nc_tiger

Tuesday can't get here fast enough.

what does this mean

there's 1 clear example of bad input to a 538 model: the 2016 presidential election. they've proven to do a good job predicting results over the last decade.

538 has never tried to be "wrong" or "right", just apply probabilities. They also never give 0 because it's impossible to give zero, the mathematics of probability prevent a 0% chance in an election based on people voting when they haven't asked every single one the voters. They don't avoid saying 0 to avoid being wrong, it's just math.

quote:

---

The issue being those "priors and available polls" are flawed to begin with.

---

of course. the pollsters (and downstream models like 538) claim to have made adjustments after learning things post-2016 election. we'll see on nov 3

quote:

---

He had Trump down in the teens, if not lower, before Election Day.

---

and at another point before Election Day he had the odds near 50/50

to build a statistical model that's based on a set of priors and available polls to predict the probability of election outcomes, not the outcome it self.

538/Nate Silver is not a pollster. the pollsters are the ones that required the most correcting, though 538 has said they've made adjustments to how the model reads the polls (bc of how off the polls were in 16)

print can still be web, just not tv/radio

curious to see where the poli board reads the news

eta: by print news I mean anything that is not tv/radio

People don’t want to get sick, more at 11