Thread by @thomasafine

There's some scientific arguments for lab origins of COVID-19 that are frankly well outside of my expertise and I can't comment on them.

But there IS a statistical component to their arguments that some are making, and that I CAN comment on.

Statistics is one of the trickiest fields of math, because it has strong interplay between the results you get, and how you state the problem. It's easy to create what seems to be some sort of statistical paradox or weirdness, but it's usually in problem misstatement.

And so there's an argument that says (more or less):

"the mutations to this virus are exactly and precisely what a lab would do to make this non-human virus very dangerous for humans."

And this makes it seem like an extremely unlikely mutation. And it IS an extremely unlikely mutation in the single mutation context. "What are the odds of this mutation happening"? Could easily be a billion or a trillion to one.

Well that sounds like damning evidence, doesn't it?

Sure until you consider that millions of natural viral mutations occur every day, day after day, year after year.

But if the odds are a trillion to one of this mutation occurring naturally, and I hand you a bucket full of ten trillion mutated viruses, suddenly the odds of this mutation actually occurring are pretty high (about 63% actually).

And in this case we have an absolute ridiculous selection bias. We have selected a virus which we know has a mutation dangerous to humans, and then we try to calculate the odds of it having a mutation dangerous to humans?

It's 100%. The odds are 100%.

This is of course way oversimplifying the complicated scientific arguments in favor of lab origins. HOWEVER, in my experience even the smartest scientists are capable of completely botching statistical arguments.

I want to add that I'm not saying I know this did not come from a lab.

I'm just saying, be careful with that statistical argument, Eugene.