[RodericDay]

I think "data science" has really let science down. 99% win predictions for Clinton, media darling Nate Silver getting the primaries completely wrong and saving face a little bit in the election, etc.

As a scientist, I instinctively trusted other data scientists to be as rigorous as I was when I did research. It turned out they were using extremely wonky simplistic models for complex human behavior, and it left me cold and clueless.

If poll-based divination is a better descriptor than political data science, then scientists should decry their stealing of the term.

[Uehreka]

I think it's important not to overlearn the lessons of this election wrt data and predictions. Most national polls were off by only a point or so. The polls in WI, MI and PA were off by a few more. As Nate said on a recent podcast "If you give me a national opinion poll, I'm inclined to believe it. If the poll is specifically from Wisconsin, I may be more skeptical."

Of course, some people built bad models and predicted 90+% chance of winning for Hillary, and those people need to rethink their approach. But this election was not a refutation of polling and modeling as a whole.