J-01 removed volume_for market terms, so the team pass/rush anchor is now [1, plays_l3, pass_rate_l3]. H-080 bounded recomposing those two windows at -0.123 pass att WITH the market in the design. Without the market, does recomposing the two team windows move the anchor, and does that reach the player line?

Owner lane: s-observation

Question

J-01 removed volume_for market terms, so the team pass/rush anchor is now [1, plays_l3, pass_rate_l3]. H-080 bounded recomposing those two windows at -0.123 pass att WITH the market in the design. Without the market, does recomposing the two team windows move the anchor, and does that reach the player line?

Prediction

P1: the nested pick beats l3 on pass att by -0.10 to -0.30 (interval excludes zero), picking a long memory (ewma h 4-12 or flat 6-8), because without the market the anchor's only team identity is three noisy games. P2 rush: -0.10 to -0.30, excludes zero. C_shuffle worse than A0 by > +0.15 pass. C_const worse than A0. H_hind beats every candidate. Stage 2 (if it runs): fantasy MAE moves by less than 0.01 pts/row in either direction, targets MAE improves by 0.00-0.02 -- the share model absorbs most anchor level via its own team features. I predict nil end to end is the likelier stage-2 outcome. | stage 2 (amendment-1): L1-L0 and L2-L0: fantasy MAE within +-0.01 pts/row, carries MAE -0.000 to -0.010, targets within +-0.005. The rush anchor gain (-0.041 team carries/game, 0.7% of MAE) spread over ~3 carrying players cannot move a per-player MAE by more than ~0.01.

Finding and verdict

REFINED. Anchor: rush recomposition (nested ewma4-8) beats shipped l3 by -0.041 team carries/game Bonf-2 [-0.078,-0.005], 7/7 seasons negative, 0.7% of rush MAE; pass nil (-0.012 [-0.042,+0.019]). Player line: MEASURED NOTHING (fantasy -0.0002 [-0.004,+0.004] Bonf-6, carries +0.0006, targets -0.0004), resolution +-0.004 pts. H-021 end to end remains MEASURED NOTHING; the anchor route of the team windows is now tested.

Reasoning

Premise moved (J-01) so H-080 bound was stale; this is the only team-window route that feeds every player line.

Next action

Anchor route of the 11 team windows: tested, player-line nil. Remaining: the share-tree copies (TEAM_FEATURES, TREND_FEATURES, opp_*) as a MATCHED_BOTH_BALL TEMP team_week shadow; and the pass mix from a script-neutral input (coaching.proe_l4, B lane). The shared-bundle override is a +-0.004 harness for any anchor question (10x finer than a refit arm).