We aggregate and archive historical data from Polymarket and Kalshi to evaluate real-world prediction accuracy. Our database tracks thousands of resolved markets and measures performance using Brier scores, hit rates, and time-horizon calibration analysis.
No other public database combines cross-platform prediction market histories with time-horizon accuracy measurement.
0
Resolved Markets Analyzed
0
Total Markets Tracked
0%
Directional Hit Rate
Sep 17, 2026
Last Updated
Refreshes hourly
—
Brier Score (2-week)
—
Brier Score (1-week)
—
Brier Score (1-day)
For each resolved market, we take the market's probability at 1 month, 2 weeks, 1 week, 24 hours, and 12 hours before close and compare it to the actual outcome. The headline Brier score uses the earliest available horizon with data. For recently tracked markets, this is typically 12 hours or 1 day. Over time, as we accumulate more historical data, longer horizons will become available.
For all open markets we observe and store probabilities at each time horizon. For each market we record the company or index name, sector, volume traded, and actual outcome — so we can report this data back to users.
Market: "Will AAPL beat Q4 earnings?"
1 month before close: probability = 72%
Actual outcome: Yes (AAPL beat earnings)
Brier Score: (0.72 - 1)² = 0.078
0 = perfect, 0.25 = random guessing, 1 = always wrong.
(prediction - outcome)². Measures how far the prediction was from what actually happened. Lower is better.
Percentage of markets where the prediction (>50%) matched the actual outcome. Directional accuracy.
Whether predicted probabilities match observed frequencies. If markets say "70% likely," the event should happen ~70% of the time.
Avg Brier Score
0.000
At best available horizon (lower = better)
Hit Rate
0%
Directional accuracy
Markets Scored
0
Resolved markets analyzed
Markets Tracked
0
Active + resolved total
How prediction accuracy changes at different lead times. The headline Brier score uses the earliest available horizon with data (currently best available). Markets become more accurate as the event approaches.
| Time Before Event | Sample Size | Brier Score | Hit Rate | Accuracy |
|---|
Bar shows relative accuracy (Brier score vs. 0.25 random baseline). Markets become more accurate as the event approaches.
Blue line: Brier score (lower = better). Green dashed: Hit rate (higher = better). Red dashed line shows the random-guessing baseline.
Points near the diagonal line indicate good calibration.
Lower Brier scores indicate better prediction accuracy.
Do higher-volume markets produce more accurate predictions? Volume indicates market liquidity and participant interest.
| Volume Bucket | Markets | Brier Score | Hit Rate |
|---|---|---|---|
| No volume data available yet | |||
Comparing prediction accuracy across different platforms. Horizon breakdowns show how each source performs at different lead times.
No source data available yet
| Sector | Resolved | Brier Score | Hit Rate |
|---|
| Company | Sector | Resolved | Brier Score | Hit Rate |
|---|