Score

Forecasting firms, model vendors, AI labs and venues all publish probabilities. Scopes scores them against outcomes the way it scores the markets: Brier and skill by horizon, calibration, with intervals, under rules fixed before outcomes. The matched question set, the rules for matching, and the scoring rules are published before any result, with timestamp receipts, so neither side can choose the comparison after the fact. Scores are published with attribution; the producer's raw probabilities are shown only if the producer allows it.

Producers of probabilities who want an independent score they can cite: superforecasting firms, election and sports models, model vendors. Evaluators who need the venue baseline, including AI labs measuring forecasting models. Venues whose self-reported accuracy claims would carry more weight with a third party behind them.

A scoring methodology document for the engagement, the matched set published before outcomes, scores by horizon with intervals on a standing page, and a scoring report the client may cite. Independent verification, not an audit.

Per engagement for a one-time scoring. A standing fee for continuous scoring on the record. Research use by arrangement.