3 ms·

Care to elaborate?

by aeze 10y ago

Care to elaborate?

InitialLastName 10y ago
Their model of the 2016 US Presidential election reflected the same polling error that all the other models showed. They gave Trump a 30% chance of winning, along with some commentary about ways that systemic error could easily swing things in his direction. Trump then won the election with results well-within their error margins, but that wasn't enough for some people. It's hard for some members of our society to acknowledge that 30%-chance events happen on occasion, so 538 takes the fall.
- oh_sigh 10y ago
  Their election night coverage was one of better ones, but in the lead up to the election, Nate Silver got a lot wrong about Trump. He even had to write a mea culpa about it. Perhaps that is what OP is talking about?
  - tacomonstrous 10y ago
    >but in the lead up to the election, Nate Silver got a lot wrong about Trump. He even had to write a mea culpa about it. That was for the Republican primary.
- geofft 10y ago
  It's not even clear if there were systemic polling errors: there were two last-minute changes that affected the election. There was the unexpected last-minute letter from Comey on October 28, and 538 did in fact report Clinton's chances to be in freefall from then until days before the election. There was also the unexpected competence of Kushner's data-science work right at the end to "micro-target" specific undecided voters and likely Hillary voters (unexpected in part because Trump had expressed skepticism at those strategies, and that organization was underfunded). It certainly seems possible that 538 had the data exactly right all along.
- LordKano 10y ago
  I suppose it's neither here nor there now but I strongly suspect that both candidates' internal polling showed much closer numbers than the nationally published polls. In the last week of the campaign, Hillary was still visiting states that should have been in the bag for her and Trump was still visiting states that he should have written off as lost. Maybe they were just tactical anomalies but I think it makes more sense if they both knew that the race was closer than everyone else thought.
- sigstoat 10y ago
  > It's hard for some members of our society to acknowledge that 30%-chance events happen on occasion, so 538 takes the fall. indeed, with a well-calibrated predictor, they should happen 30% of the time.

Care to elaborate?

by aeze 10y ago

Care to elaborate?

Their model of the 2016 US Presidential election reflected the same polling error that all the other models showed. They gave Trump a 30% chance of winning, along with some commentary about ways that systemic error could easily swing things in his direction. Trump then won the election with results well-within their error margins, but that wasn't enough for some people. It's hard for some members of our society to acknowledge that 30%-chance events happen on occasion, so 538 takes the fall.

oh_sigh 10y ago
Their election night coverage was one of better ones, but in the lead up to the election, Nate Silver got a lot wrong about Trump. He even had to write a mea culpa about it. Perhaps that is what OP is talking about?
- tacomonstrous 10y ago
  >but in the lead up to the election, Nate Silver got a lot wrong about Trump. He even had to write a mea culpa about it. That was for the Republican primary.
geofft 10y ago
It's not even clear if there were systemic polling errors: there were two last-minute changes that affected the election. There was the unexpected last-minute letter from Comey on October 28, and 538 did in fact report Clinton's chances to be in freefall from then until days before the election. There was also the unexpected competence of Kushner's data-science work right at the end to "micro-target" specific undecided voters and likely Hillary voters (unexpected in part because Trump had expressed skepticism at those strategies, and that organization was underfunded). It certainly seems possible that 538 had the data exactly right all along.
LordKano 10y ago
I suppose it's neither here nor there now but I strongly suspect that both candidates' internal polling showed much closer numbers than the nationally published polls. In the last week of the campaign, Hillary was still visiting states that should have been in the bag for her and Trump was still visiting states that he should have written off as lost. Maybe they were just tactical anomalies but I think it makes more sense if they both knew that the race was closer than everyone else thought.
sigstoat 10y ago
> It's hard for some members of our society to acknowledge that 30%-chance events happen on occasion, so 538 takes the fall. indeed, with a well-calibrated predictor, they should happen 30% of the time.