Nate Silver (538) On This Year's Election (1 Viewer)

[Grace Under Pressure]

I followed 538 every day for months in 2016. Linked it up in the main election thread here almost every day. I’ll take my hits on calling him out now. I don’t think 538 (or RCP for that matter) did anyone any favors given where we are now. I think it offered a false sense of complacency, she was ahead in his model in nearly every swing state and nationally. I think he relied too heavily on aggregation and was slow to move late. Did he give Trump a better chance than most? Sure. Was he right? Not really, maybe more right than some. Was he tracking hundreds of players at multiple positions like FBG? No. He had one call to make. 71-29, but he gets a lot of credit for the 29? I understand polling and statistics, probable outcomes and margins of error. Let’s just say I’m not a fan. But I hear where Dinsy and Woz, etc. are coming from and I respect that POV.

 

[hagmania]

I understand polling and statistics, probable outcomes and margins of error.

I have no dog in this fight but I think quite plainly that only a small percentage of people in this thread, this nation, this world actually do.

If you are one of the lucky few good on you.

 

[Zow]

I followed 538 every day for months in 2016. Linked it up in the main election thread here almost every day. I’ll take my hits on calling him out now. I don’t think 538 (or RCP for that matter) did anyone any favors given where we are now. I think it offered a false sense of complacency, she was ahead in his model in nearly every swing state and nationally. I think he relied too heavily on aggregation and was slow to move late. Did he give Trump a better chance than most? Sure. Was he right? Not really, maybe more right than some. Was he tracking hundreds of players at multiple positions like FBG? No. He had one call to make. 71-29, but he gets a lot of credit for the 29? I understand polling and statistics, probable outcomes and margins of error. Let’s just say I’m not a fan. But I hear where Dinsy and Woz, etc. are coming from and I respect that POV.

If a poker player shoves all in pre-flop with AA and loses heads up to 7s8s, was he wrong?

 

[Maurile Tremblay]

Was he right? Not really, maybe more right than some.

I don't have an opinion on whether 29% was the right chance to assign. I don't know if that was too high, too low, or roughly accurate.

I am interested in hearing from people who do have an opinion -- what was the right percentage, and how do you figure it?

(To me, the best argument that 29% was too low is simply that Trump won, so any number below 100% should be updated upward on that basis. That is evidence, but it is mathematically very weak evidence. The best argument that 29% was too high is that the general consensus of modelers arrived at a lower figure, as did the betting markets. I believe Silver himself said that betting markets are probably more accurate, generally, than his own model is. Outliers can generally be viewed with some skepticism, and 29% was an outlier on the high side. The argument that 29% may have been about right is that there are plausible, but not compelling, arguments in both other directions.)

 

[Joe Summer]

If a poker player shoves all in pre-flop with AA and loses heads up to 7s8s, was he wrong?

The problem with this comparison is that it assumes that Silver's "29% chance" was fundamentally sound to begin with.

But election predictions are not just based on simple mathematical formulas; they also involve subjective factors — things like "How much weight should we give to the latest poll?" or "How much will Comey's announcement affect the vote?" or "Is there really such a thing as a Shy Trump Voter?"

And I don't think Silver — or any analyst, even the ones that predicted a Trump victory — did a very good job of properly weighting those subjective factors.

 

[Zow]

The problem with this comparison is that it assumes that Silver's "29% chance" was fundamentally sound to begin with.

But election predictions are not just based on simple mathematical formulas; they also involve subjective factors — things like "How much weight should we give to the latest poll?" or "How much will Comey's announcement affect the vote?" or "Is there really such a thing as a Shy Trump Voter?"

And I don't think Silver — or any analyst, even the ones that predicted a Trump victory — did a very good job of properly weighting those subjective factors.

How would you have done it differently?

 

[Joe Summer]

How would you have done it differently?

I don't know.

But here's something to consider: if Nate Silver gave Clinton a 71.4% chance of winning based solely on his analysis of the polls, and he gave Clinton a 71.8% chance of winning based on his custom "polls plus" analysis (link), doesn't that imply that he's not very good at analyzing the "plus" stuff?

 

[Maurile Tremblay]

But here's something to consider: if Nate Silver gave Clinton a 71.4% chance of winning based solely on his analysis of the polls, and he gave Clinton a 71.8% chance of winning based on his custom "polls plus" analysis (link), doesn't that imply that he's not very good at analyzing the "plus" stuff?

Strictly speaking, the answer is no, it doesn't imply that.

Whether the polls-plus model should depart significantly from the polls-only model in a particular case depends on (a) how much of the "plus" stuff is signal rather than noise (since lots of "plus" stuff is just noise), and (b) whether the portion that's signal tends to point somewhat consistently in one direction or other other rather than pointing in both directions at once.

 

[parasaurolophus]

I am reminded of the movie swingers where Vince Vaugh and Jon Favreau discuss the specific instance of doubling down on 11.

 

[sho nuff]

I am reminded of the movie swingers where Vince Vaugh and Jon Favreau discuss the specific instance of doubling down on 11.

How can you say always? I lost!!!

 

[Sabertooth]

If anyone thinks this is a lock on either side then they have not learned from history.

I have not met a single person from the left who thinks this is a lock.  Only my Trump friends continue to say it’s a lock.