patrick.net

Hubris did her in, she thought she was entitled to the world.

Uh no, the last data before the night of the election, 538 had Hillary winning at 71%.
https://projects.fivethirtyeight.com/2016-election-forecast/

Even if it was 66% as you claim, that's a statistical landslide. No one had it close, absolutely no one.

lol--OK 29%. That's not a landslide. It's basically a one in three chance of Trump winning. Most reporters didn't understand probabilities, dependent variability, etc. so they didn't understand how to properly model the various potential outcomes. All it took was a normal polling error for Trump to win which is what Nate basically said. Normal polling error in favor of Trump and he wins. No polling error--Clinton wins close race. Polling error in Clinton's favor and she wins in a landslide.

joeyjojojunior says

lol--OK 29%. That's not a landslide. It's basically a one in three chance of Trump winning. Most reporters didn't understand probabilities, dependent variability, etc. so they didn't understand how to properly model the various potential outcomes. All it took was a normal polling error for Trump to win which is what Nate basically said. Normal polling error in favor of Trump and he wins. No polling error--Clinton wins close race. Polling error in Clinton's favor and she wins in a landslide.

So 71% chance to win is a "close" statistical race?

Normal "polling errors" are 3-5% MAX assuming a flat probability density, I usually see less in most MSM polls. So that would bring Trump's chance to win with a huge polling error using 538's data at 34%.

Yeah, super close dude!

"So 71% chance to win is a "close" statistical race? Normal "polling errors" are 3-5% MAX assuming a flat probability density, I usually see less in most MSM polls. So that would bring Trump's chance to win with a huge polling error using 538's data at 34%. Yeah, super close dude!"

You obviously don't understand how the model works and what the percentages mean.

538 uses polling to create a model that arrives at a probability of either candidate getting 270 electoral votes.

A polling change of 2-3% in favor of Trump, which is normal, would have swung 538s forecast completely and they would have had Trump as the likely winner.

joeyjojojunior says

You obviously don't understand how the model works and what the percentages mean.

538 uses polling to create a model that arrives at a probability of either candidate getting 270 electoral votes.

A polling change of 2-3% in favor of Trump, which is normal, would have swung 538s forecast completely and they would have had Trump as the likely winner.

No I don't understand the specialized model that Nate Silver is celebrated for because it's complete woo woo that predicted there was a 71% chance Hillary would win.

I don't try to spend time learning compromised statistical models.

A 2% change would have swung the model completely? That model is shit.

"A 2% change would have swung the model completely? That model is shit."

lol--a 2% swing changed the election. So the model was proven correct. Your understating is shit.

And you don't have to understand the inner workings of the model, just what it represents and what it's forecasting. You seem to lack that understanding.

joeyjojojunior says

"A 2% change would have swung the model completely? That model is shit."

lol--a 2% swing changed the election. So the model was proven correct. Your understating is shit.

Uh no. The election never "changed". Hillary was NEVER going to win the states Silver thought she would win. Nothing "swung or changed", Silver's data was complete shit.

Like I said, I don't spend time on Pseudo-scientist like Nate Silver who got the election completely wrong. His model was shit, most importantly, his data was shit. His prediction was shit as we all saw in November 2016.

Good to know that you can shit on something you have no understanding of.

joeyjojojunior says

Good to know that you can shit on something you have no understanding of.

Good to know you will eat pseudo-science shit that was proven wrong, and try to defend it at all cost, despite it being proven to be shit.

Nate Silver before the election:

"So we’re left to argue about the probability of an unforeseen event, or a significant polling error. It’s perhaps significant that almost no matter what news has occurred, and there’s been a lot of it — terrorist attacks, mass shootings, foreign crises, her email scandal, the Wikileaks dump, her Sept. 11 health scare — Clinton has almost always led Trump in the polls, although there have certainly been times when the election was close. What if her State Department emails are sitting on one of Julian Assange’s servers? That would be interesting, I suppose. But there are also October (or November) surprises that could work against Trump: more accusations from women, more damaging videotapes, further leaking of his tax records.

The other possibility is a massive polling failure. There aren’t really any direct precedents for a candidate coming back from this far down to win an American presidential election, although you can make a few loose analogies. Harry Truman’s comeback over Thomas Dewey in 1948 almost works as a comparison, but Truman wasn’t coming from as far behind as Trump is, and there was much less polling in 1948. Ronald Reagan had a significant late surge against Jimmy Carter in 1980, but he was ahead beforehand — and the surge came in large part because of a debate that occurred just one week before the election, whose impact was too late to be fully reflected in the polls. If Trump was going to have a Reaganesque surge, in other words, it probably would have started with a commanding performance in last night’s debate — and not another loss.

