Does the production line under-project the TARGETS of players whose club changed since his last game of S-1 (a club decision) relative to stayers, and is the miss volume rather than rate?

Owner lane: b-judgement

Question

Does the production line under-project the TARGETS of players whose club changed since his last game of S-1 (a club decision) relative to stayers, and is the miss volume rather than rate?

Prediction

Frozen in protocol.json (sha f12d5400) before any contrast: E1 NEGATIVE, movers' targets under-projected vs stayers by 0.03-0.10, interval excluding zero, same sign >= 5 of 7 seasons; E2 spans zero; E3 shrinks; E4 same sign; miss in share not room volume. REFUTED if E1 >= 0 or spans zero or inside placebo band.

Finding and verdict

REFUTED on its model half; football half reversed on rates. Held-out 2019-2025 walk-forward: the production line does NOT under-project movers' targets (+0.013 [-0.027, +0.056], inside placebo), not by week, season, rostered population, rate, share or room. The served 2026 board shows -0.18 / -0.10 on 56-68 movers, intervals spanning zero; that path (latest_rows + D-80 FantasyPros blend + matchup factor) is not the walk-forward path and has no history to grade it on.

Reasoning

H-195 was filed unreviewed off one week-1 board; the board reproduces in direction but at n=56 its interval spans zero, and seven held-out seasons of the production line show no mover bias to explain.

Next action

The served-path residue is time-blocked, not work-blocked: re-run run/h195_board2026.py (it keys either board format) once 2026 weeks 1-6 are graded (~2026-10-20). With ~350 mover player-weeks the interval should be about +/-0.08. If the board's gap holds while the walk-forward line has none, the defect is in the SERVING path (latest_rows' old-team pace, H-138; the D-80 blend), not the model.