I was taking a look at the polling averages and if the averages were reflective of the results, this is what the map would look like, a solid Biden victory. Anything that has Biden underpolling at all is just an overwhelming defeat for Trump.
If we have the polls underpolling Trump across the board by about a quarter of the value of the average margin of error, this is what the map looks like. Still a Biden victory, by slimmer margins, although AZ would almost be a toss-up in this scenario, going Democrat by only the slightest of margins. PA is the only other one kinda within reach, assuming minimal or no shenanigans (unlikely) in this scenario. He'd need PA to force a draw and an election by Congress, and both to win outright.
If we have the polls underpolling Trump across the board by about half of the value of the average margin of error, this is what the map looks like. Pennsylvania in this scenario becomes the key, and I think the recent rulings and shenanigans might be enough in this scenario to make PA go Democrat, which would give Biden the win. NV is almost a toss-up here, and MN and WI are both within reach absent shenanigans in this scenario, so doing a little better in a couple of those would negate any need to get PA.
If the polls were to be underpolling Trump by about 3/4 the value of the average margin of error across the board, the map looks like this. Shenanigans might still be enough to push MN, PA, and NV into Democratic victories, which would then give Biden the election, but absent such an impact, Trump wins.
Now, if the polls were to be underpolling Trump by the maximum value of the average margin of error across the board, the map then looks like this. Clear Trump victory. If the polls were to be underpolling Trump beyond the margin of error, then Trump wins. He'd likely get MI, then, and CO, VA, and NH would possibly be in play at that point.
If he were to be underpolling by 1 1/4 times the value of the average margin of error across the board, the map would look like this. I think this is pushing the boundary of any plausible victory, although VA, ME, and NH are all Democrat in this scenario by under 3 points. NM seem to be a really tough nut to crack at this point.
I do think he's underpolling to some degree in many if not most States (perhaps even all of them). What is important is the degree to which this is the case, and specifically to what degree in which particular States. It wouldn't be uniform like in the hypotheticals above. That'll determine whether he wins or loses, IMO. He needs to be underpolling by more than 1/4 of the margin of error, and perhaps by half or more of that value, in multiple States to win. I did think it was interesting to see what the results would be to look incrementally at the effects of underpolling on the outcome.