Another somewhat unexpected factor for Senate control: MICHIGAN.
John James is currently outperforming Trump by ~1.2 points on the margin. But there's a small discrepancy in number of votes counted for President/Senate at the moment.
If we have to wait 20 minutes per scan and Biden wins each 300 vote scan ~90% - 10%, he would net ~240 votes per scan, and we'd need 6 scans for him to cut Trump's 1267 vote lead down to zero.
In the first batch of mail votes from Clayton County GA, Biden appears to be getting the type of margin he needs to close the statewide gap once future batches report.
Why do I believe Biden is favored in Pennsylvania even though he's down 8 points?
Most of the counties that have reported almost ALL votes (including mail ballots) show a decent shift to the left versus Clinton's margins in those counties in 2016.
There are essentially two flavors of forecasting these states.
#1 - Looking Backwards - Look at counties with >98% expected vote reporting and compare their semi-final margins to their 2016 Trump vs. Clinton margins.
With this method, Pennsylvania looks good for Biden.
#2 - Looking Forwards - Look at counties with remaining votes. Estimate the margins of the remaining vote in each county and add up all the numbers. See who's ahead.
I think method #1 has made more sense for Pennsylvania so far since there is so much mail vote EVERYWHERE.
If counties with almost all their vote in shifted ~3 points left vs. 2018, we might expect future counties to do so too.
We can then apply margin estimates to turnout estimates and add everything up.
My turnout baseline is 76% statewide (10.976M), but that might be adjusted.
I chose the 2018 gubernatorial race rather than the 2016 presidential race as the baseline comparison for Florida counties because it does seem from the polling that Trump may outperform his 2016 margins with Latinos. The midterm elections showed a similar demographic pattern.