Brexit? Even that comparison doesn’t really work. The final polls showed a toss-up between the United Kingdom leaving the European Union or remaining in it, and “leave” eventually won by 4 points. If the polls were biased against Trump by that much in this election, he’d still lose, by a margin approximating the one by which Mitt Romney lost to President Obama four years ago. The primaries? They’re a reminder that one ought to be humble when making predictions. But the polls pegged Trump just fine — in fact, slightly overestimating his performance in many early states such as Iowa."

Even Nate Silver admits here that his data could be shit. Which it was. That I understand, and apparently you don't.

"Good to know you will eat pseudo-science shit that was proven wrong, and try to defend it at all cost, despite it being proven to be shit."

If you took the time to read and understand the model, you'd know that it was actually proven correct. Silver wrote before the election that a normal polling error in Trump's favor would give him the Presidency. Given that that is exactly what happened, Nate would seem to be proven correct.

"Nate Silver before the election:"

Nice--when you say before the election, you mean, WAYYYYYY before the election.

joeyjojojunior says

If you took the time to read and understand the model, you'd know that it was actually proven correct. Silver wrote before the election that a normal polling error in Trump's favor would give him the Presidency. Given that that is exactly what happened, Nate would seem to be proven correct.

I've actually been trying to goad you into explaining why you think his model is accurate. I see that you don't actually know much about his model since you aren't going into specifics. So I'll help you out.

I actually understand Nate Silver's model more than you do, it's driven mostly by demographic and historical data and aggregated polling.

You honestly don't see a huge red flag with that methodology?

"I've actually been trying to goad you into explaining why you think his model is accurate. I see that you don't actually know much about his model since you aren't going into specifics. So I'll help you out. I actually understand Nate Silver's model more than you do, it's driven mostly by demographic and historical data and aggregated polling. You honestly don't see a huge red flag with that methodology?"

bwahahahahaha. Are you a CIC account? You've been trolling me.. Please stop. You're making me laugh out loud.

Nate has multiple models with different inputs. There is no problem with the methodology--it was proven correct. Here's his article on the eve of the election:

More from that article--it's actually quite prescient:

http://fivethirtyeight.com/features/final-election-update-theres-a-wide-range-of-outcomes-and-most-of-them-come-up-clinton/?ex_cid=2016-forecast

You said "a 2% change swung the entire election."

That's not what Nate Silver is saying in that summary. You actually don't understand what he's talking about, let alone his model.

What's he really saying is that the aggregated polling numbers are within a normal polling error, but for that to happen, ALL 22-26 national polls he is aggregating would also have to be off to create that error in the aggregate. We're not talking just one or two firms here, we're talking the majority of the polls he's inserting into his model.

That would mean basically every single national poll was off, or the data was shit. Either way, in a REAL model not tainted by polling bias, that would be practically mathematically impossible (and has never happened in the modern era). Do you disagree?

So either his model is off (which you don't seem to accept) or his data was shit (which is what I've been saying all along). So which is it?

joeyjojojunior says

More from that article--it's actually quite prescient:

Whats prescient about that? He predicted black people who voted for the charasmatic black guy in 2008 and 2012 might not vote for the hoarse sounding old white woman in 2016.

Woooo. So magical and prescient.

I predicted that same exact thing last year. So did many others. I don't see you fawning over them as geniuses like Nate Silver.

"That's not what Nate Silver is saying in that summary. You actually don't understand what he's talking about, let alone his model."

hahaha. Sorry, I said 2% and Nate said 3%. Please forgive me.

And of course I disagree. Your analysis is completely off. You're wrong. His model worked perfectly. And his data was fine--within the normal range of error for a Presidential election. We had a very high number of undecided in 2016 and they broke heavily for Trump. (and what has never happened in the modern era?)

joeyjojojunior says

And of course I disagree. Your analysis is completely off. You're wrong. His model worked perfectly. And his data was fine--within the normal range of error for a Presidential election. We had a very high number of undecided in 2016 and they broke heavily for Trump. (and what has never happened in the modern era?)

