Red Mosquito

BurtReynolds wrote:

Bi_3 wrote:Apparently Nate Silver is in a twitter fight with someone over mathematical forecasting. Hurray humanity.

What do you guys think of Nate Silver's predictions for Clinton winning being so much lower than other models? I guess we'll find out soon enough who is right, but it seems kinda screwy.

From what they've said on the podcast, the differences are because the 538 model imagines that polling errors and polling accuracy correlates between states that are regionally and demographically similar. So, for example: if Clinton beats her polls in Wisconsin, she'll probably also beat her polls in Minnesota, because the two states are quite similar (in percentage of white people and minorities, types of industries, historical voting record, and shitty accents). But if there's been a polling error in Wisconsin, there's also a great chance that there's been a polling error in Minnesota. I don't know exactly how the math works, but apparently most other prediction models don't take this into account, and that affects Trump's chances.

BurtReynolds wrote:Hillary Star on Walk of Fame Destroyed by Madman In Shocking Display of Violence Against Women

You should list all the examples of times when Clinton advocated for violent or intimidating acts towards others. I'm sure we will find a trend that explains the discrepancy in reactions.

Simple Torture wrote:

BurtReynolds wrote:

Bi_3 wrote:Apparently Nate Silver is in a twitter fight with someone over mathematical forecasting. Hurray humanity.

What do you guys think of Nate Silver's predictions for Clinton winning being so much lower than other models? I guess we'll find out soon enough who is right, but it seems kinda screwy.

From what they've said on the podcast, the differences are because the 538 model imagines that polling errors and polling accuracy correlates between states that are regionally and demographically similar. So, for example: if Clinton beats her polls in Wisconsin, she'll probably also beat her polls in Minnesota, because the two states are quite similar (in percentage of white people and minorities, types of industries, historical voting record, and shitty accents). But if there's been a polling error in Wisconsin, there's also a great chance that there's been a polling error in Minnesota. I don't know exactly how the math works, but apparently most other prediction models don't take this into account, and that affects Trump's chances.

I think both of these responses contain a modicum of truth. It's hard to not be conservative about predictions when people have distilled your entire reputation to "accurate."

Nate Silver looks like a comedian who had a semi-successful sitcom in 2003 and now hosts a podcast on Earwolf

Also, how good is the Earwolf podcast in this description? Like, I Was There Too? Comedy Bang Bang Normal? Or Comedy Bang Bang with Paul F Thompkins?

tragabigzanda wrote:

McParadigm wrote:Also, how good is the Earwolf podcast in this description? Like, I Was There Too? Comedy Bang Bang Normal? Or Comedy Bang Bang with Paul F Thompkins?

One of the Nerdist episodes where they don't even have a guest.

Or Chris Hardwick interviewing himself.

tragabigzanda wrote:

BurtReynolds wrote:

Bi_3 wrote:Apparently Nate Silver is in a twitter fight with someone over mathematical forecasting. Hurray humanity.

What do you guys think of Nate Silver's predictions for Clinton winning being so much lower than other models? I guess we'll find out soon enough who is right, but it seems kinda screwy.

I think he's hedging his bets so his brand isn't tarnished if it's a really close vote on Tuesday. If Hillary loses, he can say "Told you so." If she wins in a landslide, he can still spin as him being cautiously conservative with his estimates. But if he follows the pack, and then she loses? He's got to find a new job.

I think this distorts what Nate Silver's project actually is. He's not trying to "guess" who the next president will be; his goal from the beginning has been to build a mathematical model for predicting presidential election winners based on polling and other "fundamentals," with the full knowledge that any "misses" cycle-to-cycle would be factored into the next model to help make predictions more accurate. They're not "his estimates" and he's not "following the pack," per se, because the model was built before anyone knew Clinton and Trump were going to be the nominees. If Clinton goes into Tuesday morning ahead in their model 65% to Trump's 35% (about where they are right now), I'm sure they'll publish an article saying, "538 predicts that Hillary Clinton will be the next President of the United States." But that's based entirely on the model; if it ends up being wrong, I'm sure the first article to be published the Wednesday morning will say, "Here is what the model probably missed, and here's how we're going to adjust it next time." They're working with an incredibly small sample size here--I think their data only goes back to 1980 because of the availability of polling data, but it could really only go back to 1972 at the furthest, because that's when the modern primary system really began--so there are going to be ups and downs, for sure, until there's more data.

