I've written lengthy posts about this 2020 phenomenom: pollsters being so far apart, and with several "clusters" of polls around the edge of the predicted range (as opposed to an unimodal distribution clustered around the middle of the polling range, with very few outliers, that we would expect under normality assumptions) that someone necessarily has to be way off. The huge spread between the polls that we're seeing this year cant be explained with sampling error or minor methodological choices - someone has to be fundamentally wrong.
Quote (Santara @ 2 Nov 2020 11:59)
This morning, 538 has FL at Biden +6 with a 2% margin of error. Even if Trump picked up PA, WI, AZ, and GA, he would still need to pick off both MN and IA, or NC to cross 270.
The average shows Biden up by 2.4% in Florida, not 6. There was a single Emerson poll (rated A-) yesterday which showed Biden up by 6. The same day, ABC News/Washington post ( an A+ rated pollster) had Trump up by 2.