The most useful forecast a person can make — the one that prevents the catastrophe — is the one a falsifiable record can never contain. A worked example from a private warning issued to a government in 2025, and why the discipline of refusing to score it is worth more than the credit of claiming it.
There is a forecast I will never put in the ledger, never count toward the Brier score, and never call a hit. It may be one of the most consequential calls I have made. That is exactly why I refuse to score it. This essay is about the one kind of forecast a falsifiable track record cannot contain — the prevention — and about why the discipline of leaving it out is worth more than the credit of putting it in.
The paradox is old and it is brutal. A forecast that warns of a catastrophe, is acted upon, and so prevents the catastrophe, erases its own evidence. The event that would have proven it right never happens. Intelligence services live inside this problem: the counter-terror analyst who stops an attack has, by succeeding, ensured there is nothing to point to. Success and silence look identical from the outside. You cannot tell the analyst who prevented an attack from the one who cried wolf, because in both cases the outcome is the same — nothing happened.
In October 2025 I issued a private intelligence advisory to senior officials in the Indian government. It named the [Vaishno Devi shrine near Katra](https://www.youtube.com/watch?v=IXhscEBjYHc) in Jammu and Kashmir, a date in early December, a coordinated mass-casualty attack profile, and a defined validity window. I did not post it publicly at the time. I fixed its date on public infrastructure — a briefly-public timestamp — and then deliberately kept the substance private, so authorities had room to act without public panic and without tipping an adversary. I released the advisory openly only after the window had closed; that public release is the video linked above. The window came. The window went. No attack occurred.