I trust HuffPo's model about as much as a Breitbart online poll.

BurtReynolds wrote:I trust HuffPo's model about as much as a Breitbart online poll.

I assume they currently give Beyonce a 50% chance of winning the presidency.

theplatypus wrote:Nate Silver looks like a comedian who had a semi-successful sitcom in 2003 and now hosts a podcast on Earwolf

I sometimes totally test Twitter material on the board, too.

tragabigzanda wrote:

Simple Torture wrote:

tragabigzanda wrote:

BurtReynolds wrote:

Bi_3 wrote:Apparently Nate Silver is in a twitter fight with someone over mathematical forecasting. Hurray humanity.

What do you guys think of Nate Silver's predictions for Clinton winning being so much lower than other models? I guess we'll find out soon enough who is right, but it seems kinda screwy.

I think he's hedging his bets so his brand isn't tarnished if it's a really close vote on Tuesday. If Hillary loses, he can say "Told you so." If she wins in a landslide, he can still spin as him being cautiously conservative with his estimates. But if he follows the pack, and then she loses? He's got to find a new job.

I think this distorts what Nate Silver's project actually is. He's not trying to "guess" who the next president will be; his goal from the beginning has been to build a mathematical model for predicting presidential election winners based on polling and other "fundamentals," with the full knowledge that any "misses" cycle-to-cycle would be factored into the next model to help make predictions more accurate. They're not "his estimates" and he's not "following the pack," per se, because the model was built before anyone knew Clinton and Trump were going to be the nominees. If Clinton goes into Tuesday morning ahead in their model 65% to Trump's 35% (about where they are right now), I'm sure they'll publish an article saying, "538 predicts that Hillary Clinton will be the next President of the United States." But that's based entirely on the model; if it ends up being wrong, I'm sure the first article to be published the Wednesday morning will say, "Here is what the model probably missed, and here's how we're going to adjust it next time." They're working with an incredibly small sample size here--I think their data only goes back to 1980 because of the availability of polling data, but it could really only go back to 1972 at the furthest, because that's when the modern primary system really began--so there are going to be ups and downs, for sure, until there's more data.

I'm not entirely sold on this. I get that Silver has a mission to build a working model, but that mission is surely compromised by the need for the site to make money -- whether it's via site traffic, or him getting paid as a talking head on the news circuit. And this election is already a veritable game changer, and emotions are high. If he comes out of this election cycle with egg on his face, his brand is going to take a major hit on credibility. No amount of "We'll do better next time because..." is going to reclaim his current status as the US's preeminent predictor of elections, and he knows that.

I do understand why you're predicting the public's reaction that way, trag. In the end, it is his model that he's built after years of research and study, so if people want to say, "Your model sucks," he does have to bear it. In a way, I think it might have been detrimental to his project in the long run that the model went 50-for-50 in 2012, since most people are going to expect that in every cycle moving forward.

He's also going to run into issues going forward with the fact that his model is increasingly historical in design influence. Early voting's growing prevalence alone makes his accuracy harder to maintain, the closer you get to the election.

McParadigm wrote:He's also going to run into issues going forward with the fact that his model is increasingly historical in design influence. Early voting's growing prevalence alone makes his accuracy harder to maintain, the closer you get to the election.

Yeah, they've sort of explained that "early voting" is "in the model" in the sense that pollsters who call people who have alreasy voted will mark them as "likely voters." That's a bit weak, I think, but it's really a hole that the pollsters will have to patch up, I think. They've also got to deal with the disappearance of landlines, which has been more pronounced during this final week

tragabigzanda wrote:

theplatypus wrote:Nate Silver looks like a comedian who had a semi-successful sitcom in 2003 and now hosts a podcast on Earwolf

What are your plans for NYC, Jorg?

I'm seeing some music, eating some food, and sampling a bit of my friend's new life. Only spending a week there this time around

Im voting for Hillary because David Simon told me to.