Thirty years ago, Dawes, Faust, and Meehl (1989) argued that mental health professionals should routinely use statistical prediction rules to describe and diagnose clients, predict behaviors, and formulate treatment plans. Subsequent research has supported their claim that statistical prediction performs well when compared to clinical judgment. However, many of the things we thought we knew about statistical prediction have changed. The purpose of this literature review is to describe methodological advances in statistical prediction. Three broad areas are covered. First, while statistical prediction rules are valuable for criterion-referenced assessment (e.g., predicting violence, recidivism, treatment outcomes), they are valuable only for some norm-referenced assessment tasks (e.g., diagnosis but not describing personality and psychopathology). Second, statistical prediction is particularly prominent for the prediction of violence and criminal recidivism. Results from this area will be used to describe the validity of traditional clinical judgment, structured professional judgment, and statistical prediction. The results support the use of both structured professional judgment and statistical prediction. The effect of allowing professionals to override statistical predictions consistently led to lower validity. Third, issues in building statistical prediction rules are described, including the assignment of weights to predictors, the emergence of new statistical analyses (e.g., machine learning), and the role of theory. As research has progressed, statistical prediction has become one of the most exciting areas of psychological assessment.