The technical analysis of polling has taken a giant leap forward since the ancient days of 2012. Back then, Nate Silver was the supreme potentate of election forecasting. We are now officially living in the Wild West of election analysis, with Nate Cohn, G. Elliott Morris, Rachel Bitecofer, and a slew of others all producing competing models. But the Doc Holliday, Johnny Ringo, and Wyatt Earp are definitely Silver, Morris, and Cohn (respectively, if you want). Silver still tends to be the North Star, given that his analysis gave Trump a roughly 30% chance of winning in 2016.
If you want to get into the inside baseball/nerdy technical analysis of polling, Silver has a detailed explainer of his methods and why his model isn't as bullish on Biden as others. To summarize it quickly: Silver's model takes into account the fact that there are still three months left until the election; that 2020 is an insane year; and that COVID is likely to give rise to a wave of uncertainty.
For those reasons, Silver gives Biden only a 71% chance of winning, as opposed to other models that put Biden in the 90% range. So, he has Biden roughly in the same place as Hillary in 2016, and we all know how that turned out. But, Silver himself says that if you tell his model that the election is tomorrow, Biden's chances jump to 92%.
COVID has sort of made the 2020 election a bummer, because there was a neat theory that we were going to be able to test out before it hit: The power of negative partisanship. This is largely Bitecofer's hobbyhorse, who says that election results are not primarily determined by swing voters, but by how much the opposition Party hates the opposing candidate (in this case, Trump). Bitecofer's analysis worked really well in 2018, when she projected Democrats to win 46 seats, when other models were projecting only around 23 (Dems won 41 seats).
It's really hard to test that model now, given that Trump has gotten incredibly low marks on the Administration's response to COVID, which will probably be the central issue in the election. But she gets into quarrels with Cohn on Twitter, which are fun to popcorn-read.
But election forecasting is now all the rage. The Economist has a model; Silver has his model; Cohn at the New York Times has a model; Bitecofer has her model; and there are a few other ones that have long been staples (Larry Sabato's Crystal Ball), but are less reliant on polling averages and analysis. Harry Enten used to work with Silver over at FiveThirtyEight and now produces his own analysis for CNN.
So, that’s why it’s kind of like the Wild West. There are dueling analyses and even scrap fights on Twitter. That’s part of what makes it a neat time for the layperson: You can see what the experts are saying to each other in real time, and how they are criticizing each model.
One thing that shouldn’t get lost though: Every model shows that Biden has a strong chance of winning the Presidency and that Trump will have to make some fundamental changes if he wants to close the gap.