If you are interested in doing a more detailed analysis of this, I can supply you with some boiler plate code. 538 has pretty detailed polling data that’s free to download. I’m going to be getting into it later in June for my monthly polling update.
I was considering calculating Trumps polling error differential from 2016 and seeing how it changed to 2020. I did a map of his differential polling error for 2020 for my update two weeks ago.
I’m also considering of taking the differential polling result for just swing states, applying it to current polling, and mapping out a series of ‘pathways’ to 270 for each candidate, highlighting where the pressure points are.
I think a trying to model the impact of a single news story could be pretty interesting.