The csv file which shows votes per state over time, last few days.
Csv
http://s000.tinyupload.com/?file_id=14566999163598825215
Python script
Issue:
Csv file has the vote share columns which is used to calc each candidates votes. The vote share columns only have a precision of 3 decimal places. This leads to vote shaving, needs to be at least 7 decimal places of precision.
Take 2,690,400
Votes share .493 for Trump (1,326,367 votes) Vote share .502 for biden (1,350,581 votes)
Now change Trump vote share to .492 and leave biden unchanged. Happens all over the place and by bigger margins.
Now Trump has 1,323,676 votes loosing 2,690 votes.
How can any data enginneer use only 3 digit precision for calcs like this unless you want rounding errors to hide vote shaving and use it as cover as a oversight or "glitch".
This manipulation and large voter dumps only showing up in key battkeground states and a few others.
Does anyone have any time series data by precinct?
This data being used by all media outlets is bunk.
Are you getting the impression that you're the smartest one in the room because the room has gotten dumber.
When I looked at the actual Edison data I had this same sudden uncomfortable realization. There are people on the site saying the same but mentioning "granularity", maybe they know but are using that word in an attempt to keep the obfuscation going and hide their embarrassment.
My take is Edison releases data only precise enough to drive news screen bar graph widgets but the news people go on to calculate (x*0.xxx decimal) trash vote numbers for extra eye candy. The news people do the trash math so Edison sees them put vote totals that aren't accurate on the screen and says "well what the fuck" while smiling. Fuck Edison. Ask me even 20 years ago whether news networks would some day put trash-numbers on the screen and not care and I would have said NO WAY.
Modern technology: with massively more speed and capability than ever before, deliver an inferior product at greater cost.