For example, "selection" and "fitness" can apply equally to a case in which a beetle increases its fitness by a mutated enzyme that allows it to metabolise a previously toxic fungus, as to a tree that is able to increase its reproductive investment by growing slightly longer seed "wings" than its neighbours. Very little is common physically to these two cases even if the fitness or selection coefficient is in all respects identical.
But is evolution just a statistical observation? I think not - evolutionary explanations in terms of selection (and also of drift, for I do agree that they are both the same kind of process in this respect) involve rates of change. It involves complex behaviours of systems of organisms in their environment. It involves mathematical descriptions that have greater and wider application than the immediate context in which the model was developed.
There is a distinction in physical theory between a dynamic model, which indicates the forces active in a particular case under explanation, and a kinematic model which merely describes them. Mendel's model of inheritance is in this respect a kinematic account, as he did not describe the underlying physical processes, but rather the ways in which traits, phenomenal, even epiphenomenal, aspects of inheritance, behaved.
A kinematic model is a precursor to a fuller explanation. It is not merely a statistical description of a state, or even a sequence of states - but it is not a full explanation of the underlying physical processes either. It is, as someone I now cannot recall once said, a promissory note for an explanation. It is a necessary step in the development of the understanding of a domain.
It may even be that the same model, the same mathematics, can apply in various domains. For example, epidemiological models apply to computer viruses as well as Ebola, and to demographic trends in culture as much as the progress of a pathogen. A Universal Darwinism such as Dawkins and others desire is, in the end, just a large-scale kinematic expectation. If culture is a process of Darwinian evolution, it is not because ideas have DNA, RNA or enzymatic expression.
I think therefore that an intermediate view is possible. Sober's force-theory can be recast as a theory of propensities or tendencies, and retain all the virtues of the force model. Walsh, Ariew and Lewens can reject the dynamic model of evolution and still not have to undercut the motivations for it in explanation. And we can learn something generally true about scientific, and in particular biological, explanation.