...I don't put a lot of weight on polls, and I'm not chiming in here to say they are right or wrong, but it does seem like 538's overall forecast of 89% (a near certainty) for Biden is too high. I'm probably wrong on the statistics underlying this number because they are, after all, a professional statistical modeling firm, but it just seems off to me.

Biden is going to have to win a number of states that are each reasonably close (hence called "battlegrounds" or "swing states") to win the election (e.g., some combo of FL, MN, WI, MI, PA, NC, AZ, NV, GA, OH, etc.).

Let's say that there's a 90% chance that he's going to win each of FL, OH and PA, just to take three at random, and he needs to win all three (for the sake of argument). The statistical probability of a clean sweep would be 90% x 90% x 90%, which would equal 73%. And, in this case, my guess is that there is a much lower than 90% chance (a near certainty) that Biden will win any of these states with maybe a couple of exceptions. Also, the polling itself is fraught with two errors: 1) the confidence intervals inherent in any statistical analysis and 2) the inherent vagaries of trying to poll a topic like a highly emotional and contested election about which people will lie and/or refuse to answer polling questions.

So, it just seems off to me. I guess if you go through enough permutations of potential state outcomes and assume that Biden can win with a variety of different collections of states (which is true), then you could come to a statistical probability of 89% for Biden. I'm assuming that's what happening here. In reality, though, I don't think you can predict elections like that -- in other words, it seems like a meaningless statistic to say there's an 89% chance he'll win.

It would be interesting to see what 538 would predict if you assumed that Trump is going to win Florida and Ohio, for example. Maybe they do that -- I've never been to their site.

Anyway, thanks for any comments from folks who are more familiar with 538's methodology than I am (which is not at all).