Prediction Market Calibration Scorecard

How accurate are Kalshi and Polymarket prices, empirically? We bucket every resolved market in the last 365 days by its final pre-close YES price (0–10%, 10–20%, ...) and compare the bucket's predicted probability to the realized win-rate. A perfectly calibrated venue sits on the diagonal.

Overall

Kalshi (n=49980) Polymarket (n=19856)
0% 0% 25% 25% 50% 50% 75% 75% 100% 100% Predicted YES probability Actual YES rate

Bucket detail (Kalshi)

Predicted bucketnAvg predictedActual YES rateError
0–10% 26136 1.1% 0.1% -1.1pp
10–20% 452 13.8% 2.0% -11.8pp
20–30% 229 24.0% 4.8% -19.2pp
30–40% 137 33.9% 7.3% -26.6pp
40–50% 90 44.1% 11.1% -33.0pp
50–60% 80 54.1% 31.2% -22.9pp
60–70% 68 64.7% 67.7% +3.0pp
70–80% 90 75.1% 85.6% +10.5pp
80–90% 203 84.9% 93.6% +8.7pp
90–100% 22495 99.0% 99.7% +0.7pp

Bucket detail (Polymarket)

Predicted bucketnAvg predictedActual YES rateError
0–10% 9873 1.1% 0.1% -1.0pp
10–20% 764 13.8% 0.8% -13.0pp
20–30% 432 24.0% 2.5% -21.4pp
30–40% 279 34.3% 3.2% -31.0pp
40–50% 263 44.3% 6.1% -38.2pp
50–60% 217 53.5% 10.6% -42.9pp
60–70% 131 64.1% 15.3% -48.8pp
70–80% 70 74.7% 35.7% -39.0pp
80–90% 69 83.8% 40.6% -43.3pp
90–100% 7439 99.6% 99.2% -0.4pp