H-137: 'three teams' scoring anchor is keyed to our OWN margin, so the anchor that exists to import outside information imports ours.' Is that what production runs at HEAD, and what is the version of the claim the evidence can carry?

Owner lane: Not recorded

Question

H-137: 'three teams' scoring anchor is keyed to our OWN margin, so the anchor that exists to import outside information imports ours.' Is that what production runs at HEAD, and what is the version of the claim the evidence can carry?

Prediction

Frozen in 89f2d900 before the script existed (protocol.json). Premise: P1 the three teams are now priced (stale); P2 latest_rows re_anchored 0 when MARKET_ON off, >0 when on; P3 every served K anchor matches own_team_scores (0 split). Refined claim: REFIT - SHIPPED K -0.005 to -0.04 MAE, DEF -0.03 to -0.20 MAE (errors-in-variables: a slot fitted on the sharper book regressor overstates what our noisier score carries); REFIT - BOOK24 same sign. REFUTED if REFIT - SHIPPED >= 0 on both endpoints.

Finding and verdict

H-137 as filed: NOT OBSERVED -- the 'three teams' are priced now, the latest_rows stand-in is withheld by G-TDANC (0/450 rows), and K/DEF keys all 32 teams to our own score by design (544/544). The refined claim the evidence could carry -- that the K/DEF slot fitted on the book but served ours is a fit/serve mismatch worth closing by refitting on our score -- is REFUTED on 2025: refitting makes DEF worse by +0.063 [+0.006, +0.122] and leaves K unchanged (-0.004 [-0.027, +0.020]), and a shuffled (constant) slot ties the refit. D-383's cost (+0.36 DEF) does not live in the slot's coefficients; it lives in the regressor, our team score, which is compressed to 57% of the book's spread on the same team-weeks. That points at the game layer (our own team score, gradeable through game_bets), not the K/DEF slot.