Does a new head coach break the prior season's throw tendency in weeks 1-3, and does the production volume anchor, which reads that prior season through a season-crossing window, pay for it?

Owner lane: b-judgement

Question

Does a new head coach break the prior season's throw tendency in weeks 1-3, and does the production volume anchor, which reads that prior season through a season-crossing window, pay for it?

Prediction

Frozen 21e51b89 (2026-09-26T2117Z), amended pre-scoring 758360d1: premise replicates in direction, narrower than -0.02 vs +0.43, PROE difference 95% excludes zero; endpoint E_new(0.10, throws) -0.1 to -0.5 attempts spanning zero, E_did spanning zero -- I predict the confirmatory family is nil. Served family: larger E_new (-0.3 to -1.0), E_did still spans zero. REFUTED if any E_did(throws) 99.375% interval excludes zero in favour of the shrink.

Finding and verdict

The premise holds and is stronger than filed once game script is removed (PROE r +0.37 same-HC vs -0.01 new-HC over 19 seasons). Shrinking the season-crossing throw-rate input toward the league for new-HC teams cut team throw-attempt error in weeks 1-3 by 0.24 attempts [99.375% 0.07, 0.43] on held-out 2019-2025, in every one of the 7 seasons, while the same shrink cost same-coach teams +0.05 (difference -0.30, permutation p 0.0015). My frozen prediction was wrong on the endpoint. The effect is small against the error, rush is unaffected, and the first prospective read (2026 weeks 1-2, 7 teams) goes the other way at +1.09, so this is TESTED on history and unconfirmed forward.

Reasoning

H-194 had been stopped twice on H8 without anyone checking whether its mechanism touches efficiency.fit. It does not: the site is the volume anchor, which reads team_week only.

Next action

Not before 2026-09-30: rerun research/scientist/run/h194_new_coach_pass_rate.py unchanged on a snapshot with 2026 week 3 and read the 2026 E_new (21 new-HC team-games). If it stays positive, the history result does not carry forward and the served weeks-1-3 anchor should keep the prior coach's rate. If it turns negative, a build item (new-HC shrink of the S-1 contribution to the throw-rate input) goes to semantic review. Separately: real play-caller data (DATA-WANTED), and the three wiki-string head_coach cells (mechanical crumb).