Is the existence surface's rapm_theta/uplift_cate played-vs-not-played gap position composition, and does each builder's own rule explain regime_gate and kalman_share presence exactly? Mirror: do defshare.feature_sql's per-defender joins carry a played/not-played gap?

Owner lane: construction

Question

Is the existence surface's rapm_theta/uplift_cate played-vs-not-played gap position composition, and does each builder's own rule explain regime_gate and kalman_share presence exactly? Mirror: do defshare.feature_sql's per-defender joins carry a played/not-played gap?

Prediction

(1) rapm_theta and uplift_cate within-position gap < 0.02 in RB/WR/TE; (2) regime_gate within-position gap 0.1-0.4 with 100% of presence matched by its prior-count rule; (3) defshare scan finds >= 1 joined table with gap > 0.05. SURPRISE: rapm_theta within-position gap > 0.05. CONTROL: kalman_share gap > 0.5 inside each of RB, WR, TE, else void.

Finding and verdict

TESTED. rapm_theta/uplift_cate: position mix, exactly (1.000 vs 1.000 within RB/WR/TE). regime_gate and kalman_share presence: 100% explained by each builder's rule. Frozen control FAILED on RB kalman (0.416 < 0.5), so the confirmatory reading is void by the row's own terms; counts filed as discovery. Defence: joins clean, spine is where existence breaks (H-215).

Reasoning

The surface was producing a false existence question and hiding a real one (depthmv2 RB). Fixing the instrument generates the right question from now on.

Next action

depthmv2 at RB (0.951 vs 0.829) is now on the existence surface, unexamined. The defender_week spine (23.2% of ACT roster-weeks absent; no defense_snaps = 0 rows) belongs to H-215.