So 95% of the polling firms had bum data? In his 3 month rolling aggregate only Marist College had it correct before the election. You don't see that as highly unlikely?

Why don't you just admit he had shit data? That's not even a critical comment against your hero, he even claims before the election that it's a distinct possibility (we all now know that the polls he was aggregating were completely biased and tainted).

Deplorable.

I love the rewriting of history. The Media was dancing around with Poll Numbers and Election Models like Palestinians with the bloody shirt of a Hamas Terrorist, claiming that only Nutjobs disbelieved in the DATA - the DATA mind you, That Can't Be Dismissed, Hard Numbers, dontcha understand, Don't Be a Dumbass Math Denier, Don't even Bother voting to Vote Trump!

Now it's like "We NEVER said Hillary was going to win in a landslide".

Ironically, the promotion of inevitable Hillary might have caused her to lose by giving people permission to stay home and not vote.

I think my "Cat Lady will Netflix and Chill" prediction was better than Nate.

"So 95% of the polling firms had bum data? In his 3 month rolling aggregate only Marist College had it correct before the election. You don't see that as highly unlikely?"

Nope--there is always error in polling data. 2016 was within the expected error, especially when you consider that there was a much larger undecided % vs. the last 4 election cycles. National polling had it pretty close--Clinton did win by about 2 points.

"I love the rewriting of history. The Media was dancing around with Poll Numbers and Election Models like Palestinians with the bloody shirt of a Hamas Terrorist, claiming that only Nutjobs disbelieved in the DATA - the DATA mind you, That Can't Be Dismissed, Hard Numbers, dontcha understand, Don't Be a Dumbass Math Denier, Don't even Bother voting to Vote Trump!"

Most the media are idiots and didn't understand how to interpret the data.

joeyjojojunior says

Nope--there is always error in polling data.

Yes, but 95% of the national polls were off with only one being correct? Do you know how statistically improbable that is in a supposedly unbiased, untainted sample? Go ahead, calculate it.

You still think the data he was inserting into his model was sound?

"Yes, but 95% of the national polls were off with only one being correct? Do you know how statistically improbable that is in a supposedly unbiased, untainted sample? Go ahead, calculate it. You still think the data he was inserting into his model was sound?"

First--it's been shown that late deciders broke for Trump, so it's not clear that any polls were "wrong" They measured a value at a point in time different than the point in time in which the election votes were cast.

Second-- Clinton won by 2.1%. The average of the national polls taken on Monday was 2.9%. That's pretty damn good.

joeyjojojunior says

"Yes, but 95% of the national polls were off with only one being correct? Do you know how statistically improbable that is in a supposedly unbiased, untainted sample? Go ahead, calculate it. You still think the data he was inserting into his model was sound?"

First--it's been shown that late deciders broke for Trump, so it's not clear that any polls were "wrong" They measured a value at a point in time different than the point in time in which the election votes were cast.

Second-- Clinton won by 2.1%. The average of the national polls taken on Monday was 2.9%. That's pretty damn good.

Yes, we knew late voters voted for Trump LATER, after the election was in full swing because of exit polling. That has nothing to do with Nate Silver's projection. This is why the New York Times probability graph changed so dramatically in 24 hours, but that's just proving my point about how bad the data was from aggregated national polls. They were ALL off, including Nate Silver's prediction.

I seriously think you would be doing yourself some justice by taking a simple Lynda.com course online for free and learning about the concept of Discrete probability distribution. Without you having SOME basic knowledge of how probability is actually calculated, this is not a very productive discussion for either of us because you lack basic knowledge to understand the basis of what we're debating. This is not a critique of you as a person, I'm sure you're intelligent, but I graduated from Wharton Business school and spent a decades long career in banking as a Senior Analyst at places like Goldman Sachs, a lot of what we're talking about, and what you're clearly missing are very, very basic concepts. The fact that only Marist College was the only data point out of a 3 month rolling period that even came close to showing a Trump win vs 95%+ of every other national poll is very, very improbable in light of the results (Trump win) unless the sampling was tainted (which we now know it was). That's why Nate Silver's prediction was so wrong.

CBOEtrader says

Straw Man says

Who gives a fuck?

Can't wait until DNC data team starts fingering hillary

Ewww, that's disgusting!

No--it means that people that polled as undecided actually decided after the poll was taken that they would vote for Trump. The poll wasn't biased or wrong, people changed from undecided to Trump. Data was correct at the time it was taken.

