About your overfitting comment. You've been a little selective in choosing Macomb county to suit your argument. If you watch Dr Shiva's video, he's looking at Oakland County Early voting when he introduces the horizontal line. In that data there is a cluster of points that show clearly a straight line is not the best fit - probably a curve could also fit.

He then applies that to other precincts where there is no similar evidence but the data points are sparse in that region so it's difficult to rule out there could be a similar curve. Not proof, but not debunked either.