NYT Nate Silver Doubles Down: 75% Chance Obama Wins

Obviously you haven't seen the REAL polling.

 

Actually, he's a statistical analyst.

Do you believe in stats? Or are you more like the state of Texas? I recently took defensive driving here in Texas, and this was the first thing it told me:


The next time that you get behind the wheel, consider what the odds
are that you will be involved in a collision.

This may surprise you, but we consider your odds to be exactly 50/50.

Either you will be involved in a collision or you won't.

There really is no other possibility!

 

What? I've been saying for quite a while now, thanks to the polls shown and how all other polls are bogus, that Romney will win 50 states and 100% of the popular vote.

Of course you can look at Nate Silver's polls all you want.

 

No. Nate Silver is a statistician that developed a good predictive model for baseball players. He then wanted to take it to the next level of popularity and financial reward for himself by using his predictive indicator to predict elections. But it won't work there unless he is able to get data going into his model that is accurate. Baseball data is tons more accurate than political polling data. Too many unknowns in political polling data and too many biased opinions in political polling data.

Nate is simply trying to maximize earnings using his model and trying to help his candidate at the same time. He's a smart guy. But it's not going to work. And he's going to have to come off Barack in a hurry now to save face. Watch it happen.

 

Ummm, polling data has a very long history of being very accurate and reliable. Just because you don't believe in reality doesn't change that fact.

In fact, all your nonsense about poll being skewed based on sampling is based on polling data - it's just that you're choose to believe the ones you want to believe to create the story you want to believe in your head.

 

See how it works, either you agree with Silver, or you are anti-math/science.

IMO Silver's model relies too much on state polls, which are all using 2008 turnout models when all signs point to a 2004 type electorate.

A random simulation of unweighted sate poll averages with an excel macro gets you the same results as Silver, and you don't need the title "statistical analyst" to do it.

 

What % chance did he give the Rockets to land Dwight Howard?

 

So, what we're all saying then is that there's still a 50/50 chance we'll end up with Dwight Howard next week? Or is it more of a 75% chance at this point?

 

I am choosing to believe what I feel is the truth. You can make claims about polling data being very accurate and reliable but that's hogwash. You have to understand the country, the current time, the degree of unhappiness amongst voters, and the level of excitement/buzz that the candidates are generating. And then you have to look at polls that are properly weighted for that excitement and buzz or you have no chance of being accurate. For a person not to look behind the polls into political bias is foolishness. You keep staring at polls. I'm calling it. Romney is your next president. Obama is done.

 

No - if you have a legitimate reason (as you argue) for explaining why you think the model doesn't work, that's a fair game for discussion. But the nonsense logic people use here (all the polls are wrong! or you can't predict elections!) is what makes people anti-math/science.

No one has yet to explain WHY polls are skewed this time around. Pollsters report their D/R/I samples as-is - they don't adjust them for anything. So why exactly are random calls this polling cycle all going to Democrats, when there is no history at all of that ever happening before? If you want to argue something is skewed, explain why the science of polling has failed.

What signs are those?

And yet, that's not what Silver does. It's not terribly surprising that everyone's methods from RCP averages to gambling sites to Intrade to Silver to your excel macro all come to the same basic conclusions about the race. That tends to happen when the polling data is fairly consistent.

 

Pew/Gallup/Rasmussen partisan ID surveys

Point being Silver's model offers no more insight than a random simulation of polling averages.

 

You're confusing result with process.

He has a process which you aren't able to seriously dispute.

 

So why are those pollls accurate and all the others are wrong?

Actually, it provides a lot more insight. The bottom line results are only a small portion of what the model generates. It provides measures of likelihood of winning a state; it provides tipping point information; it provides spending efficiency information; it accounts for durability of a lead and likelihood of holding a lead; and on and on. Just because you can't see something doesn't mean it doesn't exist.

 

Well, he can claim he had it at a 0% chance since he wasn't involved.

 

Global warming FTW!

 

He does provide these as well. He has a projected vote share for each state as well as a probability for winning each state. That factors into the projection of of the number of electoral votes for each candidate.

He predicts the national popular vote too, but how exactly is that more relevant when it doesn't factor into who wins the election?

 

Pretty sure had he been involved, it wouldn't have been 100%!

Howard, Dwight.

 

MojoMan's last dying primal scream before Mr Obama wins next week.

Your tears are delicious.