The approach of evolutionary games like prisoner’s dilemma, among others, is based on a consistent strategy. We propose an experimentally verified system where cooperation can arise out of two simple factors: mutation and inheritance. This system has a social dilemma property and allows each agent to set its own desired number of participants. There are two essential mechanisms: fluctuation through which mutation leads to cooperation, and natural selection which tends to promote cheaters and therefore disrupts the cooperation. It is shown in numerical simulations that the interplay between both mechanisms leads to an equilibrium and that no intentional strategies are necessary to establish and sustain cooperation. Thus, starting from a population of non-cooperating agents, natural evolution can end with a population composed of cooperating groups with the mean group size determined by the fluctuation rate and the pay-off function. A thorough analytical explanation of numerical results is provided.