Wait so you're going with, 'repubs may have committed fraud, but dems did more!' Oh dear.
The fact is, we're looking at a map of Philadelphia county, which is a Dem stronghold. Even though in 2016 PA went to Trump, Philly was 82%+ Dem. In 2012, it was 85%+ Dem. In 2008, it was 83%+ Dem. 2004, 80%+ Dem. 2000, 80%+. My point is that regardless of which way the presidency goes or who wins the state, Philadelphia always leans heavily Dem. It's hardly a surprise to find the map "riddled with precincts showing Democrat votes near the census" estimates, as you said.
I'm frankly shocked that we could even find 3 precincts where Trump won and had numbers exceeding the census estimates in such a Dem stronghold, considering there were only ~128k votes for him in total. If we instead looked at a county that leaned heavily the other way, you'd see the same pattern in reverse, with far more precincts exceeding census estimates for Trump, simply because they'll be more votes for him in total. Just for a quick sanity check, I looked at Mobile Alabama, where Trump won a modest 55.3% of the vote and 100,605 votes, Biden won 78,754
What's going on here is confirmation bias. You're only looking at (Dem) areas to confirm your foregone conclusion and not looking at (Repub) areas that could disprove it. You cannot base your view that this pattern proves Dems committed voter fraud while dismissing counterpoints of the exact same pattern as Repubs only committing negligible amounts or as simply aberrations based on looking only at an area that consistently overwhelmingly votes Dem. Quite clearly, when you have far more votes in total, it's more likely that, when distributed amongst the precincts, it'll exceed census population estimates. As the guy who created the map admitted, trying to extrapolate census data back to wards and precincts is guesswork estimates - it's not factual information, and not a good indicator of fraud.
