Improved evidence on the most common and severe types of head impacts in ice hockey can guide efforts to preserve brain health through improvements in protective gear, rink design, player training, and rules of play. In this observational cohort study of men’s university hockey, we compared video evidence on the circumstances of 234 head impacts to measures of head impact severity (peak linear accelerations and rotational velocities) from helmet-mounted sensors (GForceTracker). Videos were analyzed with a validated questionnaire, and paired with helmet sensor data. Shoulder-to-head impacts were more common than hand- or elbow-, but there were no differences in head impact severity between upper limb contact sites (p ≥ 0.2). Head-to-glass impacts were nearly four times more common, and just as severe as head-to-board impacts (p ≥ 0.4). Head impacts resulting in major penalties (versus no penalty), or visible signs of concussion (versus no signs), involved greater head rotational velocities (p = 0.038 and 0.049, respectively). Head impacts occurred most often to the side of the head, along the boards to players in their offensive zone without puck possession. Head impact severity did not differ between cases where the head was (versus was not) the primary site of contact (p ≥ 0.6). Furthermore, penalties were called in only 4% of cases where the head was the initial point of contact. Accordingly, rules that focus on primary targeting of the head, while important and in need of improved enforcement, offer a limited solution.