01 / ACCURACY
Brier score
(predicted probability − outcome)²
The squared distance between the forecast and reality. A perfect prediction scores zero.
A OnePercentBetter project
Grading the closing lines of the world's prediction markets.
A spectator’s scoreboard for the markets that predict the world.
Which prediction market is sharper? We freeze their probabilities before an event resolves, then grade them against what actually happens. Category by category. Season by season.
The same outcome. Two sets of odds. A score neither venue gets to rewrite.
Use probabilities from 0 to 1. The outcome is 1 if the priced subject wins, and 0 otherwise. Lower scores are better.
01 / ACCURACY
(predicted probability − outcome)²
The squared distance between the forecast and reality. A perfect prediction scores zero.
02 / CONFIDENCE
−ln(probability assigned to the actual outcome)
Punishes confident wrong calls much harder. Being nearly certain and wrong carries a steep cost.
03 / RELIABILITY
Stated probability ↔ hit rate
Do 70% calls win about 70% of the time? After results, 10-point probability bins will compare forecasts with hit rates and sample counts.
Polymarket vs Kalshi. Fourteen Senate races, one control test, and the probabilities frozen before the verdict.
Open the showdown