2020 Electoral College simulations

I have been compiling 2020 general election polls at the state level corresponding to one of two specific matchups: Biden vs. Trump and Sanders vs. Trump. Here I describe what happens if you generate Electoral College scenarios on the website 270towin.com in the following way:

1. Start with the actual 2016 results.
2. Average the results of 2020 polls for a given state between Trump and whichever of the two Democratic candidates you are considering.
3. For all states for which the average generated from step 2 is less than 5 points either way, make the state blank (the gray “tossup” color). For every other state, color it corresponding to whether Trump or the Dem has a lead in the aforementioned 2020 average.
4. From here, try to find the best-case scenario for Trump by pretending that he wins every state that is currently blank — so turn all these states red.
5. Do the same as step 4 but the opposite — i.e. turn all states that were blank just before you started step 4 blue instead of red. This yields the best-case scenario for whichever of the two Dems you are considering.

Here are the results:

For Biden, the best-case scenario for Trump is that he loses the Electoral College to Biden, 290–248:

The best-case scenario for Biden against Trump is a landslide Biden victory in the Electoral College, 417–121:

But for Sanders, unfortunately, the best-case scenario for Trump is that he actually wins the Electoral College by 290–248, the same margin by which he would lose it to Biden in Trump’s best-case scenario mentioned above:

The best-case scenario for Sanders is that he wins the Electoral College in a landslide, 406–132, though it should be noted that this is a bit smaller than Biden’s best-case scenario landslide mentioned above. This is because the two best-case scenarios are identical except for Arizona, which Sanders loses but Biden wins:

In conclusion, Biden appears to have a moderately better chance of beating Trump in the Electoral College than does Sanders, based on polling evidence currently available.