ID doesn't "predict" that at all. For a scientific theory to "predict" something it has to have a mechanism or mechanisms and/or to make reasonably specific empirical claims. To "predict" means that the operation of the theory's mechanism and/or deducible implications of the theory's empirical claims entail consequences that can be observed.
ID has no mechanism. It asserts that there is, by inference, an "intelligent designer" [or designers?] but it refused to describe (or even speculate about) this designer in any way which would usefully characterize it as a theoretical mechanism. Thus ID makes no empirical claims. It won't say (or even suppose) either when, where or by what means the designer actually instantiates design events.
IOW your claim about this "prediction" is exactly backward. The existence of complexity is not an implication of ID, but simply an observed fact (and almost the only observed fact) from which ID is inferred. However proponents of ID falsely imply that these kinds of complexity uniquely implicate ID, which is false. There are well known mechanisms, for instance, whereby ordinary evolution (i.e. preceding in small, progressive steps, each individually viable) can produce "irreducibly complex" systems.
I agree with you completely here. A mechanism in a scientific theory must describe observable, physical phenomena.
That said, and in accordance with the terms given here, what physical event does natural selection explain, that isn't already explained by other mechanisms such as mutation, drift, and recombination...each of which, in and of themselves, do not infer evoution.