Shortfall minimization and the Naive (1/N) portfolio: an out-of-sample comparison