The point here is right wing pollsters have form for flooding the public with dogshit consistently lean R polls. Trafalgar released a down ballot race poll that were off by 33(!) points, and the less said about the likes of Rasmussen or Echelon Insights where their chief pollster are actively shilling on Twitter for Trump the better.
Predictive polling models are like the human body, if you feed crap in, crap comes out. Silver missed it 16, 20, and 22 just like the rest of the field, and now his model includes all the polls I just listed. He can weight them all he wants, but if they are all that is there in battleground polls past recent couple of weeks there will be a movement towards Trump in his model (which, interestingly btw, never off by more than 2% from Polymarket, which he now works for). Meanwhile, national polling have shown a pretty uniform 3 point shift towards Harris in the pre-post Convention period, and recent polls for Nebraska or Minnesota shows similar margin to 2020, which is consistent with evidence we have seen elsewhere of a D+3/4 top ticket.
It’s gonna be close, but we always have to keep in mind that 1) polling errors can happen in both direction and 2) right wing pollsters are actively doing a psy op by skewing the averages with their biased, dogshit polls designed to galvanise their base.