Two recently proposed methods for predicting the optimal breed contributions to a new composite breed are discussed. It is shown that these can both lead to incorrect solutions, and a reliable method is proposed.