I’ve charted above the data from 538 showing what their model in 2016 and 2020 projected or projects for Trump.
In 2016 his probability in the model moved a lot – from as low as 12% ti as high as 48%, and ended up on 29% (and of course he won).
At this stage in 2016 he was seen at only 13% to win and today in 2020 he is at 28%, so you could argue their model has his twice as well placed as in 2016.
However I think the better argument is Trump is more of a known quantity now, and we just are not going to see the same volatility as in 2016. I doubt their model will ever have him below 20% or over say 40%. But time will tell.
The post-convention polls will be interesting but Nate Silver said the model has been designed not to give too much credence to them as any bounce from a convention tends to disappear within a week or two.