[{"id":18,"date_start":"2026-01-02","date_end":"2026-04-27","kind":"correction","ticker":null,"title":"Verification: the backtest books from 2026-01-02 to 2026-04-27 now serve their signed bytes","body":"What was wrong: for the 79 books dated 2026-01-02 to 2026-04-27, /verify served a payload rebuilt from stored values, not the bytes that were signed. The stored predicted returns and weights are rounded to six decimals (largest differences 0.0000005 and 0.0000003). The stored ranks were renumbered after signing: the signed bytes number the lowest predicted return 1, and the stored rows number the highest 1. So the served payload did not hash to the book's payload_hash, and its signature could not be checked against it. What changed: /verify now serves the original bytes, rebuilt from the backfill's input file; each hashes to the book's recorded payload_hash, and its signature verifies. In these bytes rank 1 is the lowest predicted return; /batch and the data files keep rank 1 as the highest. Tickers, picks, grades, hashes and signatures are unchanged. These books remain a backtest: they were generated on 2026-04-28, after the dates they cover, and none carries a timestamp receipt. Found in review, timefire#995.","source":"admin","created_at":"2026-10-10T06:00:52.752764Z"},{"id":20,"date_start":"2026-01-02","date_end":"2026-09-29","kind":"correction","ticker":null,"title":"Correction: every book is now graded against the S&P 500 (RSP), the equal-weight index","body":"A book's verdict now compares its return with RSP, the equal-weight S&P 500, over the same ten trading sessions. It used to compare with SPY. Every book holds its long picks in equal weights, so RSP is the like-for-like benchmark; the scoring rules name no benchmark, so no signed book changes.\n\nAnchored record (from 2026-04-30): 52 beat, 2 flat, 48 trailed against SPY; 54 beat, 4 flat, 44 trailed against RSP. Full record: 92 beat, 6 flat, 85 trailed against SPY; 105 beat, 6 flat, 72 trailed against RSP. 55 of 183 verdicts changed.\n\nEach book's own return and every stock's grade are unchanged. The SPY return and margin stay recorded beside the new ones, and the two RSP prices behind each margin are kept with it as evidence. (timefire#1103)","source":"admin","created_at":"2026-10-16T01:09:39.762076Z"},{"id":19,"date_start":"2026-03-24","date_end":"2026-10-13","kind":"correction","ticker":null,"title":"Simulation: positions held across April 8, 2026 closed on their opening price's basis","body":"What was wrong: from August 31, 2026, positions held across April 8, 2026 were shown opening at prices on one basis and marked and closed on another. Our price archive held its history through April 7, 2026 on the basis it was first recorded on, and later dates on a basis that already reflected the splits, spin-offs and dividends that came after, so a split between the two read as a price move. 1067 closed trade(s) in DD, KLAC, CRWD, SPGI, HST, PFE and 187 other tickers carried a closing price on a different basis from their opening price. The largest change to a re-closed trade's return: DD opened 2026-03-30 (long_only), 215.44% published, 5.56% corrected. What changed now: those trades are closed on their opening price's basis; the opening price, share count and cost are unchanged. Daily portfolio values from 2026-04-08 to 2026-10-13 are restated, and positions opened from 2026-04-08 on are re-sized from the corrected portfolio value at unchanged prices. Except 2 later trade(s) re-opened at the price the feed now holds for their opening day: it was revised after publication while their closing price was not, as when a dividend that went ex during the holding period is folded into earlier prices late. Their closing price is unchanged and their return now includes that adjustment. Cumulative return through 2026-10-13: long_only 16.93% (published 17.97%); regime_gated 14.26% (published 15.15%). The price archive's history through April 7, 2026 is rescaled onto the later basis, and the simulation now refuses to price a position across a change of price basis. Forecasts, hashes, signatures and grades are unchanged. Found in review, timefire#1079.","source":"admin","created_at":"2026-10-14T07:22:38.829738Z"},{"id":3,"date_start":"2026-03-25","date_end":"2026-06-22","kind":"universe_change","ticker":"HOLX","title":"HOLX: ranked coverage interrupted, then permanently removed","body":"Hologic was taken private effective 2026-04-07. In the record: HOLX is absent from books 2026-03-25 through 2026-04-13, reappears 2026-04-14 through 2026-06-18 while our price source intermittently served quotes for it, and leaves permanently with the 2026-06-22 book. Because the company stopped trading on 2026-04-07, every HOLX pick anchored on or after that date is graded as untradeable (a zero return). The staleness guard we added in July 2026 now removes such names within days instead of