Methodology and API

Every covered market is scored on five independently computed factors. Factor readings are cross-normalised within an asset class, then combined using published weights. No factor reads another factor's output, so a failure in one module degrades that factor only.

A market enters the compute set once it holds sustained notional above the coverage floor for seven consecutive days across at least one qualifying venue. Markets that fall below the floor are marked stale rather than removed, so their history stays intact.

Weights are fixed for the life of a scoring version. When weights change the version increments, every covered market is recomputed, and prior scores remain queryable under the previous version.

Given the same normalised inputs and the same scoring version, a computation reproduces exactly. Every published score retains its full factor attribution and the identifier of the computation that produced it.