Rank uses realized historical outcomes, not AI probabilities. Each sample is measured from that group's own first call through +6H. 30D is the primary window; U means fewer than 30 evaluated samples.
30D realized performance · 6H after each group's first call · Minimum 30 evaluated samples for rank
Rank uses realized historical outcomes, not AI probabilities. Each sample is measured from that group's own first call through +6H. 30D is the primary window; U means fewer than 30 evaluated samples.