Previous work has shown that during adolescence, classrooms vary greatly in the extent to which aggression is rewarded with popularity (the 'popularity norm'). Aggressive popularity norms may promote the proliferation of aggression and negatively affect the classroom climate. It is, however, unknown how these norms emerge in the first place. This longitudinal study therefore investigated whether aggressive popularity norms can be predicted by the classroom composition of students. We examined whether the prevalence of six student types - socially and non-socially dominant prosocial, aggressive, and bi-strategic adolescents (adolescents who are both highly prosocial and aggressive) - contributed to the norm by establishing a popularity hierarchy: strong classroom asymmetries in popularity. We collected peer-nominated data at three secondary schools in the Netherlands (SNARE-study; N-students = 2843; N-classrooms = 120; 51.4% girls; M-age = 13.2). Classroom-level regression analyses suggest that the classroom percentage of socially dominant aggressive and bi-strategic students predicted higher aggressive popularity norms, both directly and by enhancing the classrooms' popularity hierarchy. Instead, the presence of non-socially dominant aggressive students and socially dominant prosocial students contributed to lower aggressive popularity norms. Socially dominant prosocial students also buffered against the role of socially dominant aggressive adolescents in the aggressive popularity norm (moderation), but not against bi-strategic adolescents' role. Our findings indicate that interventions aimed at reducing aggressive popularity norms should first and foremost take the composition of classrooms at the start of the school year into account; and should not only encourage prosocial behavior, but also actively discourage aggression.