Research supports interventions for high-risk juvenile offenders to reduce recidivism. Methods for assessing delinquent risk vary, however. Aggregate risk scores (i.e. number of risk factors) and specific risk profiles (i.e. types of risk factors) are both empirically supported techniques. This study compared aggregate scores versus profiles for predicting measures of criminal severity among detained adolescents (n=292). Twenty-four risk factors from the Structured Assessment of Violence Risk in Youth (SAVRY) were summed to calculate aggregate scores. Using latent class analysis (LCA), profiles were identified based on scores from the following theoretically important SAVRY risk factors: Risk Taking/Impulsivity, Anger Management Problems, Low Empathy/Remorse (CU traits), and Attention Deficit/Hyperactivity Difficulties. LCA identified one low-risk profile, plus two high-risk profiles differentiated by levels of CU traits. Aggregate scores significantly predicted four out of six criminal severity indicators, while profiles failed to predict any measures. Results support aggregate scores over profiles for assessing delinquent severity.