_recent_season_blend
Jason Adam	Setup Man	4.1	1.0	0.989	0 / 0.752	1 / 0.507	1 / 0.385	0 / 0.312	all_players_stats_recent_season_blend
Mason Miller	Closer	4.2	1.1	0.989	1 / 0.679	0 / 1.781	0 / 0.584	0 / 0.426	all_players_stats_recent_season_blend

Audit: NO DEFECT FOUND

Invariant	Verdict	Evidence
blend arithmetic away_runs	PASS	4.0000 + 0.78 x (3.5655 - 4.0000) = 3.6611, recorded 3.6611
blend arithmetic home_runs	PASS	5.0000 + 0.44 x (4.2176 - 5.0000) = 4.6558, recorded 4.6558
blend arithmetic home_win	PASS	0.5782 + 0.00 x (0.6792 - 0.5782) = 0.5782, recorded 0.5782
published rounding faithful to the blend	PASS	blended 3.6611-4.6558 rounds to 4-5, layer published 4-5
published probability on the score's side	PASS	4-5 with p(home) 0.578153
Log5 additivity136 checks	PASS	136 of 136 pass
away club fields exactly one starter	PASS	1 starters in a staff of 3
away staff totals a regulation workload	PASS	24 outs recorded, 24 expected
away starter carries a continuous line	PASS	expected_IP 4.006
away starter batters faced inside the real band	PASS	Dustin May projected 17.0 batters faced; real starters run p01-p99 4-30, median 23, mean 21.75
away reliever batters faced inside the real band2 checks	PASS	2 of 2 pass
home club fields exactly one starter	PASS	1 starters in a staff of 5
home staff totals a regulation workload	PASS	27 outs recorded, 27 expected
home starter carries a continuous line	PASS	expected_IP 3.908
home starter batters faced inside the real band	PASS	Nick Pivetta projected 15.9 batters faced; real starters run p01-p99 4-30, median 23, mean 21.75
home reliever batters faced inside the real band4 checks	PASS	4 of 4 pass
away hitters satisfy every coherence rule	PASS	9 hitters checked; H<=AB, HR<=H, 2B+3B+HR<=H, SO<=AB, continuous line present
home hitters satisfy every coherence rule	PASS	9 hitters checked; H<=AB, HR<=H, 2B+3B+HR<=H, SO<=AB, continuous line present

Provenance. Read only from published artifacts and stored KPI value documents. Nothing is recomputed, so this document cannot disagree with what was published. Per-player metric adjustments are a measurement of the catalog at player level and are labelled as such throughout; the pipeline itself moved team runs, the win probability, and hitter lines by one club-wide ratio.

Verdict across the slate

How to read this, and what a defect looks like

Slate overview

Per game

Stage chain: every stage that touched the score

KPI team layer: the blend, with its arithmetic

What drove the KPI forecast for away_runs

What drove the KPI forecast for home_runs

Per-player metric movement

LAD against Evan Phillips (RHP)

ATL against Dylan Dodd (LHP)

Published pitching, integer draw beside the continuous model line

Audit: 2 CHECK(S) FAILED

Stage chain: every stage that touched the score

KPI team layer: the blend, with its arithmetic

What drove the KPI forecast for away_runs

What drove the KPI forecast for home_runs

Per-player metric movement

MIL against Chad Patrick (RHP)

SD against Joe Musgrove (RHP)

Published pitching, integer draw beside the continuous model line

Audit: NO DEFECT FOUND

What drove the KPI forecast for `away_runs`

What drove the KPI forecast for `home_runs`

What drove the KPI forecast for `away_runs`

What drove the KPI forecast for `home_runs`