This thesis presents a procedure to collect, process, and analyse data for use in machine learning models within the rifle marksmanship domain. The machine learning model, features, and analysis presented in this thesis provide a first step towards models that can provide automated assistance for rifle marksmanship practice.A quasi-experiment is designed with eye movement and rifle handling as independent variables, and shooting result as dependent variable. The data is collected by letting 14 participants with different levels of marksmanship experience perform a rifle shooting exercise. Eye movement data is gathered using a Tobii Pro Glasses 3 eye tracker, and rifle handling data is gathered using an iCubeX Orient3D IMU. Analysis from the data shows a stratification of participant rifle marksmanship experience into five classes. The highest performing group, to a larger degree than the other groups, aimed with both eyes open, has a faster rifle acceleration, has a lower time between peak motion and shot, and are overall more consistent with their shots. A prototype random forest classification model trained to predict these classes shows a Cohen’s Kappa of 0.526, balanced accuracy of 0.599 and a one-vs-all AUC between 0.83 and 0.95 for the classes.