Do the prices come true?
Every settled print, grouped by the price actually paid for it and compared against how often that leg really won.
Actual vs price paid
Calibration error by price
About this page — read before quoting any number
The x-axis is the price actually paid, not the band midpoint. The published calib_error and clears_2se columns measure each band against its midpoint, which is the convention the other venue files use and is kept so they stay comparable. In the tails the midpoint is nowhere near what anyone paid, so everything here is recomputed against sum_price_contracts / n_contracts.
One game is one observation, not one per print. Thousands of prints on a single ball game settle on the same result. Every interval is clustered on the event — field 4 of the DKeX symbol, shared across every market type and period on that game, so a player prop and the moneyline on the same game go in the same cluster. Treating prints as independent would shrink these bars by roughly 3–7×.
Voids are excluded, and that exclusion is load-bearing. A refunded event has no outcome: both legs of a voided or postponed event settle at $0.50. Void prints, pro-rated partials and not-yet-settled markets are removed before binning — about 0.43% of the tape. Before the 2026-08-06 settlement fix, voids were scored as wins, which alone moved the 10–20¢ band from −0.18¢ to +4.01¢.
Whose price. A binary has two legs. DKeX prints one price per row and the symbol names the specific leg, so the x-axis is that leg's own price and the win rate is that same leg's own settlement. DKeX publishes no aggressor flag, so this is not a taker curve; Kalshi bins the taker's own side, which is the one place the two series are not strictly comparable.