Datasets
Data status
What has stopped updating, what is empty, how far behind each dataset runs, and how history is repaired and restated.
Flagged STALE by the engine (3)
- STALEoptions underlying dailylast row May 15, 2026, 147 days before the day measured; expected every 7 days or sooner.
- STALEcrypto price eventslast row Mar 28, 2024, 925 days before the day measured; expected every 60 days or sooner.
- STALEtennis matcheslast row Sep 13, 2020, 2,217 days before the day measured; expected every 60 days or sooner.
Empty tables (6)
fails to deliver, readjust queue, edgar facts, edgar filers, edgar quarters, edgar tickers.
Empty tables. fails to deliver and all four edgar tables. readjust queue is empty, which is normal.
Freshness by dataset
| Dataset | Last row | Behind |
|---|---|---|
| Tennis results and odds STALE | Sep 13, 2020 | 2,217 days behind on the day measured |
| Options activity STALE | May 15, 2026 | 147 days behind on the day measured |
| Fund holdings (13F) | Jun 30, 2026 | 101 days behind on the day measured |
| Point-in-time fundamentals | Aug 31, 2026 | 39 days behind on the day measured |
| Weather | Sep 2, 2026 | 37 days behind on the day measured |
| Short interest | Sep 15, 2026 | 24 days behind on the day measured |
| Soccer results and odds | Sep 28, 2026 | 11 days behind on the day measured |
| Crypto perp funding | Sep 30, 2026 | 9 days behind on the day measured |
| House factor returns | Oct 1, 2026 | 8 days behind on the day measured |
| Analyst ratings | Oct 3, 2026 | 6 days behind on the day measured |
| Shares and float | Oct 3, 2026 | 6 days behind on the day measured |
| Earnings estimates | Oct 4, 2026 | 5 days behind on the day measured |
| Prediction markets | Oct 4, 2026 | 5 days behind on the day measured |
| US stock and fund prices | Oct 8, 2026 | 1 day behind on the day measured |
| Daily short volume | Oct 8, 2026 | 1 day behind on the day measured |
| Factor library | Oct 8, 2026 | 1 day behind on the day measured |
| House momentum model | Oct 8, 2026 | 1 day behind on the day measured |
| Currencies | Oct 8, 2026 | 1 day behind on the day measured |
| Index membership | Oct 9, 2026 | Up to date on the day measured |
| Insider trades (Form 4) | Oct 9, 2026 | Up to date on the day measured |
| Company news | Oct 9, 2026 | Up to date on the day measured |
| Commodity futures and spot | Oct 9, 2026 | Up to date on the day measured |
| Crypto prices | Oct 9, 2026 | Up to date on the day measured |
| The 24/7 tape | Oct 9, 2026 | Up to date on the day measured |
| Memecoins | Oct 9, 2026 | Up to date on the day measured |
| Earnings calendar | Feb 5, 2027 | Holds dates 119 days ahead (scheduled rows) |
| Splits and dividends | Sep 14, 2029 | Holds dates 1,071 days ahead (scheduled rows) |
| FRED economic series | Oct 1, 2036 | Holds dates 3,645 days ahead (scheduled rows) |
| Security master | - | No dated rows |
| Fails to deliver (SEC) | - | No dated rows |
| SEC filing facts (XBRL) | - | No dated rows |
History repairs
Every repair to history is logged in the engine with who made it and why. The catalogue holds 35 logged repairs, from Oct 5, 2026 to Oct 9, 2026; the rows themselves are in the engine (your AI can read momentum.history_repairs with query_warehouse).
- history repairs: 35 rows, Oct 5, 2026 to Oct 9, 2026. Upkeep: every repair to history, who and why.
- readjust done: 28,511 rows, Jan 3, 2000 to Oct 8, 2026. Upkeep: splits and dividends re-applied to history.
- readjust queue: 0 rows. Upkeep: re-adjustments waiting.
- readjust scanned: 115 rows, May 1, 2026 to Oct 8, 2026. Upkeep: days scanned for new splits and dividends.
- price hygiene: 9,572 rows, Sep 16, 2026 to Oct 9, 2026. Upkeep: tickers whose early history was cut, and why.
- adj step review: 2,703 rows, Oct 4, 2026 to Oct 9, 2026. Upkeep: adjusted-price steps under review.
Still open: 48 splits are still unadjusted, 30 of them in 2026. (ACCURACY, Broken splits)
How restatements are handled
When a company restates a quarter, the engine keeps the earlier figure in a restatement store (fundamentals point in time versions) with the day it was seen, and the main table carries the newest. A backtest reads each quarter from its filing date, so it never sees a figure before it was public.
The store holds 23,682 earlier versions for 940 names, for periods from Jan 31, 1993 to Jul 31, 2026. Which figures changed and by how much is in the engine, not the catalogue.
Other tables keep no version history: a corrected price, filing or series replaces the old one. FRED series are their latest revised values, with no earlier vintages.
Every count and date here was measured on Oct 9, 2026 from the QuantGPT engine's warehouse and is read from the catalogue file when the site is built. Metadata only: no vendor rows are shown.