iluvatar5150
Well-Known Member
- Aug 3, 2012
- 25,257
- 24,155
- Country
- United States
- Faith
- Christian
- Marital Status
- Married
- Politics
- US-Democrat
In post 47, you said, "I don’t know where 71:29 is close, let alone very close."
In post #8, you said, "The poll is saying out of 100 people surveyed 87 are voting Biden and 13 are voting Trump."
In post #105, you said, "20 sided dice aside, a poll that shows an 85:15 is a virtual landslide, not a very close race. Avery close race would be 48:46."
In post #114, you said, "87:13 are not close odds"
It's quite clear to the rest of us that you fundamentally don't (or at least didn't until very recently) understand how the 538 model works, because in the context of that model, every one of those statements of yours is absurd.
Yes, they got the electoral college prediction wrong, but if you had read and understood the links provided to you in post #115 where they broke down what went wrong, you'd understand why it was still a near miss on their part. One big factor that you seem to not be grasping at all is the way in which polling error is correlated. If polling errors are uncorrelated (i.e. the error in one poll has nothing to do with the error in another poll), then if 10 polls are taken and for some reason they're all off by, say, 3 points, those errors go in somewhat random directions such that when you average all the polls together, the errors cancel out to some degree. When polling errors are correlated, then the errors tend to all go in the same direction. In politics, polling errors are correlated and there were a lot of states in which Trump's support wound up being under-reported. And with the way the electoral college works, a couple-point swing in several states results in big shifts in electoral counts.
Upvote
0
