I’m finally reading the THE ODDS ARE ONE QUADRILLION TO ONE “expert” testimony.

https://electioncases.osu.edu/wp-content/uploads/2020/12/TX-v-PA-Appendix-first-half.pdf

I’m going to summarize as best as I can! It has been a long time since I’ve done statistics (since grad school probably? 2000?), but I can basically follow along.

So first our “expert” asks if Clinton and Biden’s performance are statistically similar, and finds that there is a less than one-in-quadrillion chance that they are not statistically similar.

Okay, now, here’s the thing: I am *pretty* sure you can’t just assume votes are normally distributed etc etc, which all of this does, so this is wonky math from the outset.

But.

I would just like to point out that what this proves—if the math is right, which it isn’t—is that there is a one in quadrillion chance that the vote in Georgia in 2016 for Hillary Clinton was statistically similar to the vote in Georgia for Joe Biden.

I... I.... I don’t think you need to be a statistician to determine that?

They are obviously different? For many reasons?

Okay so anyway, having proven that we are NOT stuck in a time loop, and that Joe and Hillary are two different people, this “expert” goes on to mention ONE OBVIOUS way that this is all different: more ballots were cast!

And he says, OKAY, let’s get rid of that influence!

But it’s cool, he still concludes that we are NOT stuck in a time loop and that Joe and Hillary were different people.

Okay, LEMME ZOOM IN ON THIS LAST PART BECAUSE IT IS SO INCREDIBLE.

Recall that this man used shitty math to “prove” that there is a less than one-quadrillion-in-one chance that nothing changed between 2016 and the present day.

You’d think he would conclude that something has changed between 2016 and the present day.

But no. This absolute genius, the lead witness in this utterly ridiculous case, concludes that *he* cannot figure out what could have changed, and so this means that actually there is a less than one-quadrillion-in-one chance that the results were valid.

I’m just going to stop and point out something REALLY obvious.

If you applied this shit-stupid analysis to EVERY election in human history, you would conclude that they were all invalid.

Literally every single one.

You will *always* find that past elections are not statistically similar to present ones, and you will find this in just about every election, because—hello—THINGS CHANGE.

What you then have to do, if you’re doing any kind of scientific analysis, is at least *try* to disprove the impact of those changes on the electorate.

This one-in-a-quadrillion number comes out because he is baking into his (terrible mathematical) model the assumption that nothing should have changed.

He starts with the assumption that Trump should have won, finds that the actual result is statistically different than Trump winning (lol you do NOT need statistics for this), and concludes that it is invalid because it statistically does not match his assumption.

All of the statistics is a smoke screen for that.

This is a *very* fancy way of saying, “I expected Trump to win, he did not, I could not possibly be wrong, and therefore fraud happened.”

Anyway, this is embarrassing and I’m embarrassed for Ken Paxton and every last state attorney general who signed on to this horrific piece of garbage, OH MY LORD, it’s so bad.

*contemplates going on*

*contemplates hitting myself in the head with rusty nails*

*latter seems preferable*

Ultimately, here’s the thing. If you are collecting data (and voting is a form of data collection), if you do not know what your data *should* look like in some fashion, you have to be very careful about discarding data without independent evaluations.

You can’t come up with a statistical test comparing one election to another and just discard votes until this election fits last election!

This is a method that gives you a dictatorship, not a democracy.

Anyway, I personally find it statistically incredible (and an equal protection violation) that Ken Paxton has not read all my books, and I think Colorado should sue the state of Texas to enforce this result.

*starts reading on*

god this dude REALLY likes the word quadrillion.

Am I doing this?

This “expert” also concludes that the early votes reported were not drawn from the same sample as the later votes.

...yes, this is why nobody called Georgia after the first 1% of votes had been reported.

In this person’s defense, he DOES mention that maybe there’s a reason why the later reporting ballots were different?

But then he says: “I am not aware of any actual data supporting that either of these events occurred.”

Did he....look...at the prior paragraph?

First, I want to scream about the use of “anecdote” and “data” here. Statisticians talk about anecdote versus data, because someone saying “I took hydroxychloroquine and I did not get covid, so it proves that HCQ stops transmission” has not proven anything.

But saying, “Nobody in the country of MadeUpLand took Hydroxychloroquine“ is not an anecdote. It’s a discussion of the statistical character of an entire sample.

”They counted all the absentee ballots last“ is not anecdote in the statistical sense.

Second.... if you want data that supports the proposition that the later-counted ballots were *different* in character....

Would you accept evidence that they had statistically different results than earlier-counted ballots?

It’s like, yes, there is data that shows they were different. You just...pointed to it.

You can’t say, “they’re different, but I’m not going to accept reasons why they’re different, they should be the same.”

That doesn’t prove anything.

....I am on page 6 of 75. And this is...the first half, because the exhibits are split.

*runs screaming*