From an analysis of GA at libertyauthors.com:
“This is what you are looking at: First column is Trump’s initial lead margin (after 9 rows), second column is his final loss margin and the third column is the name of the precinct.
I took every precinct with 9 or more rows in the CSV because fewer rows than that is probably not enough time-series data to give meaningful information in respect to a reversal. That comes out to 2,653 precincts. Of those precincts, 1,721 were showing for Trump by the 9th row in the CSV, but then went to Biden by the final row. Just 3 precincts were showing for Biden in the 9th row but went to Trump on the final row. I wonder if this is what Sidney Powell is referring to by “mathematical impossibility.” You will notice that the final win margins are remarkably uniform. I haven’t run a distribution on them, but they’re pretty much all clumped up on one value.
OK, now for possible ways to explain this away. I have clumped together all data rows as they are sequenced in the CSV file, without respect to vote-type (absentee, early, election-day, provisional). I think the D-narrative is that “early voters broke for Trump but late voters broke for Biden”. But the data is too “smooth” for this. I don’t know exactly how to describe it, I’m trying to think about how to graph it to show it. The time-series do not have the amount of variability you would expect in a “random” dataset. Very, very weird. Digging...”
More at: http://libertyauthors.com/viewtopic.php?f=109&t=3606
More from the GA analysis thread:
“Smoking gun...
What you’re looking at: Percent win (positive) or loss (negative) for Trump for every precinct. -1000 means 100.0% of votes went to Biden, 1000 means 100.0% of votes went to Trump. The smoking gun is the spike on the far left. A large hump would indicate “very popular but polarizing candidate” that strongly inspired his base to turn out to the polls. But that spike is anomalous, to say the least.”
More at: http://libertyauthors.com/viewtopic.php?p=30640#p30640