The newest 538 Election Forecast page at
https://projects.fivethirtyeight.com/20 ... -forecast/ has one really good graphic that gives an idea of just how much uncertainty there is. I really do not think most people understand.
As of now the graphic looks like the image below. It's a sample of 100 simulations they ran. If you go to the page you can hover over each dot and see the results of the simulation for that dot. But here is the bottom line:
10 of the dots...or 10% since there are 100 of them...represent scenarios in which, if the election were held tomorrow, Trump would win by more than 77 electoral votes. So you would not be wrong in betting that if you had to bet that Trump would not win by more than 77 electoral votes. But, at the same time, I don't think you would say him winning by more than 77 electoral votes is "extremely unlikely."
Here's another way to look at it: If you constructed a 90% confidence interval for the outcome, it would be from Trump winning by 113 electoral votes to Biden winning by 368 electoral votes. So if you went by the conventions of statistics Trump winning by 77 electoral votes would not be considered remarkable at all. It is well within the range of reasonably expected outcomes.
The reason I reference 77 electoral votes is because that's the margin by which Trump beat Clinton. When he did that, the election eve 538 forecast showed 71% confidence that Clinton would win. The stuff I just discussed with respect to the current 538 display below is associated with 79% confidence that, if the election were held tomorrow, Biden would win. Trump winning by 77 electoral votes in 2016 was even less remarkable than Trump winning by 77 electoral votes would be if the election were held tomorrow. It was not nearly the surprise that popular perception suggests. An upset, yes. An unbelievably unlikely outcome that should make people think the polls were completely whacked, no.
