Now for some poli-sci victory laps.

The 'citizen forecast', the favourite model of all quacks based on asking people "who do you think is going to win?", had Trump winning 334 Electoral college votes.

Helmuth Norpoth's primary model, the favourite of all bad-faith partisan journalists with too much time on their hands, which is based on performance in early primaries, had Trump with 91% chances of winning the election.

The IFAAR-University of Neuchatel's internet search model: also wrong. Also quackery.

Trafalgar Group's adjustment for "shy Trump voters" - basically adding 6 points to Trump - was also quackery. Colour me shocked.

Bela Stantic's media sentiment analysis: interesting, but nothing to do with forecasting. Quackery. Wrong. Bye.

The Democracy Institute, which had Trump ahead in the popular vote thanks to (1) low-turnout assumption, and (2) accounting for 5-6% of shy Trump voters is probably going to end up 7 points off on the national popular vote. Quackery. Never listen to them again.

Polls are imperfect (some much less imperfect than others, h/t @jaselzer). "Fundamentals" are vague.

But most things that have been tried to supplant them are charlatanism, and the fact that the people pushing them have PhDs doesn't make them any less charlatans.