weeks.","source":"admin","created_at":"2026-08-22T05:09:58.424233Z"},{"id":17,"date_start":"2026-04-10","date_end":"2026-09-24","kind":"correction","ticker":null,"title":"Simulation: published trade and NAV values restored after a vendor price re-adjustment","body":"What happened: the price provider re-adjusted the price history of EIX and LEN, and the nightly simulation replay rebuilt already published rows from the re-adjusted prices. The live API showed 14 trade(s) and 24 daily NAV row(s) that differed from the published export. EIX and LEN trades carried the re-adjusted prices and share counts; AES and DUK trades differed only by rounding. The largest change to any money amount was $0.0001, and to any return 0.000001. What changed now: the stored rows were restored to the published export values. This repair changed only simulation trade and NAV rows; the export files and the signed forecasts were not touched. Replays now keep published simulation rows (timefire#972). Found by the export guard, timefire#957.","source":"admin","created_at":"2026-10-10T03:17:39.429372Z"},{"id":5,"date_start":"2026-06-09","date_end":"2026-06-09","kind":"universe_change","ticker":"SBNY","title":"SBNY joined the prediction universe with the 2026-06-09 book","body":"SBNY is ranked from this book onward. Earlier books are not backfilled - its record starts here.","source":"admin","created_at":"2026-08-22T05:09:59.152624Z"},{"id":1,"date_start":"2026-06-29","date_end":"2026-07-01","kind":"publisher_gap","ticker":null,"title":"No books were published June 29 - July 1, 2026","body":"After the June 26 book was sealed, a GPU stage of the daily prediction pipeline began hanging during feature extraction. Rather than risk publishing predictions built from incomplete data, we halted the pipeline entirely while we diagnosed it. The stage was moved to different hardware and publishing resumed with the July 2 book (July 3 was the Independence Day market holiday).\n\nThese three days will never be backfilled. Every book in this record was committed before its outcomes were knowable; a book added after the fact - even an honest one - would break that rule, and the rule is the record. The gap stays, and this note is its explanation.\n\nWhat changed as a result: the pipeline stage that failed was replaced, and the system now pages us the same evening if a scheduled book fails to seal, rather than relying on someone noticing.","source":"admin","created_at":"2026-08-22T05:09:57.707633Z"},{"id":4,"date_start":"2026-07-23","date_end":"2026-07-23","kind":"universe_change","ticker":"CTRA","title":"CTRA left the prediction universe with the 2026-07-23 book","body":"Coterra Energy merged into Devon Energy effective 2026-05-07 and stopped trading. Stale quotes kept it in the ranked set until the 2026-07-23 book removed it. Every CTRA pick anchored on or after 2026-05-07 is graded as untradeable (a zero return - never a synthetic price). This episode is what led to the automatic staleness guard now applied at every seal.","source":"admin","created_at":"2026-08-22T05:09:58.772080Z"},{"id":6,"date_start":"2026-08-05","date_end":"2026-08-05","kind":"universe_change","ticker":null,"title":"95 names joined the prediction universe with the 2026-08-05 book","body":"Ranked coverage expanded from 437 to 532 names on 2026-08-05, adding S&P 500 constituents that previously lacked sufficient history in our data sources. Each added name's record starts with this book; earlier books are not backfilled. Removals remain individually annotated.","source":"admin","created_at":"2026-08-22T05:09:59.523274Z"},{"id":10,"date_start":"2026-08-10","date_end":"2026-08-10","kind":"correction","ticker":null,"title":"Benchmark precision correction for the 2026-08-10 book","body":"A later NAV rebuild repeatedly reconstructed a six-decimal benchmark/alpha pair, moving both values by 0.0002 percentage points. We restored the SPY 10-day return to -1.2367% and the public book alpha to +1.08005%, matching the immutable export originally published for this book.\n\nPredictions, realized stock returns, verdicts, the signed canonical payload, payload hash, IPFS CID, and OpenTimestamps proof were unchanged. The database correction and NAV rebuild were explicitly approved by Ray. Incident: timefire#852 / #856.","source":"admin","created_at":"2026-09-25T22:09:08.053228Z"},{"id":2,"date_start":"2026-08-17","date_end":"2026-08-17","kind":"universe_change","ticker":"EA","title":"EA left the prediction universe with the 2026-08-17 book","body":"Electronic Arts was taken private; its last tradeable session was 2026-08-04. Its price feed then froze (identical closing price, zero volume), and our staleness guard flagged it on 2026-08-17, removing it from the ranked set rather than ranking it on a dead quote. EA appears in books through 2026-08-14; any EA pick whose measurement window crosses the delisting is graded as untradeable (a zero return - never a synthetic price). Nothing is removed