This paper examines the effect of a high school student's relative position in the class achievement distribution on school violence. We identify this effect by exploiting idiosyncratic differences in the distribution of earlier academic achievement across classes. Such differences generate quasi-random variation in rank for students with the same initial achievement. We consider distinct types of school violence, namely, verbal, relational and physical violence. We find that rank has a negative effect on both the probability and frequency of perpetrating school violence for all the specific types of violence considered. The effect size is economically significant, especially in the case of physical violence. We find that rank is less or not effective in reducing physical violence for low-background students, migrants, in lower-quality schools and in high-crime areas, consistent with the lower perceived opportunity costs associated with misbehavior for disadvantaged students in low quality schools and located in violent local contexts.