With the numbers it has to be more than "shy Trump voters" though....It has to be "shy Trump voters lying and saying they are voting for Biden" and if the stats are to be believed, it would have to be double the number that supposedly did this last time. And, of course, the polls would have to be using the exact same assumptions they did last time.
One of the main differences in 2020 vs. 2016 is Biden is polling 50%+ in: CA, CO, CT, DC, DE, HI, IL, MA, MD, ME, ME1, MI, MN, NE2, NH, NJ, NM, NY, OR, PA, RI, VA, VT, WA, and WI. He's 49.8 in NV (but has been +/- 0.2 from 50% across the last week). All those added together gets JB to 279 electoral votes. Trump could win every other state and it wouldn't matter.
In 2016, Hillary was leading but not above 50% in polls in NC, FL, PA, MI, and WI (all of which she ended up losing). She won every state where she polled 50%+. Trump picked up a lot of undecideds and independents down the stretch last time, but there are a lot fewer of those this time around.
Put another way, compared to where he finished in 2016, Trump is currently running: -7 in MI and ME2; -6 in IA, -5 in NC, WI, AZ, GA, and NE2; -4 in PA and OH; and -2 in FL. He won FL, PA, MI, WI, and NE2 without getting 50% of the votes (and doing so again in all those states seems improbable).
For Trump to win: the polls in battleground states have to be more wrong than last time, he will have to flip Democratic votes to GOP votes, they will have to have registered a ton more Republicans, or they have to suppress votes from Democrats (or run interference through the courts). Of course, there could be a combination of all of those, but IMO the chances are decreasing with each passing day . . . and that probably has to happen in more than one state.
I thought there was a chance of a drop off for Biden in the polls after the last debate or the laptop news, but neither of those seemed to have made much of a difference. So the question becomes what will the GOP do to try to influence the outcome.