Shows how accurate Polymarket odds were at different points before they resolved.
Polymarket forecasts closely align with actual outcomes across all probability ranges.
| Bucket | Expected Percentage | Resolved Percentage |
|---|---|---|
| 0% | 2.6% | 2.4% |
| 10% | 7.8% | 7.6% |
| 20% | 17.9% | 17.1% |
| 30% | 27.6% | 26.8% |
| 40% | 37.3% | 35.4% |
| 50% | 47.4% | 44.6% |
| 60% | 57.6% | 56.5% |
| 70% | 67.5% | 68.2% |
| 80% | 77.2% | 78.5% |
| 90% | 87.8% | 86.9% |
| 100% | 97.9% | 95.6% |
Lower is better. Brier scores show how accurate predictions really are — not just if they were correct, but how close the prediction was to being right.
| Bucket | Brier score |
|---|---|
| <1K | 0.065 |
| 1-5k | 0.061 |
| 5-10k | 0.056 |
| 10-25k | 0.052 |
| 25-50k | 0.050 |
| 50-100k | 0.048 |
| 100-250k | 0.041 |
| 250-500k | 0.034 |
| 500k-1M | 0.022 |
| 1M+ | 0.016 |
Shows the percentages of all polymarkets that ended up resolving to "Yes" or "No".
All metrics are calculated from resolved polymarkets, using price snapshots taken 1 month, 1 week, 1 day, 12 hours, and 4 hours before resolution. Accuracy reflects how often the final leading outcome matched reality, and Brier scores measure the average squared error of predictions.
Learn more about Polymarket Brier Score.
A market counts as accurate at a given time if the outcome it favored then — the side priced above 50% — is the one that actually happened. "4 hr accuracy" is the share of resolved markets whose leading outcome four hours before resolution turned out to be correct.
The Brier score is the mean squared difference between a forecast probability and what happened (1 if the event occurred, 0 if not). It rewards being confident and right and penalizes being confident and wrong. 0 is a perfect score, always guessing 50% scores 0.25, so lower is better.
As the resolution date approaches, more information becomes public — polls close, games are played, data is released — and traders price it in. There is simply less uncertainty left a few hours before a result than a month before, so prices converge on the true outcome.
Many markets ask whether a specific thing will happen by a deadline, and most specific things don't. Multi-outcome events are also split into one Yes/No market per candidate or team, so a single event with ten options produces one "Yes" and nine "No" resolutions.
The figures are computed from historical Polymarket price data and the final resolution of each market, sampled at fixed times before resolution. The numbers on this page are a published snapshot rather than a live feed.