Children with ADHD often display aggression. Ascertaining aggressive episodes in an objective manner reduces the need for relying on informants and maximizes the opportunities for targeted interventions. As such, this study sought to examine whether physical activity (PA), quantified using a wearable sensor, can reliably predict aggressive behavior in children with ADHD. Children diagnosed with ADHD (n = 11; aged 8.5 ± 1.0 years; 27% female; BMI: 16.7 ± 1.5 kg/m2) participated. A wearable sensor (ActiGraph GT3X+, Pensacola, FL, USA) attached to the waist was used to remotely monitor PA per hour for 7 days. Parents completed the Retrospective-Modified Overt Aggression Scale (R-MOAS). Five PA variables, including light PA (%), moderate PA (%), MVPA (%) (ie, averaged sum of moderate and vigorous PA), steps per minute, and average steps, were extracted from the wearable sensor using ActiLife (Pensacola, FL, USA). To compare PA level between hours over the no-aggression episode and aggressive episode, and at 1, 2, 3, and 4 hours before the aggressive episode, all PA variables were assessed statistically using generalized estimating equations analysis. All participants exhibited aggressive outbursts at least once (mean number of outbursts = 6.3 ± 7.3). All PA variables were significantly higher during the aggressive outbursts compared to other times when no aggression occurred (p < 0.0001). Multiple pairwise comparisons revealed that all PA variables at 1 hour and 2 hours prior to the aggressive outbursts were significantly higher compared to times outside the aggressive outbursts but no difference to times when an outburst was occurring. Our data suggest that tracking PA variables may enable identifying an aggressive outburst with a 2-hour lead time. Future studies are needed to further validate these findings and to examine their potential to improve treatment of aggression.