I just stumbled across an audit of voting tallies at thedonald.win.  What they are doing is auditing vote tally updates.  Each update is a cumulative total of votes cast, plus cumulative percentage of votes for Biden and Trump.  What they are finding is that the total votes for Trump is being systematically decreased.  You don’t see it on cursory inspection, you just see Biden’s percent total slowly increasing and Trump’s slowly going down.  This of course would appear natural if votes were trending for Biden.  Here is an example of how the data looks:

Here it is in table form:

So a new set of votes came in, more for Biden than Trump, so Biden’s share goes up and Trump’s goes down.  The odd behaviour is apparent when actual vote totals are tabulated::

Trump’s cumulative total went down!  They are running this audit everywhere and finding similar circumstances everywhere.  This isn’t happening in one or two data sets, it is happening everywhere.  It is exactly what you would do if you wanted to flip an election – make small changes everywhere that appear to support a larger trend.  If the data set is valid (which I don’t exactly know), someone has alot of explaining to do…

Having gone through the data a bit, it appears to be legit.  The source data is simply a dump from the NY Times website and contains about 80% unrelated data.  i.e. they downloaded an archived page that contained the data voting time-series along with a bunch of other information.  Also, it appears that there are quite a few switches from Biden to Trump in the data.  But, what is heavily one-sided is vote-counts switched from 3rd-party to Biden.  Those numbers are huge.  Here are a few examples:

How is there not a single instance of more 3rd-party votes getting flipped to Trump than to Biden?