The best shot game applied to networks is a discrete model of many processes of contribution to local public goods. It has generally a wide multiplicity of equilibria that we refine through stochastic stability. In this paper we show that, depending on how we define perturbations, i.e. the possible mistakes that agents can make, we can obtain very different sets of stochastically stable equilibria. In particular and non-trivially, if we assume that the only possible source of error is that of an agent contributing that stops doing so, then the only stochastically stable equilibria are those in which the maximal number of players contributes.