Scorecard
Weight = centrality × boldness (1–9). Score: confirmed or supports 1, partly or mixed ½, refuted or contradicts 0. Final uses resolved claims; interim uses the latest supporting, mixed or contradicting evidence on open claims; neutral evidence and expired claims are not scored. Brier is the average of (probability − outcome)² over claims recorded with a confidence: 0 is perfect, 0.25 is what always saying 50% earns, lower is better. Follows the source and forecaster filters.
Calibration
Claims grouped by the probability given when they were recorded. Well calibrated means the share that came true is close to the probability in each row. It takes a few dozen scored forecasts before this means much.
Record claims
Write the test before you know the answer. For a metric test the current reading is stored as the baseline. Claims written here are permanent.
Record one claim
Import a set of claims from a file
A JSON array of claims with id, group, source, at, claim, test, refute, significance (centrality, boldness, reason), reviewBy, and optional forecaster, confidence (p from 0.01 to 0.99, basis own, stated or implied) and observations. Claims whose id already exists are skipped, never overwritten.