A feasibility verdict is not a point prediction. It is a structured answer with four parts: the verdict class, the margin to the constraint boundary, a confidence interval, and the assumptions the verdict rides on.
ARIAA logs every verdict at issuance. When the outcome arrives, the platform computes the Brier score against the verdict and records it against the calibration table. Over time this record is the asset.
Related
- Calibration (Brier Score) — The discipline of grading probabilistic forecasts against realised outcomes. Low Brier score = well-calibrated.
- Computational Feasibility — The class of problem that asks, under live constraints and uncertainty, whether an outcome is still reachable — and with what confidence.