People compare 2020 to 2016 polls but miss the big picture. Here's 2016 vs. 2020 polling avg in key states.

PA
Clinton 46.3%, Trump 42.4%
Biden 49.6%, Trump 44.9%

MI
Clinton 44.5%, Trump 40.4%
Biden 49.7%, Trump 42.8%

WI
Clinton 46.4%, Trump 40.5%
Biden 50.3%, Trump 43.8%

Biden's lead is both larger than Clinton's lead but also he's at or near 50% in all of these states! Clinton and Trump were both lower in 2016, meaning there were a ton of undecided voters. We credit those undecideds for a late swing towards Trump for the win. We can actually...

Look at Clinton's final vote percentage in these states and her polling averages were spot on! However, her support was low enough that a full undecided swing to Trump made it not enough to win. Let's take a look at her polling averages in these states vs. actual vote.

PA
Polling avg: 46.3%; Actual: 47.5%

MI
Polling avg: 44.5%; Actual: 47.3%

WI
Polling avg: 46.4%; Actual: 46.5%

The polls didn't overestimate Clinton's support. Instead, they either (a) underestimated Trump's support or (b) couldn't foresee a last second undecided shift!

So IF polls are "off" like they were in 2016 (big if), Trump still really has very little opportunity to be ahead of Biden since Biden's support is so high. You'd actually need Biden's support to lower and then Trump to get almost all the undecideds for Trump to win.

Of course, the big old caveat to all of this is we're still 38 days out from Election Day and, even though polls have been strikingly stable, shifts could happen based on debate performances, October surprises, or other events.