We have to get over this countermathematical notion that if you say something has a 20 percent chance of happening and it happens, you were wrong.

If you roll a fair die, you have a 16.6 percent chance of rolling a two. If you do roll a two, guess what? Model was still right! https://t.co/yqxo5M3trr

— Linda Holmes Thinks You’re Doing Great (@lindaholmes) October 1, 2020

Unfortunately, elections are not math and occur only once. Therefore, models aren’t much use. At least not nearly as useful as people like Holmes like to pretend they are.

This is where quant/math people utterly fail to understand the real world. I can’t believe we’re still re-litigating the “but the polls were right!” bullshit from 2016, when the “experts” got their ass handed to them. The problem is, they are talking about the national polls and it’s the state polls that matter due to the electoral college.

Aside from all that, there’s too much philosophy and epistemology here to write about. But…pretending that no matter what your model says that you’re right has nothing to do with what anyone who is not a math geek wants or needs. In my opinion, election polls and models should never be made public. It’s harmful and also fairly useless.