I think his methodology was fine and pollsters were shit. If you look carefully at the map, he basically called all the swing states hilary and trump took from one another. His model saw the holes in the pollsters data. I think to swing trump to winnning, he'd probably just have to adjust a weighting factor by fractions of a percent. Look at how close michigan, florida and wisconsin were. The tiniest of margins won it for trump in the end. Literally a handful of factories that had closed in the past decade between those states determined this race.

The scarier thing is that for the next 10-20 years, boomers in the rust belt are going to swing republican unless dems can get a Bernie. To me, whatever democrat wins the michigan primary, is their candidate above all else.

He should call it quits together with all other pollsters. All that predicting is at best a totally useless activity or even worse, it influences voter behavior. Bet on football games instead. At least that doesn't do damage to the country.