The U.S. Army is interested in extending the application of intelligent tutoring systems (ITS) beyond cognitive problem spaces and into psychomotor skill domains. In this paper, we present a methodology and validation procedure for creating expert model representations in the domain of rifle marksmanship. GIFT (Generalized Intelligent Framework for Tutoring) was used as the architecture to guide development efforts and was paired with an Army marksmanship simulator that collects behavioral information through sensor technologies. The models were based on expert data from eight members of the U.S. Army Marksmanship Unit's Service Rifle Team. The goal is to establish validated models that serve as artificial intelligence assessment criteria for driving a self-regulated training environment. We review the techniques applied to the data for model construction, the trends found in the data that are generalized across each expert informed through cross-fold validation practices, and discuss how the models will be used for driving real-time assessment. Results support the utility of generalized expert models across the fundamental components of rifle marksmanship as outlined in U.S. Army doctrine.