It seems they are using economic data back since 1880(!) but not Presidential Approval. I think that is strange.
Yeah, this seems like the kind of arbitrary decision that Silver would challenge other forecasters about (quite reasonably) were they to make it. Nate Cohn mentioned it in his Twitter thread, although he didn't exactly @ Silver over it. It's not quite as head-scratching as the NYT Headline factor, though. That's a bit amatuerish, really.
Not sure what to say that hasn't been said already. The model seems slightly conservative, the nowcast less so, except we don't get to see that one for some reason. Having it available might make the model seem less indecisive. Don't love that there's no national map. Unsure if Nate's fursona adds much insight - idk if people going to 538 need to be told that the underdog could win, but it is what it is.
And this is entirely subjective, but comparing
the 2016 model to
this year's one I think 2016's page is just a bit better. There's some cool stuff which didn't make it to this year's page, like the error bar plot, the hexmap, and the scenarios down at the bottom.
It's also more obvious where to go for individual states - you've got the national map as the first thing you see, but you've also got the sidebar, with the competitive states right at the top. The 2020 page has this list, but it's in a visually indistinct dropdown menu that's not anywhere near the content -which only takes up the middle 1/4 of the page, as if it were an essay with a page limit. Bit silly really. The 2016 page is also rather more pleasant to look at- seriously, what's with the two separate shades of grey and weird background swoosh on the 2020 model. That thing ugly. The 2018 page and the primary model are practically identical in design to the 2016 page, so not sure what inspired the switch. Maybe it works better on mobile?