Eight forecasts on the record. The first verdict is in.

The forecasts

The probabilities below are retained from the original published snapshot. The range from 0.15 to 0.90 makes confidence visible; the reference line marks 50%.

50% — the coin-flip line

When they resolve

Resolution is front-loaded and then back-loaded: one early signal, followed by a cluster of verdicts at the six-month mark.

Scorecard snapshot · September 23

Four calls across the August ledger, September portfolio and latest weekly report are graded below, with one void shown separately. Updated September 23, 2026. The Treasury call uses its original published 0.72; the weekly report’s 0.62 was corrected. The Resident Evil probability and FedEx void are supplied in the weekly report; their original dated forecast records are not published here.

Brier score = (probability − outcome)², with yes = 1 and no = 0; lower is better. “Hit” means the more likely outcome occurred (P > 0.50 predicts yes; P < 0.50 predicts no; P = 0.50 is unclassified). The mean uses unrounded scores and excludes the void. Four outcomes do not establish calibration or forecasting skill. Of the original 18 forecasts in the August ledger and September 15 portfolio, 3 are graded and 15 remain ungraded. The Resident Evil miss and FedEx void are additional records introduced by the weekly report. New forecasts and ungraded calls are excluded from the mean; the weekly report’s three-call score excludes the August counter-tariff result.

The ledger

Original forecast fields with current grading status. No independently verifiable pre-outcome hashes or timestamps are published for this batch; this page does not prove that the original record was immutable.