Skip to content

Every call, graded

Track record

Sports has results so far, and one world-events question has been graded. Elections and the other world events are still open: we keep tracking them and grade each call once it's decided.

How we grade

We take the number we published before the game, race or deadline, and check it against the result.

Two questions: did the side we favored win, and was our confidence the right size? A miss at 90% counts for far more than a miss at 55%.

The scoring rule

We use the Brier score: the squared gap between our probability and what happened (1 if it happened, 0 if not), averaged over every call. 0 is perfect. Saying 50% every time scores 0.25.

Saying 80% on something that happens scores 0.04. Saying 80% on something that doesn't scores 0.64.

What we hold to

Only real calls
The record counts forecasts we published before the event, nothing run after the fact.
Fixed before the event
The method is set before a game or race. We don't tune it after.
Misses stay up
Every miss is published next to the wins.
Every figure sourced
Each number names the data it came from.