It is assumed that predators mainly prey on substandard individuals, but even though some studies partially support this idea, evidence with large sample sizes, exhaustive analysis of prey and robust analysis is lacking. We gathered data from a culling program of yellow-legged gulls killed by two methods: by the use of raptors or by shooting at random. We compared both data sets to assess whether birds of prey killed randomly or by relying on specific individual features of the prey. We carried out a meticulous post-mortem examination of individuals, and analysing multiple prey characteristics simultaneously we show that raptors did not hunt randomly, but rather preferentially predate on juveniles, sick gulls, and individuals with poor muscle condition. Strikingly, gulls with an unusually good muscle condition were also predated more than expected, supporting the mass-dependent predation risk theory. This article provides a reliable example of how natural selection may operate in the wild and proves that predators mainly prey on substandard individuals.