All the rest of your condescending post is obviously you trying to cover up for your mistakes.

I'm quite certain that I am much more well versed in probability and statistics than you are. You don't even seem to be able to differentiate between a raw polling percentage and a percentage derived from model calculating a candidate's probability of winning 270 electoral votes.

joeyjojojunior says

Second-- Clinton won by 2.1%

Earth to jojo! Trump won the election! He didn't win a poll and didn't have to! Popular vote only counts at the state level not the national level. Duh! The OP stated that HRC ignored inconvenient data from three critical state polls! Maybe they were so focused on the favorable national polls that they forgot the rules of the election?

"Earth to jojo! Trump won the election! He didn't win a poll and didn't have to! Popular vote only counts at the state level not the national level. Duh! The OP stated that HRC ignored inconvenient data from three critical state polls! Maybe they were so focused on the favorable national polls that they forgot the rules of the election?"

No shit. And Trump did lead in polls, national and state.

And I'm agreeing with the DNC. Read what I write please.

Also--could you post this Marist poll that you think had the correct methodology?

Here's what I find:

http://maristpoll.marist.edu/115-final-push-clinton-and-trump-close-nationally/

Looks about the same as most polls although it underestimated Clinton's support nationally.

try prunes?
joeyjojojunior says

No shit.

joeyjojojunior says

The poll wasn't biased or wrong, people changed from undecided to Trump.

Yes, they were biased and wrong. This isn't even something that's debatable anymore since we have the data and have gone over it for months since the election. In fact these same polls CONTINUE to oversample/undersample for everything from approval polls to issue polls. Take for instance the ABC News/WaPO National Poll in January gauging Trump's favorables, a poll that featured prominently in Silver's data. Also consider that according to Gallup the most recent breakdown of their party identification statistic breaks down like this; 29% Republican / 40% Independent / 28% Democrat / 3% other.

31% democrats, 23% republican, and 37 percent independent. That's a 8+ point SWING towards Democrats in that poll alone. You think that poll is not tainted by bias?

CNN has also been caught loading 8-10 point partisan gaps into their polls regularly.

Like I said, Lynda.com is free, and I'll chip in for an upgraded account so you can take a few courses.

It's not even debatable anymore?? Really? The average of polling data on Nov. 8th was Clinton +2.9 and the final election tally was Clinton +2.1. It's not possible that undecided voters accounted for 0.8% of the vote in Trump's favor?

You really are a CIC alt, aren't you? That was his favorite topic--the supposed polling bias in favor of Dems. Only the results are amazingly accurate for having any bias. Or are you saying this bias accounted for 0.8% of the vote?

joeyjojojunior says

You really are a CIC alt, aren't you? That was his favorite topic--the supposed polling bias in favor of Dems.

Are you actually denying that ABC News/WaPO and CNN engaged in poll loading?

Are you actually saying that pollsters whose job and livelihood depends on accurate results purposely made such an obvious mistake as oversampling Dems? That's what you think?

joeyjojojunior says

Are you actually saying that pollsters whose job and livelihood depends on accurate results purposely made such an obvious mistake as oversampling Dems? That's what you think?

CNN depends on accurate polling results for its livelihood?

CNN doesn't equal pollsters.

The network doesn't rely on it, but the employees who CNN pays to get polling data certainly do.

joeyjojojunior says

CNN doesn't equal pollsters.

The network doesn't rely on it, but the employees who CNN pays to get polling data certainly do.

Yeah that makes a lot of sense, the pollsters who work for CNN wouldn't dare manipulate data even if it didn't agree with the narrative CNN was trying to push. CNN being a network of the utmost trustworthiness would always expect perfectly sampled, non-loaded, completely unbiased polling data.

Ah, the CNN is biased narrative. So, I'm assuming the CNN polls were wildly different than other polling firms then, right? Since they were trying to push a specific agenda that others weren't?

Oh, wait, the CNN results are similar to everyone else? I guess you're wrong again.

Goran, just ignore joey/tatupu. It's obvious he lives in denial or is just trolling you.

He's like the fat obnoxious kid at the party who thinks he's always right and won't shut the fuck up. Even when he's blatantly wrong.

joshuatrio says

Oh, wait, the CNN results are similar to everyone else?

Yes! Did you notice how 95% of the MSM media got the election completely and utterly wrong? It proves my point. Nate Silver's model was compromised by loaded, tainted data.