Why FiveThirtyEight Gave Trump a Better Chance Than Almost Anyone Else | Hacker News Reader

Why FiveThirtyEight Gave Trump a Better Chance Than Almost Anyone Else

fivethirtyeight.com

15 pointsmbgaxyz8 comments

They were pretty clear about the level of uncertainty leading up to the election. They said the data they were seeing and Donald Trump winning could both occur in a random universe slightly more than flipping a fair coin twice and both times getting tails (29%).

Probability exists to quantify uncertainty, not to predict the future (though this can help). One thing is clear - we don't live in a random universe, and there are concrete biases present in the way we collect election data. FiveThirtyEight carefully avoids and accounts for many of these sources of uncertainty. For example, herding of polls at the end of an election and the unusually high percentage of undecided voters at the end of the election.

Should be titled: "How FiveThirtyEight Desperately Tries to Regain its Credibility"

This!

I was tracking their polling aggregator in "real time" daily and all it did was give me a false sense of security. Now we have Trump.

If someone gives you a 1/3 to 1/4 chance of something bad happening, relying on that as "secure" seems ... risky. Criticism of specific mistakes they made are good and well, but just saying they predicted the wrong percentage (when we have no good idea what the real one was) seems to me like faulting somebody because they lost a coinflip.

The bigger issue IMHO is how much attention is given to polls and predictions, when they try to amplify useful data from tiny differences in key states. Results that indicate "we can't really know" seem likely, but make for really bad headlines. And worst case, some people actually think 70:30 means they can stay at home and not go vote or something...

As I said in another thread[1], I think this is the psychological explanation for a lot of the heat Nate Silver took before and after the election (being "wrong" both ways according to people who don't understand what a 70% chance means).

People saw 70% and rounded up to "Hillary".

For a psychological counterbalance, imagine your doctor tells you you have a 70% to be alive this time next week. Yet in the political realm people were calling that a "sure thing".

[1] https://news.ycombinator.com/item?id=12937736

I'm confused. Was their 29% chance "wrong"? Would a 40% chance still be wrong? 45%?

In general, people don't have such good intuition when it comes to probabilities. Their heuristic is that if the predicted outcome doesn't match the actual outcome, you are wrong.

This whole election, every time Nate Silver made a prediction about Trump, it turned out to be wrong.

That's a 100% failure rate.

Now, they're trying to claim that they were the least wrong? LOL.

Reply on news.ycombinator.com