One of the things I found interesting in the analysis by Dr. Frank was the entertaining of the question "Did you find any counties or locales that did NOT fit your hypothesis?"
That is a question less asked than it should be, and I sometimes fall prey to not looking at that more closely. We all do. We think of a reason why something is, develop a hypothesis, then find examples that buttress the hypothesis.
The last step, and one that is often ignored, is seeking out examples that DO NOT fit.
In his case, as an apparently conscientious scientist, he DID ask that question, and there was one county he found that did not fit exactly...Bent County in Colorado:
It was humorous, yet revealing: Notice the curve is volume of residents by census (on the Y axis) by age (on the X axis) and the blue curve absolutely does not match like it does everywhere else...but look at the age range there: 18 to 50 year olds, then it matches up again.
(FYI, The black line is registrations, the blue is predicted ballots (by the algorithm) and that red is actual ballots cast. The oddity is that the blue line is population by census...the 2020 census, and part of the descrepancies he is finding is because he thinks the algorithms were based on the 2010 census )
Odd, but...hmm. What would that population range out of whack between 18-50 signify? Not military, too weighted with older ages. It is prison population. In a county of 6500 people of whom 4700 are over the age of 17 , 1300 of them are prisoners in a correctional institution. That's about 28%!
LOL, they tried to plan for everything, but...weaving lies can be complicated when it goes a bit deeper...I just found that interesting.