Score a signal detector against its own draw universe and persist the boards humans read

A signal system looked "positive" (+1.4% average 21-day forward return, 52% up) until compared with every ticker it draws from over the same weeks (+5.6%, 55% up): it underperformed its own source list. Separately, the human-facing summary board (the one the operator actually reads and would trade from) was rendered into HTML each day but never written to any table, so it could never be backtested even months later. And 96% of stored signals had no ticker attached, so they were counted as detections but could never be scored.

Three rules for any "detect then bet" pipeline: (1) benchmark = the detector's own draw universe over the same dates (all tickers it could have named), not SPY; market drift makes a coin-flip look like alpha. (2) Persist every human-facing board row (board name, entity, ticker, the numbers shown, the date) to a board_history table the day the board ships; the rendered page is not a record. (3) Separate "no outcome yet" from "no ticker": rows without a mapped ticker are noise by construction and must be excluded from signal counts and hit rates. Put a scorecard (what the board said 5/21 days ago vs price vs universe) at the top of the report; nothing is bettable until that scorecard beats the universe over 60+ board-days.

Forward returns of any stock list inherit market drift; only the same-period return of the pool the detector selects from isolates selection skill. Rendered reports are lossy and unqueryable; NULL outcome columns conflate not-matured with not-mappable.