Polls are always tricky, but I was looking closely at the margin of error. For the recent UK election and Brexit, the polls and the final results were both within the margin of error. The pundits read the numbers wrong.

In looking at the numbers here, the benefit of the doubt is two points on top of the margin of error. For example, Pennsylvania may end up being far tighter. But giving a two point benefit of the doubt on top of the margin of error is statistically an awful lot of room.

Of course, could be completely wrong (see the Bradley Effect). But in this case, I think the cushion on top of the polls is pretty significant.