Abstract: Adjusted +/- (APM) has grown in popularity as an NBA player evaluation technique in recent years. This paper presents a framework for evaluating APM models and also describes an enhancement to APM which nearly doubles its accuracy. APM models are evaluated in terms of their ability to predict the outcome of future games not included in the model’s training data. This evaluation framework provides a principled way to make choices about implementation details. The enhancement is a Bayesian technique called regularization (a.k.a. ridge regression) in which the data is combined with a priori beliefs regarding reasonable ranges for the parameters in order to produce more accurate models.