Weight schema selection by domain-specific track record

prioritize schemas that have produced accurate predictions in structurally similar problems over schemas that feel familiar or recent.