Twin studies suggest a substantial role for genes in explaining individual differences in aggressive behavior across development. It is unclear, however, how directly measured genetic risk is associated with aggressive behavior at different moments across adolescence and how genes might distinguish developmental trajectories of aggressive behavior. Here, a polygenic risk score derived from the EAGLE-Consortium genome-wide association study of aggressive behavior in children was tested as predictor of latent growth classes derived from those measures in an adolescent population (n=2229, of which n=1246 with genetic information) and a high-risk sample (n=543, of which n=335 with genetic information). In the population sample, the polygenic risk score explained variation in parent-reported aggressive behavior at all ages and distinguished between stable low aggressive behavior and moderate and high-decreasing trajectories based on parent-report. In contrast, the polygenic risk score was not associated with self- and teacher-reported aggressive behavior, and no associations were found in the high-risk sample. This pattern of results suggests that methodological choices made in genome-wide association studies impact the predictive strength of polygenic risk scores, not just with respect to power but likely also in terms of generalizability and specificity.