As the amount of data, facilities, and tools for understanding students' programming process are improving, the time is ripe for analyzing students' actual programming process. In our current work we are investigating how students' behavior during her programming process (e.g. eagerness to start working on freshly released exercises, following best programming practises) affects the course outcome. We purposefully utilize only data gathered automatically using snapshots from the students' programming process, and do not gather any additional background information. Currently, we are able to predict whether the student is a high-performer, passes the course, or fails the course with a 78% accuracy.