Hindcast

How accurate are prediction markets — and which won't resolve the way you think?

The independent record of how Polymarket and Kalshi markets are priced versus how they actually resolve — plus which open markets are most likely to settle ambiguously or be disputed. Calibration and resolution risk, scored. Not hype. Right now Polymarket's prices sit within 1.6% of reality on average (24h before resolution).

Calibration (ECE)
1.6%

avg gap · 1,072 markets

Favorite wins
66%

priced 61% · 107 races

Confident & correct
95%

at 95% confidence

Brier score
0.041

0 = perfect · context only

Polymarket calibration curve
Do markets priced X% actually resolve YES X% of the time? On the dashed line = perfectly calibrated. 24h before resolution.
Sharpest vs. softest categories
Lower ECE = better calibrated.
CategoryECE (avg gap)BrierMarketsCalibration
Crypto1.6 pt0.018221Sharpest
Politics2.3 pt0.054519—
Culture3.4 pt0.02595—
Sports4.2 pt0.044147—
Economy5.2 pt0.04540Softest
Riskiest open markets right now
Which open markets are most likely to resolve ambiguously or be disputed — scored by settlement language, complexity, and a 4.0% historical dispute rate. A free preview of the resolution-risk radar; 1,500 open markets scored.
MarketRiskWhy
Will Trump speak to Ahmed al-Sharaa in June?62highsubjective/interpretive language: considered, credible
Will Trump speak to Yoon Suk Yeol in June?62highsubjective/interpretive language: considered, credible
Will Trump speak to Elon Musk in June?62highsubjective/interpretive language: considered, credible
Will Trump speak to Xi Jinping in June?62highsubjective/interpretive language: considered, credible
Will Trump speak to Kim Jong Un in June?62highsubjective/interpretive language: considered, credible
Will Deutsche Bank fail by June 30, 2026?56highsubjective/interpretive language: considered, substantially
Will JPMorgan Chase fail by June 30, 2026?56highsubjective/interpretive language: considered, substantially
Will Scotiabank fail by June 30, 2026?56highsubjective/interpretive language: considered, substantially
Will HSBC fail by June 30, 2026?56highsubjective/interpretive language: considered, substantially
Will UBS fail by June 30, 2026?56highsubjective/interpretive language: considered, substantially

Showing the top 10. Premium scores every open market with full drivers and alerts.

Go deeper with Hindcast Premium

Per-category track records over time, both lead times (24h + 7d), venue comparisons, alerts, and data export.

See Premium