Currently, surface electromyography (sEMG) prostheses are characterized by low control capabilities and long training times. This is in contrast with recent advances in mechatronics, thanks to which mechanical hands have often many degrees-of-freedom and force control. Therefore, there is a need of techniques able to increase control capabilities with sEMG signals. Several reasons determine differences in the signal patterns, and make the classification of sEMG signals a challenging task. One of the reasons is the positioning of the electrodes on the subjects. In this paper we evaluate the positioning effect on the Ninapro database using automatic classification of the data for its evaluation.