Polls don't mean anything though. We learned this lesson in the last election.
Most of those polls in the image were uneducated guesses by people working for the respective companies.
Of the two that weren't uneducated guesses (first and last listed respectively):
1. The Princeton Election Consortium (PEC) 99% prediction wasn't final. Their more conservative guess was 95%, and in a blog post after that the person who made the calculations said they had made a bad assumption about a particular variable, which led to such a high prediction percentage. They corrected this in a later post on the website. PEC's was inaccurate but only because of operator error.
2. The Five Thirty Eight (FTE) poll was completely accurate and that merely reaffirms FTE's fantastic track record with election predictions. I don't think you realize just how uncertain their prediction was:
They predicted that Annoying Orange had a better chance at winning than getting 2 heads in a row in a coin toss. These are fundamentally probabilities, and occasionally they'll go in the 28% instead of the 72%.
So in short, polls are an extremely helpful metric AS LONG AS YOU'RE GETTING THEM FROM SOMEONE WHO KNOWS WHAT THEY'RE DOING. (Like FTE) Because statistics is hard.
EDIT: phrasing because statistics is hard
Poll