We can test our data, too, to see what would happen if the polls are underestimating Trump like they did in 2016.
What I found was Biden would still be ahead, even with a 2016 sized mishap.
The polls underestimated Trump by 1 point (RealClearPolitics) or 2 points (FiveThirtyEight) in the aggregate of the states we currently have polling from. Applying that 2016 bias to our current data, Biden would have a 6- to 7-point lead nationally.
Concentrating on just the competitive states, the polls undersold Trump by 2 points (RealClearPolitics) or 3 points (FiveThirtyEight). If the polls in the competitive states were off by as much as they were at the end in 2016, Biden would still be ahead in states like Florida, Michigan and Pennsylvania.