Comment by dmurray

The abstract does say

> limitations, principally the small number of tournaments available for validation and the risk of in-sample weight selection

But I agree this model is no more valuable than Paul the Octopus.