> The first sentence says Hillary has a "98.2 percent chance of winning the presidency". It only gets better from there. That is why physicists demand 99.99% (

> The first sentence says Hillary has a "98.2 percent chance of winning the presidency". It only gets better from there. That is why physicists demand 99.99% (three sigma) significance before they are calling it hints. The thing is, it is absolutely true that (some) polling models predicted a 98% chance of an Hillary win. It is just the problem that these models have trouble to predict the behavior of Trump voters. And frankly that is the problem with models. I have predicted, on reddit, that Trump would win. And I managed to do that precisely by not having a model, I just looked at the model outputs of Real Clear Politics and added a fudge factor, because usually models underestimate right wing populists, for perfectly understandable reasons. However, having a model means that you can not just add in a fudge factor, honesty demands that you have to justify each and every parameter and actually all forecasts where I did bother to look, did that. But the thing is at rather uncertain predictions like 98% the fudge factors reign.