{"approved_by":"owner, 20 Oct 2026: \"lawyers say ok\" (Knightsbridge Law's read of SCORING-ADDENDUM-DRAFT-2026-10-20.md, relayed by the owner); plan approved the same day","batch":1,"file":"SCORING-ADDENDUM-DRAFT-2026-10-20.md","hash":"7b30bf562542ce4a140e96dd0eb58a299314bf0085b77307043a47d0cb79a7ff","kind":"scoring_rules","prev":"0000000000000000000000000000000000000000000000000000000000000000","recorded":"2026-10-20T06:27:24Z","seq":1,"sha256":"a3dcfc08a1d13a71a45499d76c9999346f1310c805c03d11e1b8394b0b17deed","text":"# Scoring addendum: DRAFT for counsel and the owner, 20 October 2026\n\nA rule text, to be signed with the portfolio's ML-DSA-87 key and its fingerprint anchored on chain BEFORE the FDA acts on any filing it covers. It changes no signed call, no book weight and no sealed text. It fixes how the signed calls and the sealed books are reported at each FDA decision, adds the two comparisons any reader will make, and settles two points the sealed book texts leave open. Kept private until counsel's word and the owner's.\n\n## The signed calls (ledger.jsonl, batches 1 to 3)\n\n- OUTCOME, per filing: approval at the FDA's first action on that filing counts 1; a complete response letter counts 0. A withdrawal or a refusal to file is reported and not scored. A goal date the FDA moves carries the filing to the new date.\n- FORECAST, per filing, exactly as signed: the rate of the call's group, 83.5% for a filing called above average and 64.5% for one called below average, the rates sealed in the books (book-ledger.jsonl, batches 1 to 3). A filing with no call is not scored.\n- REPORTED at each decision, over every filing decided so far: the hits at a 50% cut-off and the Brier score (the average squared gap between the forecast and the outcome), beside two baselines: \"always approve\" (100%) and the cohort base rate of 63% (Avalere, as the room cites it). The number of filings decided is stated every time.\n- PLAIN SENTENCE: at these rates every forecast and both baselines predict approval at a 50% cut-off, so the hit counts will be identical until a filing is refused. When a baseline ties or beats the calls, the page says so in one sentence.\n- Counted across the groups, never filing by filing, as the room already states.\n\n## The sealed books (book-ledger.jsonl, Books 1 to 3)\n\n- EXIT: the close of the first trading session after the day the decision is announced, the window the room's decision-day study already uses.\n- OPEN POSITIONS: marked at that same close at each decision.\n- PRICES: daily closes from Yahoo Finance's chart service, each with the time it was read. The sealed texts name stockanalysis.com for the entry close; the entry closes stored at signing (osint/book-entry-closes-2026-10-14.json) were read from Yahoo Finance the next morning. Both report the exchange's official close. Book 3 enters at the close of 15 October 2026.\n- REPORTED at each decision: each book's value, the equal-weight book of the same positions, and XBI, all from the same entry close. A loss is published exactly as a gain. Book 3 is reported only once its switch is on.\n\n## Fallback\n\nIf this addendum is not anchored on chain before the FDA acts on INO-3107 (target date 30 October 2026), INO-3107 is left out of every score this addendum defines. The sealed book texts already apply their own fallback to the books.\n"}