Microlearning, in which lecture recordings are segmented into parts, saw renewed focus as a means of maintaining student engagement amid the challenging conditions of the COVID-19 pandemic. While many institutions shifted to remote provision with segmented lecture recordings, there is a lack of consensus about the length that these segments should be in order to best maintain engagement. Using a self-reported system of Likert-based diagnostics, 135 videos in use at Solent University's computer games area were analysed. Ninety-four students were asked to agree or disagree with statements in the format 'I understand X', each tailored to the subject material of the video in question. Repeated questions before and after the video allowed for a change in confidence to be measured, as an indicator of engagement. The resulting 4198 responses showed an optimum range of 5-8 min overall. However, the year of study emerged as a significant factor in this regard -- with an optimum range for first years at 6-12 min, and for second and third years at under 8 min. There is a need for institutional-level change in this area, as many institutions currently recommend use of lecture video segments far longer than either figure.