I was listening to a Left Right and Center episode from about two years ago, just before the 2016 election, they were talking about what to expect. Josh Barro, giving me sweet release from the hell that is popular understanding of statistics, pointed out that while not probable, based on the 538 models at the time, it was extremely possible for Trump to win.
The post-election shock enraged me for some time. People would hear, "it is 20% likely" for something to happen and jump to loudly proclaiming it to be obviously settled. As if it is inconceivable you'd roll a die and it would land on 1. Don't view probability like that, and if you're going to make that mistake, make it for something orders of magnitude more certain, like the 10+ sigma financial models that went kaput during the 2008 crisis – at least those people had a model suggesting something close to certainty.
One of the guests (Keli Goff, I believe) said, she thought the 538 statistical models only put the chance of Trump winning as high as it was (13% at the moment they recorded, although obviously it changed a lot before that election occurred) because they would look bad if their models said it was 100% certain Clinton would win.
Let that sink in a little. I've been stewing over this on and off for hours, and I'm not even sure I'm done with it.
So, obvious things first: she got it wrong, and that looks very foolish in hindsight. Also, having 100% absolute certitude was childishly absurd overreach, even if she was correct – saying she was an arrogant fool seems appropriate. There is also a level of statistical innumeracy that is at play here that makes me a little concerned about her competence to read and understand polls, but it is very possible she has educated herself in the two years since that recording happened. But suffice it to say, to achieve 100% certitude you need infinite information. Certainty is not really a probability so much as an abstract notion, in the same way infinity isn't a number but an idea about numbers.
But the part that bothered me about it (because, if you can imagine, the above bothered me no more than almost everyone bothers me when speaking about statistics) was the tone. She thought they were rigging their models because it would humiliate them to come to a simple 100% vs 0% conclusion. She called them liars with no purpose other to avoid their own shame at being so stupid, compared to her.
This was... a staggeringly insulting comment. You'll notice my own remarks were unvarnished, but at least mine were considered. I've made an attempt to imagine others as I would imagine myself, with flaws, certainly, but flaws driven by passion and a willingness to be wrong given the alternative is to waste everyone's time with equivocation.
But the sneer of her comment, the dismissive calumniation, the confidence in the utter lack of integrity of others, it stopped me in my tracks. I suppose just before the election Josh Barro had a lot else to discuss. But that comment.. it wasn't a good look, and I'm surprised he let it go unchallenged.