Voting machines that saw more voters had a larger skew. Up to about 300 votes per machine, the results look pretty random and natural. From there on, the pattern changes drastically and shows unexpected clustering. It's unexpected because larger samples should slowly converge but still be normally distributed (law of large numbers). Instead of matching a bell curve, the distribution shows a "russian tail" (search for it on the page), a sign of vote manipulation.
All this is only for the early votes. If you compare the scatter plots and distributions of early votes vs. election day, they look completely different.
Voting machines that saw more voters had a larger skew. Up to about 300 votes per machine, the results look pretty random and natural. From there on, the pattern changes drastically and shows unexpected clustering. It's unexpected because larger samples should slowly converge but still be normally distributed (law of large numbers). Instead of matching a bell curve, the distribution shows a "russian tail" (search for it on the page), a sign of vote manipulation.
All this is only for the early votes. If you compare the scatter plots and distributions of early votes vs. election day, they look completely different.