538 prediction is back up to 88%...it's been there before, then fallen back to 87%.
Senate 75%-80%, depending on the model
Popular vote spread has closed a tiny amount - down to 10.2 from 10.8.
Nate said something about comparing 2016's polling failures to 2020.
He basically said 2016 is a sample of 1. 538 derives its odds from not just one election, but from dozens of them, over decades. Looking at larger samples, and you don't see many polling failures like we had in 2016.
On the other hand, we never had a politician like trump, we never had cell phones, we never had half the R party thinking the Democratic Party is a front for a child trafficking ring, etc. Maybe 2016 is more relevant than 2008, 2004, etc.