How the Call has done.
Every question the Call put a number on before its date is scored when it settles, and the machines are scored on the same questions. No score prints until 10 results carry a number recorded before the date; until then every result is listed below, so the record can be checked line by line.
4 answered yes or no; 572 more questions with words approved before their date settle by December 31
the score prints at 10; 7 to go
0 of 4 settled: the source did not answer the question
dated questions that reached their date with no approved words: shown, never rated
The score
Brier score: lower is better; 0 is perfect3 of 10 results so far carry the Call’s number recorded before the date. The score, the machines beside it and the Call’s calibration by band print from the 10th; a score over fewer results says more about luck than about the Call.
Every result
What each one did to the Call| Settled | Question | Answer | The Call | The machines | Gap² | Words |
|---|---|---|---|---|---|---|
| Oct 25 | PJM home power prices up 4+ points over the rest, to August? | yes | 82 | — | 0.032 | before the date |
| Oct 21 | Did Meta report Muse usage beyond downloads in Q3? | yes | 55 | 58 | 0.202 | before the date |
| Oct 19 | Japan's Treasury holdings higher in August than July? | no | 35 | — | 0.122 | before the date |
| Oct 10 | Investment-grade spread tighter on Oct 2 than in August? | no | — | — | — | after the date: not rated |
How it is scored
The Call gives each question a chance of yes, from 0 to 100, when its words are approved. The Brier score is the mean of the squared gap between that chance and what happened: 0 is perfect, and a constant 50% scores 0.25. A result counts when its words were approved before the date and the Call’s number was recorded before it; words written afterwards are shown and never rated, and an inconclusive result is counted and never rated. The machines are scored on the same results by the median of the numbers they had on record when the question settled. The season’s yes rate is known only in hindsight, which is why it sits beside the Call as a bar to clear, not as a rival. We score claims, not people.