Several talent selection programs in elite sport schools are based on motor diagnostics for the purpose of recommending or transferring promising talents to general groups of sports; game sports, combat sports or endurance sports, and to more concrete sports such as gymnastics, skiing, or tennis. However, in most sports, the predictive value of such testing is unclear. PURPOSE: The aim of the talent prediction was to assign each individual of the Under-15 athletes to his own sport. METHODS: The sample consisted of N = 97 youth athletes from Shanghai Elite Sport school belonging to six different sports including basketball (n = 7), fencing (n = 23), judo (n = 20), swimming (n = 10), table tennis (n = 15), and volleyball (n = 22). The performance diagnosis took place between September 2016 and March 2017, and comprised eighteen anthropometric parameters, two motor tests on back strength and complex reaction speed, five physiological measurements of the heart rate at rest, vital capacity, systolic and diastolic blood pressure, and hemoglobin concentration in the blood. The prognostic validity of the morphological, motor, and physiological tests was determined using linear discriminant analysis and nonlinear neural networks (multilayer perceptron). RESULTS: The between-sports differences in a battery of generic anthropometric, motor, and physiological tests allow one to distinguish the young athletes’ talents according to their individual sport provenience. The linear and nonlinear statistical methods that were used in parallel to identify the most relevant talent characteristics of each of the six sports by means of the leave-one-out procedure reversely confirmed the quality of the results. CONCLUSION: All diagnostic methods exhibited medium to high validity to discriminate between the six different sports. The relevance of the eighteen body dimensions, five physiological measures, and two motor tests for talent identification was confirmed.