Theory predicts that when individuals live in groups or colonies, male–male aggression peaks at intermediate levels of local average relatedness. Assuming that aggression is costly and directed toward nonrelatives and that competition for reproduction acts within the colony, benefits of aggressive behavior are maximized in colonies with a mix of related and unrelated competitors because aggression hurts nonkin often, thereby favoring reproduction of kin. This leads to a dome-shaped relation between male–male aggression and average relatedness. This prediction has been tested with bacteria in the laboratory, but not with organisms in the field. We study how male–male aggression varies with relatedness in the social spider mite Stigmaeopsis miscanthi. We sampled 25 populations across a wide geographic range between Taiwan and Japan, representing a gradient of high to low within-population relatedness. For each population the weaponry of males was measured as the length of the first pair of legs, and male–male aggression was tested by placing pairs of nonsibling males together and scoring the frequency of male death over a given period. As these two morphological and behavioral variables correlate strongly, they both reflect the intensity of male–male conflict. Our data on the social spider mite show that male–male aggression as well as weapon size strongly peak at intermediate, average relatedness, thereby confirming theoretical predictions.