Males that behave violently to achieve social and reproductive objectives are a widespread phenomenon among mammals, as well as humans. Because this kind of behaviour can be socially very disruptive, its continued survival in human populations, in particular, remains a puzzle. We use historical data on the pedigrees of medieval Icelandic Vikings to test the hypothesis that males who killed had higher fitness than other males. Whereas most studies that examine the evolutionary benefits of behaviour focus on the numbers of offspring sired, in this paper we measure fitness more directly in terms of the number of grandchildren produced both directly and indirectly via collateral relatives, as well as determining the costs of pursuing alternative strategies in terms of experienced mortality rates. We show that, on average, killers gain a very significant fitness advantage despite the often high costs they pay and, more importantly, that they had a dramatic effect on the fitness of their male kin. We suggest that such behaviour represents a phenotypic, rather than genetic, response to opportunities provided by local circumstances.