Ask HN: Did Nate Silver End Up Listening to Nassim Taleb's Critique? | Hacker News Reader

Ask HN: Did Nate Silver End Up Listening to Nassim Taleb's Critique?

13 pointsTomminn8 comments

I noticed listening to 538 today that Nate Silver produced a model that gives Biden a 71% chance of victory, which is anomalously low chance given current polling data. The explanation is given that the election is in the future, and therefore there is future uncertainty. This can be interpreted as there is future uncertainty in the polling data, which I believe was precisely the pertinent fact Silver's 2016 model ignored, according to Taleb's critique.

I'd be interested to hear from people, especially people who took Nate Silver's side in the debate, whether they still believe he hasn't absorbed Taleb's critique.

View on HN

> whether they still believe he hasn't absorbed Taleb's critique.

No. Taleb's an idiot, he just does psychobabble. Silver knows that.

I can't find the data but I think 3 months ago Trump would have won going off polls. So clearly in 3 months he could win again.

We are in a period of huge variability.

For instance Biden gets the coronavirus is a 10% chance he'll die. 10% chance he can't run in the election. Biden can't keep away from people, it'll could drain the polling numbers, he'll have to mix. CNN hinted the VP is very important this election.

Silver also knows he stuffed up last time, from everyone commenting. 30% is be a good face saving estimate at this point in time, he easily can up or lower it as the election comes closer.

Nothing to do with Taleb.

[edit] > "The explanation is given that the election is in the future, and therefore there is future uncertainty.

This is well known and always accounted for, along with known polling and mail-in biases etc. What's less well know is how to do this accurately, knowing Taleb I'd guess he's pretends to know this with a vague statement. I'm not sure what he actually said?

> 71% chance of victory, which is anomalously low chance given current polling data.

Your question is based on 71% being low, which I don't think it's true. It's way way higher than I would expect or predict and certainly higher than what the polls indicate. Have a look at 538's methodology for creating it and also look at Nate Silver's article on how you shouldn't count Trump out yet.

As far as Taleb goes - who cares, the more people ignore him, the better.

> Your question is based on 71% being low, which I don't think it's true.

Assuming I got my maths correct the betting market has him at ~ 65%

https://www.oddschecker.com/politics/us-politics/us-presiden...

538 are claiming that their model has a lower percentage for Biden than others (random example - Economist has Biden at 87% currently) because they're considering more than just the polls. The implication seems to be that their prediction is actually lower than their analysis of the polls would indicate.

Frankly I am surprised pollsters are still taken seriously as a profession. 2016 should have shown them to be nothing more than fortune tellers, skilled in the art of telling people what they wish would happen. Yet, here we are...

The problem is that people do not understand their findings.

Election forecasting isn’t useful for those who don’t understand enough about statistics and forecasting to have the correct context to know how to read and use the information. That’s not a criticism of you, it’s a problem in the field. I work in data science/forecasting, and explaining forecasts is so hard that it’s often useless for those who don’t understand how to read them.

Puncher's chance. Account for all the improbable things; Biden literally shitting his pants on the debate stage with Trump, Biden pissing his pants, Biden showing up wearing no pants, etc. Any individual such event is not very likely, but if you add them up you probably get at least 29% of one of them occurring.

Reply on news.ycombinator.com