Accuracy

How well our probabilities match what actually happened. Each claim is checked against the primary source named in advance, in the 06:00 run on the date the edition bears.

10claims resolved

0.183Brier score (0.250 = coin toss)

70%calls that went our way

0.54log score (0.69 = coin toss)

Calibration

Each dot is a group of claims with similar probabilities. On the dashed line, our 70% calls come true 70% of the time. Dots below the line mean we were overconfident; above it, too cautious.

Forecast band	Claims	Average forecast	Came true
10–20 %	1	15%	100%
30–40 %	2	32%	50%
50–60 %	1	55%	0%
60–70 %	1	65%	100%
70–80 %	1	70%	100%
80–90 %	2	83%	100%
90–100 %	2	90%	100%

By section

Section	Claims	Went our way	Brier
Front page	4	2	0.284
Europe	1	1	0.010
World	5	4	0.137

Against benchmarks

Each row compares our Brier score with a benchmark on the same claims. Lower is better. A benchmark only counts where it existed when we froze the forecast.

Benchmark	Claims	Ours	Benchmark
Coin toss (50%)	10	0.183	0.250
Our informative claims only (15–85%)	8	0.227	0.250

Largest misses

Edition	Claim	Forecast	Outcome
23 Oct	The 2026 Sakharov Prize is awarded to the International Criminal Court (alone or jointly).	15%	Came true
23 Oct	Russia releases Stephen Hubbard, or his release is officially announced.	35%	Came true
23 Oct	The 2026 Sakharov Prize is awarded to Tundu Lissu (alone or jointly).	55%	Did not happen
23 Oct	The Datafolha runoff poll published on 22 October shows a gap of 2 points or less in valid votes.	65%	Came true
23 Oct	No US–Iran agreement based on the Qatar–Pakistan proposal is officially announced.	70%	Came true