2 counterpoints: 1. buying data is expensive 2. running a fund is a completely different skill than building predictive models (back-office, compliance, etc.)