538 isn't using pseudoscience, obviously. This is probably their biggest miss ever (b_b or someone may be able to weigh in) but statistics is a real field. Unlike psychology. Anyway, it may be time to pay attention to the primaries. I'm still tagging this #sillyseason.
I'm pretty sure the statistical fuck up came from the fact that Michigan is an open primary and Sanders got a disgusting amount of independent voters in michigan. I'm not 100% sure (my knowledge on pollsters is limited to interactions with a few professionals), but they probably didn't account for the independent numbers.
I wonder how many of the GOP voters decided they just didn't want HRC, and thus voted for Bernie. It's interesting that he not only won very liberal counties, such as Washtenaw, but he swept the most rural and conservative counties. Either that, or rural Dems prefer the Bern.
I don't think thtis would be much of a surprise really. Especially now, after he swept Nebraska and Kansas. His policies definitely cater towards the lower class, but he's had issues connecting with minorities. So, to me, it makes sense that rural voters, being generally of lower income levels than urban voters, would go for Sanders over Clinton.rural Dems prefer the Bern.
Still, election statistics comes from polls. Silver's brilliance was that he didn't take polls at face value, he measured how they were measuring, as well as what, in order to turn them into psychohistory a la Hari Seldon. If you haven't read his book, you should.