Forward claim scoreboard: every hypothesis, kept on the books
Per claim family (sport x hypothesis) from the 4 validation ledgers: verdict history ordered by run_ts, current status = last entry. Families with >1 distinct status flagged flipped.
How to read this figure
Each row is one sport x hypothesis family, its status is the most recent verdict in the ledger, and 5 families flipped verdict across reruns (e.g. NBA three-in-four fatigue went NULL then CONFIRMED).
What it means
The system tracks every sport-by-hypothesis family it has ever tested and keeps the failures on the record. Of 259 distinct families, 121 currently sit verified, 99 came back null, 28 were not testable, and 7 were retracted. A grader that only ever confirmed would not be credible; this one publishes more misses than hits.
Caveats and confounds
Current status is just the last ledger entry, and many rows predate run_ts stamping (run_ts is null), so ordering within a family can be undated.