Does `bets` hold a lock for 2026 week 1, and is the locked 'ours' row the object that shipped -- same number as the stored pre-slate board, with a spread that could be graded?

Owner lane: calibration

Question

Does `bets` hold a lock for 2026 week 1, and is the locked 'ours' row the object that shipped -- same number as the stored pre-slate board, with a spread that could be graded?

Prediction

Row as filed: zero 2026 week-1 rows. No fresh prediction was frozen before this run; the enumeration below was run once, as written, and every output is reported.

Finding and verdict

TESTED. Premise as worded is stale (8,057 rows exist). The substantive defect stands in a different form: for 2026 weeks 1-2 the lock ledger is a post-hoc backfill by a later commit, differs from the stored board by a one-signed -0.28 to -0.61 pts mean, and carries no sigma at all. Any CLV/coverage grade read off `bets` for these weeks grades a reconstruction, not the object that shipped (CLAUDE.md rule 3).

Reasoning

The calibration lane's subject is whether the published spread is honest. The only artefact that declares a lock has no spread column populated, so the honesty of the locked spread is unmeasured, not measured-and-fine.

Next action

(1) freeze must write sigma (or lo80/hi80) from the same band the board prints, or bets cannot be a coverage source; (2) a live week-4 freeze with frozen_at < placed_at + 15 min is the first real lock -- check it on 2026-10-01. (3) Name-join outliers (max 20 pts) not audited; a gsis-keyed board would remove the join risk.