Cohort
3Repeated experimentA group of operators measured together for comparison — the unit of comparative analysis.
Last updated: 2026-09-01Spec: SigRank Standard v1.0
Definition
A group of operators measured together for comparison. The unit of comparative analysis.
Inputs
Operator identifiers + their token pillars + cohort membership.
Derived variables
Cohort-level distributions, percentile ranks, quartiles, IQR fences.
Claim
Cohort comparison reveals relative performance — an operator's metrics are only meaningful in the context of a reference cohort.
Test
Compare operator metrics against cohort distributions (medians, quartiles, IQR).
Observable
Percentile ranks, z-scores relative to cohort, outlier flags.
Falsifier
If cohort composition does not affect relative rankings, cohort is not a meaningful unit.
Evidence
Field-analysis schema provides medians, quartiles, and IQR fences for yield, SNR, leverage, velocity.
Limitations
Cohort selection bias — small cohorts produce unstable percentiles.
Lineage
SigRank field-analysis methodology, leaderboard percentile computation. Architecture: mos2es.com/architecture.