Posting on a throwaway account so I don't get doxed.
When looking at other pedes' analysis of vote switching occurring in the Edison data from the NYT, I noticed something:
Almost all the switches happen at night
They also mostly happened in early morning hours on Nov 4th. See image below. I plotted the switches from 11/3 through 11/11 on a histogram:
https://i.maga.host/85O14cj.png
Now it makes sense why election coverages from that time got deleted.
Note that a "Switch" is defined as one candidate losing votes while another gained votes between updates. A "Switch" is NOT a single vote switching from one candidate to another, but a batch of votes switching, sometimes thousands.
And here is the time of day that switches happen, compiled from 11/3 through 11/11:
https://i.maga.host/qM4XizL.png
There is a very distinct pattern in when these Switches happen.
Now, let's look at November 4th only, again for all 50 states + D.C.:
https://i.maga.host/F7UqnFW.png
Notice the pattern of most Switches happening in early morning hours. This is important, because often the same pattern happens in individual states:
Individual States
Here is the November 4th data for a few of the most heated states. Notice they tend to follow the same pattern:
Wisconsin
https://i.maga.host/mFzmX0I.png
Michigan
https://i.maga.host/rvxprFo.png
Pennsylvania
https://i.maga.host/1C9DPFg.png
Georgia
https://i.maga.host/UwVTvGh.png
Arizona
https://i.maga.host/qfHi4Nc.png
"Switched" and "Lost" votes
The above analysis only looks at Switched votes. There are also times where both Biden and Trump lose votes, and they are "Lost". Here is the analysis including Switched and Lost votes:
https://i.maga.host/x5geMoS.png
https://i.maga.host/REGNeTB.png
Sources
Data is pulled from NYT. You can download data via:
https://static01.nyt.com/elections-assets/2020/data/api/2020-11-03/race-page/alabama/president.json
...where "alabama" is replaced by the state you are downloading. Use hyphens for states w spaces like "new-york".
THIS DATA IS DISPUTED Too long to put full explanation here, but basically this is the data that all the major news networks pull their live update counts from. Hence why when votes Switched in this data, the votes Switched on live TV.
Python Script: I took a python script from a fellow pede who seems to have deleted their account, and modified it to generate the histograms. Thank you, unknown pede!
I'm skeptical about this Edison data finding, but it is very peculiar. Long explanation, so I will post in the comments below.
My Conclusion
Vote Switching primarily happened in the 4am-ish hours of Nov 4th, when some of the crazy stuff was happening regarding ballot counting in certain states. Then it continues to happen after nov 4th, almost exclusively at night.
It rattles my tinfoil hat that these switches occurred in the middle of the night when few people would be watching, synchronized across states, and synchronized with the shenanigans at about 4am on Nov 4th.
However there are many other explanations than this being nefarious. I theorize it has something to do with how this NYT data is pulled from servers.
Here is my theory for votes getting lost
One server lags behind another, and the nyt data comes from both.
Here's a simple example: Think of one state, say: Texas. Let's say Texas has server A and server B with the vote counts. Let's say server B lags 1 hour behind server A.
Now, let's say new data comes in at 12:00 and immediately sever A gets updated. Immediately, NYT pulls the data from server A and their count updates. Then, 5 minutes later, NYT pulls data from server B, which is lagging behind, and those votes get removed. Thus votes subtracted from all candidates.
Here is my theory for votes switching
Picture the same example above, but another layer of complexity. Let's say server A is up-to-date on a certain region e.g.: West Texas, but is lagging 1 hour behind in East Texas. Let's say server B is vice versa. Pulling data from these servers at different times would give you a switching behavior.
HOWEVER, this still does NOT explain how exactly 19,958 votes switched from Trump to Biden on live TV in PA. (And that switch is captured in the nyt data). That is, Trump lost exactly 19,958 and Biden gained exactly 19,958. It would be mathematically nearly impossible to coincidently have the same exact number. Perhaps there is an explanation, though.
Also, why does there seem to be no Switching/Lost votes whatsover on Nov 3rd? Only after the vote count grinded to a crawl do we see this behavior.
Thoughts on Dr Shiva
I am unimpressed with the way he presented the data. Namely, he plotted on a cartesian plane with two dependent variables. That will just about guarantee a distinct pattern. But he knows more than I do, take my criticism with a grain of liberals' salt.
I think we are on to something, but it's inconclusive. Keep digging pedes, and be careful not to spread misinformation and make us look bad. Trust but Verify! 👍