News for hackers. Stuff that matters.
Re:Facebook Must Really Suck At Machine Learning by dragonwriter on Friday November 18, 2016 @08:15PM (#12991416)
No mainstream prediction I saw predicted a massive Clinton landslide; most predicted a slightly larger popular vote victory than actually occurred.

Some, though not the best regarded, gave a Clinton victory a very high probability, largely because they modelled state-level variability as independent rather than highly correlated, an error that 538 called out explicitly.

Re:Facebook Must Really Suck At Machine Learning by rdtsc on Friday November 18, 2016 @08:35PM (#12991483)
Most were also biased in her favor. That was the big learning experience for me in this election. She had an unprecedented support from media, unlike few other candidates. I think even Fox (the traditional Republican faux-news channel) didn't like him very much.

Yeah I don't know the answer, what the mistake was. I'd want to know. I suspect some never responded or wanted to participate. Those who participated, said they'd vote for someone else but changed their mind. They used registered phone numbers and some people more consistently paid to have their number de-listed...

Re:Facebook Must Really Suck At Machine Learning by hga on Saturday November 19, 2016 @07:07AM (#12993114)
Yes, from what I've read Fox was very anti-Trump, aside from a few of its figures who's contracts have not inspired.

I mean, this election was used as an opportunity to mount a successful coup ousting its founder, who'd, BTW, announced a couple of years ago that the network was moving to the left anyway....