Evaluation
How future outcomes may resolve persisted research decisions.
#Declared horizon
The canonical track-record policy stores a resolution horizon. Evaluation must wait until that horizon and then use real stored observations for both covered asset and benchmark.
#Outcome record
An outcome should store start and end snapshot IDs, asset movement, benchmark movement, relative result, resolution time, and any integrity flags. The current release does not fabricate this table.
#Metric gate
Arena ranking remains disabled until at least 20 resolved decisions, 30 days of age, complete benchmark data, and no unresolved integrity flags exist.