CrisisBrief vs the market
A calibration curve proves we’re internally consistent. This proves something harder: that on the same questions, scored from the same information cutoff, our sealed forecasts hold up against liquid prediction markets. We match each forecast to a Polymarket or Kalshi market that resolves on the same criteria, pin the market’s price as of the day we sealed our call, and score both with the Brier rule when the question resolves. Every matched call appears here — win or lose.
The head-to-head record publishes once 20 matched questions have resolved — a thin record is a noisy one, and we won’t dress up small numbers.
Back to /track-record. Prices via the public Polymarket and Kalshi APIs; this is a descriptive accuracy benchmark, not trading advice.