What's noteworthy about 2020 vs 2016 was in the variability of the race. The timing with Comey at twenty-one days out made that obvious, as Clinton's numbers varied dramatically throughout the race. Clinton also ran a more divisive race, and was a far more controversial candidate. Trump was at 35% on 538 a week before the election and while he dropped dramatically, that number was sufficient for a lot of mail-in and early election ballots to be cast. Had the election been a week later, Clinton would have won the EC, assuming no other shenanigans.
Silver did in fact predict that an upset was entirely possible, but more to the point, he also made it clear that the Margin of Error on the race was very high. The 2020 MoE is looking to be much smaller, given the degree to which the race has barely budged.
My expectation is, given both the fact that there are more mail-in ballots and longer voting periods (good things, both), there's going to be a significant locked in advantage for Biden even IF there is an October surprise of some sort. I also suspect that Nate's model may actually be biased towards Trump precisely because of the 2016 election.
I agree, none of this should be taken to say we shouldn't get complacent. I fully expect that you will see every attempt made to cheat, especially at the Senate level, by the GOP. They're not even attempting to be subtle about it anymore. However, I think that the very explicit nature of that cheating is actually backfiring.