Source credibility leaderboard
Which sources actually predict well? We rank news outlets, think tanks, and analyst feeds by the Brier-improvement of forecasts that cited them. Updates nightly. Sources with insufficient sample size or CI overlapping zero are not ranked.
How to read this
- Predictive value = (avg Brier of predictions not citing this source) − (avg Brier of predictions citing this source). Positive = source improves accuracy.
- CI is a 95% confidence interval. If the CI overlaps zero, we can't claim above- or below-baseline; the source is classified Neutral.
- Sample is the number of resolved forecasts that cited the source. Minimum 10 to be ranked.
- Decay: source quality decays 0.95/month if no new resolved data, so the leaderboard reflects current usefulness, not legacy reputation.
Loading leaderboard…
Want this for your team?
Drop your email and we'll be in touch when access opens for your vertical.