5 ms·
> When polls predicted a 99% chance of Hillary Clinton winning, you and I and most of our peers know that the polls weren't wrong. A bit of a tangent, but poll
by justinpombrio 5y ago
> When polls predicted a 99% chance of Hillary Clinton winning, you and I and most of our peers know that the polls weren't wrong.
A bit of a tangent, but polls don't make predictions, people make predictions based on the polling data. Some people predicted Hillary with 99% chance, and 538 predicted Hillary with 66% chance. When dealing with probabilities it's not as simple as "wrong" or "not", but if people were betting then those that bet 99% would have lost a lot of money, and 538 only a little. As a result I trust 538 a lot more now; that's the nice thing about people making specific predictions, you can over time learn who to trust.
Anyhow this tangent ultimately agrees with your point, as you and I know that if a talking head makes a "99%" prediction, it's got at best an 80% chance of actually happening.
- jonnycomputer 5y agoAnd if 538 were consistently wrong, say wrong 80% of the time, then ~538's prediction gets you right 80% of the time. Win!
- thetallstick 5y agoDoesn't this only works if an election has two possible outcomes? Revotes and voids are possible as well in most cases. In 2022 these things somehow feel like they're much more likely than before...
- jonnycomputer 5y agoSure. Though in many cases these types of events are pretty low probability events.
- bcrosby95 5y agoAnd lots of people still said 538 got it wrong. No, it didn't. Or about how Omicron defies statistics. No, it doesn't. People just don't fucking understand statistics.
- chmod600 5y agoHow can you objectively say that 538 was right or not? They use numbers to come up with probabilities, but there are a lot of baked-in assumptions about things like turnout, various sampling errors or biases, etc. They have reasons for their assumptions, such as using information from previous elections. But the current election may violate those assumptions, but it's hard to say how much and with what probability. I trust 538 for the most part. But it's just models and those models have assumptions and it's quite possible the assumptions were just wrong in 2016. We only have one data point, so we'll never really know.
- thrashh 5y agoYou can’t over one poll. It’s really consistency that you’re looking for over many, many polls. The worst part of figuring out the trustworthiness of a source is putting in the time and effort to comb over all of its output and comparing it to reality, an effort that can take a long time. Which lends me to ask: how can you say that you trust 538 if you also say there is only one data point? Because there shouldn’t be: 538 made big news during the election of 2008. 538 has been around since 2008 so you should have a lot of polling predictions to comb over. I actually don’t have an opinion of 538 as I haven’t put in that work.
- belter 5y agoYou can't blame People. Even some Statisticians, with Phd's and written books, don't understand Statistics and confuse basic concepts of their domain: "The p value and the base rate fallacy" https://www.statisticsdonewrong.com/p-value.html https://www.statisticsdonewrong.com/p-value.html "Misinterpretations of Significance: A Problem Students Share with Their Teachers?" "https://www.researchgate.net/profile/Heiko-Haller/publication/27262211_Misinterpretations_of_Significance_A_Problem_Students_Share_with_Their_Teachers/links/0deec53198736a7dae000000/Misinterpretations-of-Significance-A-Problem-Students-Share-with-Their-Teachers.pdf https://www.researchgate.net/profile/Heiko-Haller/publicatio..."
- TrispusAttucks 5y agoPolls also only count the people you poll. If your poll pool has an inherent bias the poll results will reflect that bias.