They gave Trump a 30% chance of winning when the NYT said 2%. I fail to see how 538 got owned
Edit: Trump had a 10% chance of winning before the Comey letter. I think things changed a lot in the last week and other pollsters were slow to factor that in
I hate seeing people give them partial credit for being less wrong in 2016. It's their job to predict it, and they predicted wrong. "But they were 20% less wrong than other sites!" So what? I don't want to subscribe to the notion that their predictions work because they're less wrong but still wrong.
It's incorrect because they didn't get it right. No one is saying 30% can't happen, but we're saying why are you putting value into the predictors if they consistently make predictors that turn out wrong.
Then if we're talking about an event that will only happen in these circumstances one time, what exactly is the value of these types of predictions? They could have given Trump a 1% chance of winning and still said "well that's not impossible". That's not incorrect, but it sure sounds a lot like ass-covering and it makes you wonder why we even listen to this
It's valuable because it's interesting. Might help you plan for future events. I might take an umbrella if there is a 30% chance of rain but not if there is a 1% chance
Keep in mind, it's not just a prediction of the overall result, they are also forecasting each state, D.C., and the popular vote, so it's more like an ensemble of 52 predictions, so you can also judge the accuracy of each of those forecasts as well.
I mean, not really, but you're essentially saying "If he said it was 50-50 instead of likely Hillary would he have been wrong?" Well, no, but he didn't say that.
I thought they stopped doing predictions cus they kept getting owned by reality?
They gave Trump a 30% chance of winning when the NYT said 2%. I fail to see how 538 got owned
Edit: Trump had a 10% chance of winning before the Comey letter. I think things changed a lot in the last week and other pollsters were slow to factor that in
I meant more recently than 2016, they've largely been getting less correct in their predictions.
Second to that, comparing them to NYT is probably unfair cus the NYT is more often incorrect than correct.
What other publication or group gave Trump more than a 2 or 3% chance of winning. I said NYT but basically every other predication was the same
deleted by creator
Yeah 30% is basically a one in three chance of winning, that's fucking huge in statistics actually.
I hate seeing people give them partial credit for being less wrong in 2016. It's their job to predict it, and they predicted wrong. "But they were 20% less wrong than other sites!" So what? I don't want to subscribe to the notion that their predictions work because they're less wrong but still wrong.
How is 30% odds incorrect though? You can still die playing Russian Roulette despite the odds being low
It's incorrect because they didn't get it right. No one is saying 30% can't happen, but we're saying why are you putting value into the predictors if they consistently make predictors that turn out wrong.
If a weatherman says the chance of rain is 30%, and it rains, his forecast is not necessarily incorrect. You need repeated trials to confirm accuracy.
If 10/10 times he predicts a 30% chance of rain it rains, then yeah, he's a bad forecaster.
If 3/10 times he predicts 30% chance of rain it rains, then he's pretty accurate
Then if we're talking about an event that will only happen in these circumstances one time, what exactly is the value of these types of predictions? They could have given Trump a 1% chance of winning and still said "well that's not impossible". That's not incorrect, but it sure sounds a lot like ass-covering and it makes you wonder why we even listen to this
It's valuable because it's interesting. Might help you plan for future events. I might take an umbrella if there is a 30% chance of rain but not if there is a 1% chance
I think they're refering to predictions for a presidential election, not weather predictions
In that case it's just interesting. There's nothing you can do with this information
Keep in mind, it's not just a prediction of the overall result, they are also forecasting each state, D.C., and the popular vote, so it's more like an ensemble of 52 predictions, so you can also judge the accuracy of each of those forecasts as well.
Giving 30% odds to an event that happens is not incorrect though. It's not getting it wrong
I don't really get this logic. If he had given Trump a 49.9% chance of winning would he have still been wrong?
I mean, not really, but you're essentially saying "If he said it was 50-50 instead of likely Hillary would he have been wrong?" Well, no, but he didn't say that.
We can't even say they were wrong. Things with a 30 percent chance of happening happen all the time.
How often? I can't say.
You don't understand how statistics work
Removed by mod
Have they done statistical analyses on how meaningful a statistical prediction is this far out from the election?
The model says if the election were held tomorrow Biden's chances would be 93%.
I can't think of a more meaningless metric.
What if the election happened yesterday
Then hopefully your model will take that information into account.
No shit, Einstein. My minor was in stats.
Just because some dork with glasses can work an excel spreadsheet doesn't make him a good statistician not does it mean that his models are correct.