Can the Weatherman Be Wrong?
People are notoriously bad at internalizing probability. In 2016, when 538 said that Donald Trump had a ~30% chance of winning, the mainstream narrative was that 538 “got it wrong”. This fundamental misunderstanding of probability runs rampant everywhere, and I think it would be helpful if everyone could understand the following.
1. If your weatherman says there is a 90% chance of snow, and it does not snow, they are not wrong by default — stating a probability of something is not the same as making yes/no prediction about something.
2. If your weatherman made mistakes while calculating that 90% proability, and thus there was not actually something like a 90% chance of snow, then they are wrong. They are not wrong because it did not snow, but their statement “there is a 90% chance of snow” is wrong. Key difference.
3. If your weatherman made a yes/no prediction statement, like it is going to snow tomorrow, and then it does not snow — they are wrong. In this case the thing they are wrong about is whether or not it snowed.
Separately, a ‘forecast’ is generally just a statement of probabilities. When you say you think your favorite football team will win, you are not making a forecast — you are predicting an outcome. When some ESPN analytics software says there is an 88% chance your team will win, it is making a forecast, and if your team loses the ESPN tool was not wrong.
Send to anyone you think doesn’t get this.
1. If your weatherman says there is a 90% chance of snow, and it does not snow, they are not wrong by default — stating a probability of something is not the same as making yes/no prediction about something.
2. If your weatherman made mistakes while calculating that 90% proability, and thus there was not actually something like a 90% chance of snow, then they are wrong. They are not wrong because it did not snow, but their statement “there is a 90% chance of snow” is wrong. Key difference.
3. If your weatherman made a yes/no prediction statement, like it is going to snow tomorrow, and then it does not snow — they are wrong. In this case the thing they are wrong about is whether or not it snowed.
Separately, a ‘forecast’ is generally just a statement of probabilities. When you say you think your favorite football team will win, you are not making a forecast — you are predicting an outcome. When some ESPN analytics software says there is an 88% chance your team will win, it is making a forecast, and if your team loses the ESPN tool was not wrong.
Send to anyone you think doesn’t get this.