This paper describes a novel approach for segmenting surgical gestures in an eye trauma treatment simulator, based on the analysis of system and environmental events, generated in order to track user performance. Subdivision of surgical procedures into tasks & sub-tasks allows for application of more accurate metrics (optimized per task), which may more appropriately evaluate user performance. This approach may also be applied in other surgical contexts. For the specific case of eye trauma treatment this system can generate different classes of events that, when combined, may drive a state machine with the capability of subdividing the macro operation(s). The results of this process may be utilized to develop informative feedback to the user in the form of performance metrics and training guidance.