Interesting discussion here. I don't think you're representing the author's point correctly, though, given the larger context of the work he's doing and current debates in the historical field. Here's what he says:
"I suppose the point of this post is to articulate my growing concern that we are so damn good at coming up with post-facto historical explanations to contextualize any given observation, that we are particularly susceptible to confabulating these post-facto rationalizations with the idea that we somehow knew the results of this quantitative work already."
The implication here, I think, is simply that historians should be open to multiple ways of testing explanations (i.e., combining qualitative and quantitative approaches, if those are available). And in the larger context of debates about digital humanities, he's swatting down the argument that DH simply tells us what we already knew, because if done right it can be used to add a new perspective on existing arguments. The point isn't that history should be science but that "scientific"/quantitative approaches can be used to test the validity of historical claims based on more traditional historical work, like finding manuscripts in archives, doing a close reading of a key text, tracing correspondence networks, tracing the material history of a painting or object, etc. I agree with him.
Edited to add: also need to be careful about confusing prediction with explanation. Historians don't predict much, but we certainly try to explain things a lot! I read this post as being exclusively about the latter.