There's no reason to believe that Election Day will be even more R though, when many of those Rs could easily be voting now in Early Voting, especially with turnout going over 2016's EV.
Not to mention, NPAs broke Trumps way in 2016, but if they break Biden's way this year, which it appears they are based on most polls, that changes everything.
Um...every single poll we've seen indicates Trump is winning the ED vote by a lot, certainly more than he did in 2016. This goes across pretty much all states surveyed. It also makes sense given that Ds disproportionately requested mail-in ballots compared to Rs and Ds generally have a much more favorable view of early / mail-in voting than Rs do.
The NPAs point is confusing to me. Hillary won by more actual votes than the registration gap would've indicated (-50k vs 250k actual lead), so clearly NPAs broke her way in 2016. Are you saying they are likely to break more towards Biden this year? That certainly seems possible, but what then would be your estimate for actual vote lead for Biden going into ED if Dems have a 100k registration advantage vs. a 2016 50k registration disadvantage? I cited 400-500k actual vote lead, but I suppose it could be more if Biden is winning by enough among NPAs and is winning some Rs.