https://www.voterstudygroup.org/publication/political-divisions-in-2016-and-beyond
IDK this sort of thing is basically a trope at this point. I think straining to read the tea leaves of the electorate for latent libertarians can only get you so far. And over the past decade I'm not sure that process has aged well in any case.
While I do think people overstate the Voter Study Group plot- there is only so much you can say about the electorate from a two dimensional composite score from a series of responses to a set of 12 questions- the cited rebuttal by Karl Smith misunderstands what was done. The plot isn't trying to find the median among the responses, it was plotting their absolute scores on the scale. Why would we rescale to make it a relative scale when we are trying to determine how many people are actually libertarianish? Libertarianism isn't defined on relative terms, but on absolute terms.
I understand why some people are committed to the idea that there are a lot of latent libertarians. But the whole idea of "libertarian-ish" voters sort of is self-defeating to me.
A better rebuttal to the VSG plot is that it sort of isn't even measuring libertarianism to begin with, or at least can only do it weakly. But I still think the effort to find libertarianish voters is fairly doomed from inception.