Note
Sometimes the honest verdict is that nobody can tell
The pressure is always to produce a verdict. Refusing is the harder discipline and the one that keeps the rest of the numbers worth anything.
A prediction can carry a number, a date and a named source and still fail to resolve. When that happens the honest verdict is that nobody can tell, and recording it as such is a real outcome rather than a broken one.
This is unpopular. Everyone wants the verdict. A page that says “undetermined” feels like the machinery failed.
The ways a well-formed claim still fails to settle
The source stopped publishing. A test written against a series that gets discontinued, rebased or paywalled has no referee any more. The claim was fine. The world moved.
The revision changed the answer. GDP, CPI and unemployment are all restated after release. A prediction can be met on the advance estimate and missed on the third. Unless which print counts was fixed in advance, both readings are defensible and neither should win.
The wording carried two readings and nobody noticed until it mattered. Ambiguity that seemed harmless in January becomes the whole argument in December, because the two readings now point at different verdicts.
The evidence is not reproducible. Something that would let us call it, but only from a page that has since changed, with no independent copy. We know; a stranger could not confirm it. That is not good enough to publish as a result.
Why refusing is the expensive, correct choice
Every settlement this archive publishes is worth exactly as much as its weakest one. Produce a verdict on a claim that genuinely could not be decided, and you have not added a data point — you have devalued every other verdict, because now they all carry the possibility of having been forced.
The resolver here is deliberately dumb, and that is the feature. The deadline arrives, the locked source is read, the locked operator is applied, and the answer is met, not met, or undetermined. No model opinion, no editor, no vote. When the evidence will not support a stranger re-deriving the outcome, it stays open with its test public so anybody can settle it themselves and show their working.
What it means when you see it
An undetermined verdict is not a mark against the person who predicted. It is usually a mark against the wording — or against the world, for removing the referee.
And it is information. A forecaster whose claims frequently cannot be settled is telling you something real about how they write, which is the subject of a separate note. Half of what is on this record — 50 of 101 — was never scoreable in the first place. Pretending otherwise would have been the easier product and a worse one.
Every figure above is a count from the record on August 29, 2026, and the archive is exportable if you want to check the arithmetic yourself.
Also here: People who write predictions get scored. People who say them don't. · What makes a prediction testable · The goalposts problem · What a 96% accuracy score actually measures · How a track record lies without a single false statement · How to check whether someone actually predicted what they say they did · Everyone predicts in January and finds out in December