Skip to main content
Every night after the US close, the index writes a snapshot of itself and commits it to the repository. Not a database, not a hosted service: a JSON file in git, next to the code that produced it. As of 11 October 2026 there are 13 snapshots, the first on 26 September 2026. Three nights are missing: 28, 29 and 30 September. The 29th was taken, found to have come off a stale cache, and deleted the same day rather than kept. The git log shows it going in and coming back out.
It is short, and saying so is more useful than implying an archive. The point is not how deep it is today. The point is that a day that passed without being recorded cannot be recovered by anyone, including us.

​
What a snapshot holds

Deliberately small. Each record keeps the shape of the category rather than a frozen copy of every row. sets is what makes a change postable. The counts say how many gained a pool; the names say which company it was.

​
Reading it

The file is in the repo, so there is nothing to call and nothing to authenticate.
Or clone it and read the git log, which is the honest version: you can see when each record landed and that none of them were edited afterwards.

​
Why it matters more than the index

The live index describes now, and so could anybody with a weekend and the same public sources. The method here is published precisely because copying it is not the problem. What cannot be copied is the record. A rating only means something if you can watch it move, and a category only tells you anything if you can see where it came from. Someone starting tomorrow starts at zero and waits.
Figures from a snapshot will not match the live index, because one is last night after the close and the other is the last five minutes. That gap is yesterday against now, not a disagreement.