Abstract
Free-throws are essential to basketball game outcomes, but there are no existing methods for training athletes with objective free-throw feedback. This study analyzes free-throws by measuring plantar pressure (PP) distributions during free-throw movements and proposes a new method for developing free-throw feedback using PP-based machine learning. PP data were collected from seven university basketball players who performed 50 free-throws. Most subjects had at least one PP parameter (e.g. maximum pressure, center-of-pressure velocity) that significantly differed between success and failure (p<0.05). Additionally, optimally separating PP parameters were unique to each subject. A trained computer was able to predict free-throw results of three follow-up subjects, alternately using support vector machines and k-nearest neighbors, with average classification rates of 0.69 and 0.63, respectively. These results suggest the possibility of new methods for developing free-throw skills using PP feedback.
