. And the fact was that this election was a toss-up with too many variables to predict accurately.
> 71%/92%/98%/99% chance Hillary wins
> election was a "toss-up"
lol
High enough to be totally wrong.
You realize that probabilistic predictions, by definition, cannot be "disproven" just because the (projected) less-likely event occurred? 3 times out of 10 on average, an event with which we accurately predict will happen with 30% probability will indeed occur. We can estimate whether or not their predictions are likely to over-predict or under-predict certain events if we have enough data, but there is no reason to draw this conclusion from one event (Trump's election).
Of course, if someone predicts an event with 99.99% certainty and that event does not come to pass, one must take very seriously the idea that the prognostication was incorrect, but events with a 30% chance happen all the time (just check weather reports), so you'd need far more instances of over-prediction for the discrepancy to be statistically significant.
I think 538 indeed gave a very reasonable prediction of Trump's chances of election at 29%. The polls generally pointed toward an HRC victory, but with substantial variance and margin for error.
