Forecast Calibration: Predicted vs Actual

Every macro scenario Convex has tracked, with the probability at peak conviction, the final reading before resolution, and the recorded outcome. The point is accountability: if the AI editorial system is wrong, the record shows it.

Methodology: scenario probabilities are produced by the Bayesian engine described on the scenario methodology page. A scenario is "resolved" when the underlying event has occurred or the resolution date has passed without it occurring. Both directions are logged here.