This is based on their proprietary "Edison" data source which would ordinarily be impossible to access for people outside the press.
The CSV is available here:
s000.tinyupload.com/?file_id145669…
And the script to generate it is here:
What we are looking at will be time series analysis and you will see that it is extremely difficult to create convincing synthetic times series data. By looking at the times series logs of the ballot counting process for the entire country, we can very easily spot fraud.
One of the first things noticed while exploring the dataset is that there seems to be an obvious pattern in the ratio of new #Biden ballots to new #Trump ballots.
As we can see on this log-log plot, for many of the counting progress updates, we see an almost constant ratio of #Biden to #Trump. It's such a regular pattern that we can actually fit a linear regression model to it with near-perfect accuracy, barring some outliers.
How could this be possible? Is this a telltale sign of fraud?