1. You Can Torture Data Until It Confesses To Anything You Want

Which is precisely what some people are doing to the Georgia vote counts from 2020 election. It certainly piqued my interest, as official data is readily available from the State's website. What does it say?

2. For starters, Georgia has some unique characteristics in this election that lend themselves to an interesting bit of analysis that is either not possible or not easily doable for any other state. In addition to the President, both State Senators were also on the ballot in GA.

3. And that allows each party's senate candidate to be used as a control data sent against that party's presidential candidate for each voting channel (in-person versus mail-in) to spot any glaring statistical anomalies/impossibilities in voting that might suggest fraud.

4. And that is exactly what @PandaTribune did in the enclosed thread. All his data calculations are correct as far as I can see. I can vouch for them based on my own look at the same data downloaded from Georgia's state website. https://twitter.com/PandaTribune/status/1337377252884029440?s=20

5. You can trust the data. But what about inferences @PandaTribune has drawn from the data? That is where I diverge. I looked at the same data and I didn't see the eye-popping anomaly that is implied by, for example, this bit of analysis. https://twitter.com/PandaTribune/status/1337377254851096580?s=20

6. Here was my response: https://twitter.com/bansisharma/status/1337931580279500800?s=20

7. Then someone said this: https://twitter.com/Sophisto5/status/1337938691797340162?s=20

8. Well, maybe and perhaps quite likely a few little shenanigans did take place, but nothing at the scale being implied by many. Here are a few things to keep in mind.

9. It doesn't seem all that anomalous when you consider:
a) Biden got 58K more mail-in votes than Ossoff (Trump got 7K votes less than Perdue).

b) While Biden underperformed Ossoff in election day votes in 39 counties, he didn't underperform by more than 100 votes in any county.

10. And:

c) As a matter of fact, in 19 counties Biden underperformed Ossoff in election day votes by less than 10 votes.

d) Ossoff is a much weaker candidate than Perdue, hence no surprise Biden bested him in every county in mail-in votes.

11. Most of the questionable adjudications very likely were decided in Biden's favor in large cities that are predominantly Democratic. That's, however, true of all large cities in the country, not just in Heorgia. That's why Democrats are so hot on universal mail-in voting.

12. Frankly, the more I look at the data, the less anomalous it looks. I believe fraud detection is only possible through signature verification at this point in Georgia. The rest of the data doesn't look too bad. A few minor data anomalies that do not add up to a hill of beans.

13. That is still not to say that some serious fraud that overturned the presidential win in Georgia did not occur. All I am saying is that you are not going to find it in any irrefutable statistical analysis of the vote counts data, no matter how hard you try.

14. What's worse for statisticians who are looking for preconceived inferences of fraud, the deeper one looks into the data, the more convinced an objective observer is likely to be that there is no result-changing anomaly in the date. That is exactly my own inference.

The End