See my link below. I was referring to MSNBC which actually published a 99% prediction. New York Times was at 85%. 538 at least had the dignity to stop doing predictions.
15% and 1% are not similar predictions. 15% is roughly the probability of throwing a D6 and getting a 6, that is, fairly unlikely but very possible. 1% is roughly the probability of throwing three D6's and getting either 6-6-6 or 5-5-5, i.e. extremely unlikely.
I'm not sure it's really very useful to assign probabilities to election outcomes, since they're rare, one-off, highly unpredictable events. But predicting that something will happen with probability 85%, and then it not happening, is not an inherently bad prediction. Iirc the 99% one was pretty stupid and made some obviously flawed assumptions.
Also I think 538 just haven't got theirs ready yet. There's no way Nate Pewter can resist building a wildly overcomplicated model that basically just outputs "yeah, the person who's leading the polls will probably win, but we don't know for sure" in a numerical form that he knows most people don't understand.
538 gave Clinton a 70% chance of winning on Election Day. It was 90% before the Comey Letter.
there wasn't anything wrong with the polls
See my link below. I was referring to MSNBC which actually published a 99% prediction. New York Times was at 85%. 538 at least had the dignity to stop doing predictions.
15% and 1% are not similar predictions. 15% is roughly the probability of throwing a D6 and getting a 6, that is, fairly unlikely but very possible. 1% is roughly the probability of throwing three D6's and getting either 6-6-6 or 5-5-5, i.e. extremely unlikely.
I'm not sure it's really very useful to assign probabilities to election outcomes, since they're rare, one-off, highly unpredictable events. But predicting that something will happen with probability 85%, and then it not happening, is not an inherently bad prediction. Iirc the 99% one was pretty stupid and made some obviously flawed assumptions.
Also I think 538 just haven't got theirs ready yet. There's no way Nate Pewter can resist building a wildly overcomplicated model that basically just outputs "yeah, the person who's leading the polls will probably win, but we don't know for sure" in a numerical form that he knows most people don't understand.