This tweet is incredibly dumb.

Firstly you can’t just look at one candidate’s support in isolation, because usually there are significant numbers of third-party and undecided voters. Actually these were really high in 2016, so it’s pretty common for both the main parties to beat their 2016 polls, but the Democrats usually beat theirs by much more.

Secondly the 2016 election was on the 8th, not the 3rd, so it makes no sense to have a common x axis with the calendar date - “number of days from election” would be more appropriate. When you’re a long way out from the election, 5 days isn’t very much, but the difference between being 7 days out and 12 days out is pretty big, especially bearing in mind the high rates of early voting this time.

Thirdly there is no reason to believe that the swing state polls will overstate Biden’s support the same amount that they overstated Hillary’s support. Historically the errors have jumped around - it’s possible that they will be out even further in the same direction, but it’s also possible they will be out in the opposite direction from last time. So it’s hard to draw any conclusions from comparing the 2020 polls to the 2016 polls.