retroactively.","source":"admin","created_at":"2026-08-22T05:09:58.167170Z"},{"id":7,"date_start":"2026-08-27","date_end":"2026-08-27","kind":"universe_change","ticker":"EQR","title":"EQR left the prediction universe with the 2026-08-27 book","body":"EQR is not ranked in this or later books: no real closing price since 2026-08-21, which usually means a corporate action (delisting, acquisition, or rename); removed from the ranked universe rather than ranked on a stale quote Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-08-27T21:32:37.500280Z"},{"id":8,"date_start":"2026-08-28","date_end":"2026-08-28","kind":"universe_change","ticker":"AVB","title":"AVB left the prediction universe with the 2026-08-28 book","body":"AVB is not ranked in this or later books: no real closing price since 2026-08-17, which usually means a corporate action (delisting, acquisition, or rename); removed from the ranked universe rather than ranked on a stale quote Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-08-28T21:29:47.073143Z"},{"id":9,"date_start":"2026-08-28","date_end":"2026-08-28","kind":"universe_change","ticker":"EQR","title":"EQR joined the prediction universe with the 2026-08-28 book","body":"EQR is ranked from this book onward. Earlier books are not backfilled — its record starts here.","source":"auto_seal","created_at":"2026-08-28T21:29:47.073149Z"},{"id":12,"date_start":"2026-09-17","date_end":"2026-09-30","kind":"correction","ticker":"CTVA","title":"CTVA: spin-off distribution omitted from the September 17 grade","body":"Wrong: the September 17 CTVA forecast initially showed a -84.5824% stock return. Right: the corrected distribution-inclusive return is -0.8586%; the directional hit remains True. On October 1, each CTVA share carried one distributed VYLR share. Entry CTVA close was 81.529998779296875; exit CTVA and VYLR closes were 12.569999694824219 and 68.260002136230469. The calculation is (exit parent + distributed child) / entry parent - 1. Cause: the parent-only price series used by our grader did not account for the distribution. This explicitly disclosed corporate-action exception uses unadjusted parent/child closes; the original signed adjusted-close recipe is preserved, not rewritten. Scope: one existing grade corrected; CTVA forecast windows from September 18–30 use the same distribution-inclusive method when due. No long-basket NAV, simulation position or broker holdings changed in this correction. Original observations and grade are retained in the audit record. Previously frozen mirror/export files retain their original value; this note and the current API identify the correction. Forecast payloads, hashes, signatures, IPFS CIDs and OTS proofs are unchanged. Found by price-integrity alert timefire#899; corporate action previously flagged in #857. Terms: https://www.sec.gov/Archives/edgar/data/2128626/000119312526409909/d115079d8k.htm","source":"admin","created_at":"2026-10-02T00:52:41.811005Z"},{"id":11,"date_start":"2026-09-30","date_end":"2026-09-30","kind":"universe_change","ticker":"SBNY","title":"SBNY left the prediction universe with the 2026-09-30 book","body":"SBNY is not ranked in this or later books: no real closing price since 2026-09-24, which usually means a corporate action (delisting, acquisition, or rename); removed from the ranked universe rather than ranked on a stale quote Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-09-30T21:40:23.286150Z"},{"id":13,"date_start":"2026-10-02","date_end":"2026-10-02","kind":"universe_change","ticker":"CTVA","title":"CTVA left the prediction universe with the 2026-10-02 book","body":"CTVA is not ranked in this or later books: corporate action awaiting input/trading validation (spin_off 2026-10-01) Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-10-02T21:45:36.570365Z"},{"id":14,"date_start":"2026-10-05","date_end":"2026-10-05","kind":"universe_change","ticker":"QRVO","title":"QRVO left the prediction universe with the 2026-10-05 book","body":"QRVO is not ranked in this or later books: corporate action awaiting input/trading validation (stock_and_cash_merger 2026-10-05) Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-10-05T21:56:10.714819Z"},{"id":15,"date_start":"2026-10-06","date_end":"2026-10-06","kind":"universe_change","ticker":"PSKY","title":"PSKY left the prediction universe with the 2026-10-06 book","body":"PSKY is not ranked in this or later books: corporate action awaiting input/trading validation (name_change 2026-10-06) Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-10-06T21:40:59.222217Z"},{"id":16,"date_start":"2026-10-06","date_end":"2026-10-06","kind":"universe_change","ticker":"WBD","title":"WBD left the prediction universe with the 2026-10-06 book","body":"WBD is not ranked in this or later books: corporate action awaiting input/trading validation (cash_merger 2026-10-06) Its full history through the previous book remains in the record; nothing is removed retroactively.","source":"auto_seal","created_at":"2026-10-06T21:40:59.222222Z"}]