Calibration record

early period · n<30

Every number below is computed weekly from live resolutions only, with bootstrap confidence intervals (1000 resamples). When an interval crosses zero we say so; nothing here is a claim until it separates.

snapshot 2026-07-27 15:24 UTC · 27 resolved · 45 scored forecasts · 84 shadow scores · 6 abstained

Brier by verdict class

the core experiment: does the verdict class carry information about error?

Verdict	n	Brier	Bootstrap CI
CONSTRAINED	45	0.236	95% CI [0.176, 0.296]

ABSTAIN rows never publish a probability, so they cannot appear here; their value is measured below, through the shadow.

Value of abstention

an ungated shadow model forecasts every question; we compare its error on the questions we abstained from

shadow on published set

0.248

n=63

shadow on abstained set

0.285

n=21

abstention value (diff)

+0.037

95% CI [-0.088, 0.167]

Positive means the questions our gate refused were genuinely harder, even for an ungated model. The interval currently crosses zero: direction only, not yet a claim.

Layer vs shadow on the published set

same questions, same evidence pool; the only difference is the decision layer

diff -0.012 (layer minus shadow; negative = layer better) · 95% CI [-0.086, 0.066]crosses zero: not yet separated

Risk-coverage

verdict classes are the natural thresholds of a selective forecaster

Publish up to	Coverage	Risk (Brier)	n
CONSTRAINED	88.2%	0.236	45
CAUTION	88.2%	0.236	45

naive AURC 0.208 · with few points this is descriptive, not comparative