BTW Fivethirtyeight posted the final model talk episode of the podcast. Having listened, I feel that I now understand Silver's model, what it really is. I echo what he says about the importance of DISCUSSING polls but at the same time the 538 model is his own product and is not a rigorous or scientifically sound thing.
( Polling science is real and mathematically proven, but not HIS poll of polls. )
When he says 'Biden to win with 90% chance' it is indeed a very well crafted and educated GUESS. Unlike a real poll.