The goalposts problem
Everyone knows to screenshot the claim. Almost nobody writes down what would settle it, which is the half that gets quietly rewritten.
August 29, 2026 · figures counted October 4, 2026
Any claim can be made true after the fact if you are allowed to edit the test. That is the central difficulty in scoring predictions, and it is usually not dishonesty — it is that nobody wrote the test down, so the only version that exists is the one recalled afterwards.
Fixed before, shown after
A record here carries its resolution criteria from the moment it is registered, and those criteria are part of what is anchored. They can be disputed, and a dispute is appended, but they cannot be quietly rewritten once the outcome is known.
It is also why 63 of 126 records are marked as never settling automatically. A claim with no agreed test is not scored on a guess. It stays unresolved and says so.
The cost of doing it this way
The archive settles slowly, and much of it never settles at all. A site that wanted a satisfying scoreboard would resolve those claims on judgement and move on. The reason not to is that the first time you do, every verdict on the site becomes an opinion.
Every claim here is checkable.
The exact words, the source and its date, the test fixed before the outcome, and what happened when it landed.