Elite military programs such as the 75th Ranger Regiment's Ranger Assessment and Selection Program (RASP) see rates of attrition often in excess of 50%, and amplify the need to identify and screen candidates based on their probability of successful matriculation. Models were developed (and cross-validated) to predict attrition from RASP using the physical abilities, cognitive abilities, and personality scores collected during candidate admissions screening. We report both regression weights and standardized odds ratios for optimum models of candidate success over three program timeframes to enable an understanding of the relative importance of each predictor. In spite of physical abilities scores being used to select RASP candidates, they were the strongest predictors of RASP attrition. Personality scores accounted for more variance in predicting candidate success than cognitive ability scores. Personality predictors, especially dimensions related to Openness, were better at predicting week one attrition than attrition in later weeks. The use of a single, aggregated candidate probability score for making admissions decisions is discussed, along with additional practical and scientific implications.