0023Z split the season-share lead into two opposing MEAN properties: (a) the lines_for targets line is over-dispersed (calibration slope 0.83) and (b) given the line, the season-to-date target share f is under-read (partial -0.15..-0.25, 9/9 seasons). It then scored f on MAE and found it nil. The start/sit decision is an ORDERING consumer: a common slope on the line with position intercepts cannot reorder two men at one position, so (a) is invisible to it and only (b) can move it. Does adding a walk-forward fitted f term to the FANTASY line change same-position start/sit calls, and does it close any of the consensus's decision lead?

Owner lane: decision

Question

0023Z split the season-share lead into two opposing MEAN properties: (a) the lines_for targets line is over-dispersed (calibration slope 0.83) and (b) given the line, the season-to-date target share f is under-read (partial -0.15..-0.25, 9/9 seasons). It then scored f on MAE and found it nil. The start/sit decision is an ORDERING consumer: a common slope on the line with position intercepts cannot reorder two men at one position, so (a) is invisible to it and only (b) can move it. Does adding a walk-forward fitted f term to the FANTASY line change same-position start/sit calls, and does it close any of the consensus's decision lead?

Prediction

A1 - A0 flex_sym accuracy in [-0.003, +0.006] with the 97.5% interval containing 0; regret interval containing 0. A1 flips under 6% of flex_sym calls. A1s - A0 interval contains 0. C - A0 accuracy positive (CP-H17 direction) and A1 closes under a quarter of it. A0r - A0 exactly 0. PLANT detected. Refutation clause: an A1 - A0 flex_sym accuracy interval that excludes 0 in either direction refutes this prediction; excluding 0 positively would mean the ordering consumer carries what MAE did not.

Finding and verdict

REFINED. H-021 end to end is still MEASURED NOTHING on MAE; on the start/sit ORDER the season-share under-read is real on close calls (+0.50 pp [+0.32,+0.69], regret -0.059 [-0.081,-0.038] pts per call, 2018-25, 8/8 seasons, held-out 2018-22 alone +0.50 pp) and at the edge of zero on the primary flex_sym 2023-25 cell (+0.46 pp, Bonf-2 lower bound -0.000). The prediction (primary nil) stands; the secondary is the finding. It is small beside the consensus lead (+4.4 pp) and lives at weeks 5+.

Reasoning

The decision lane exists because MAE and the order rank arms differently. 0023Z closed f on MAE; this asks the consumer that isolates part (b).

Next action

(1) Frozen, prospective: score the A1 term on 2026 ISSUED lines weeks 5+ (after 2026-10-13) on flex_sym against the consensus. (2) The same contrast on the SERVED product (0.40 blend), which is the order a manager sees. (3) A frozen family of season-level usage terms (snap_pct_season, carry_share_season, fp season-to-date) to separate 'target share' from 'season usage' as the under-read quantity.