Does the 80% band cover the men who were NOT in the room their trailing windows describe (played none of the team's previous 3 games for this team) as well as it covers the men who were, at the same projection level? H-222 measured the served band only on rows with a previous played row, so this population was outside it by construction.

Owner lane: calibration

Question

Does the 80% band cover the men who were NOT in the room their trailing windows describe (played none of the team's previous 3 games for this team) as well as it covers the men who were, at the same projection level? H-222 measured the served band only on rows with a previous played row, so this population was outside it by construction.

Prediction

Frozen at prereg 2026-09-27T01:17Z, before the runner existed: (1) out-of-room share 8-18% of played rows, not 48%; (2) PRIMARY shipped-band ceiling exceedance, out-of-room minus full-room, level-matched: +3 to +10pp, Bonf-12 interval excluding zero, 2025 alone same sign; (3) below-floor contrast smaller than above; (4) out-of-room level-matched mean residual positive (under-projected); (5) AAR band same sign as shipped; (6) partial-room contrast between 0 and the out-of-room one. REFUTED if the primary Bonf-12 interval contains 0.

Finding and verdict

REFUTED AS REGISTERED, BY 0.05pp. Shipped-band ceiling exceedance for men outside the room (8.1% of played rows) is +3.13pp over full-room men at matched level, Bonf-12 [-0.05, +6.08]; same sign in 2024 (+2.1) and 2025 (+3.1, held out, Bonf-12 excludes zero) and on arm S (+2.9). The version the data carries: it is spread, not bias (matched residual -0.04), and the AAR band, which reads rating spread, already covers these men (+0.09pp). H-111's 48% is a week-1 number (41.6% out of the room, 55% not full); in season it is 5-6%.

Reasoning

The claim is about calibration of the band that ships; the shipped band is level-only, the AAR band is level + itt + rating spread, and the two were never compared on the same rows by room membership.

Next action

(1) Frozen confirmatory on 2026 weeks 1-6 once they settle: shipped-band M0-M3 matched ceiling exceedance, same runner, bar Bonf-1 excluding zero -- week 1 alone will carry 40% M0 rows. (2) Walk-forward a shipped_band that reads rv_target_sd (or membership as a Mondrian cell) and grade pooled pinball90 and M0 exceedance on identical rows; 17% of M0 rows have NULL rv_target_sd, so the fallback matters. (3) Scope: F-08 serves RB/WR/TE imminent bands from sim_world, so the shipped band this grades serves QB and rest-of-season; QB is where the exploratory gap is largest (+13pp, n=194, interval spans zero).