For QB forecasts whose error aar_attribution labels REGIME (2024-25 walk-forward) plus the 2026 served zero-line QBs, what did the club's pre-kickoff judgement (injury designation of the incumbent starter, depth chart) say about his room? Knowable from judgement = a B-layer method defect; not knowable = outside this layer.

Owner lane: b-judgement

Question

For QB forecasts whose error aar_attribution labels REGIME (2024-25 walk-forward) plus the 2026 served zero-line QBs, what did the club's pre-kickoff judgement (injury designation of the incumbent starter, depth chart) say about his room? Knowable from judgement = a B-layer method defect; not knowable = outside this layer.

Prediction

Prereg research/scientist/prereg/2026-09-23T1958Z-h114-qb-regime-judgement.md: J4 (incumbent silent: not designated, did not play) largest, J2 (incumbent Out) second, J0 >= 15%. Falsifier: J0+J1+J5 (no pre-kickoff signal of a change) is a majority.

Finding and verdict

FALSIFIER FIRED as registered: 40 of 61 regime QB misses carry no pre-kickoff judgement signal of a starter CHANGE (25 are non-starters over-projected, 15 are a continuing backup starter). J4 (silent absence) was predicted largest and is the smallest started class. Disclosed post hoc: 15 of 36 started misses are a backup who had already started the previous game and was still charted rank 2 -- a knowable signal the model does not read; a counterfactual replay is not run here.

Reasoning

The prereg asked about a change of starter; the largest recoverable class turned out to be a non-change the depth chart misreports.

Next action

New hypothesis filed: replace depth_chart_rank with 'started the team's previous game' (observable Monday) for the QB room and replay the J1 forecasts through the production path at the stored arm; separate from J3 (promote on Doubtful/Questionable-and-inactive) which needs the inactives clock (Lock.GAME).