In this study, we discuss the application of K-means clustering technique on classification of NBA guards, including determination category number, classification results analysis and evaluation about result. Based on the NBA data, using rebounds, assists and points as clustering factors to K-Means clustering analysis. We implement an improved K-Means clustering analysis for classification of NBA guards. Further experimental result shows that the best sample classification number is six according to the mean square error function evaluation. Depending on K-means clustering algorithm the final classification reflects an objective and comprehensive classification, objective evaluation for NBA guards.