Research Papers

Farrell, Sean; Laity, Ethan; Laughlin, Dave; Pennebaker, James W.

Risk assessment of potential recruits is of prime importance to all NBA franchises. Although most scouting focuses on physical performance statistics, there is general agreement that psychological factors also play an important role in determining success. Accurate psychological assessment of potential recruits can be difficult due to limited access to athletes, the time required to complete the assessments, and the self-reporting nature of traditional psychology questionnaires. In this paper we explore applications of language psychology metrics using machine learning and survival analysis techniques to predict success in the NBA. We found that we could predict which athletes would make it onto an NBA roster with an accuracy of 63% without any physical attributes included in the model. In contrast, a model built just using NCAA playing statistics achieved an accuracy of 78%, and combining physical statistics with the psychological features boosted the performance to 83%. Adding in physical attributes such as age, height and weight, along with the NCAA conference the athlete played in increased the accuracy further to 87%.