General Discussion
Related: Editorials & Other Articles, Issue Forums, Alliance Forums, Region ForumsWhat I don't understand why people have been worshiping Nate Silver all election season.
Nate Silver correctly predicted the outcome of the election in most states. That is good.
All election season people have been touting his predictions as good news. Good news is nice I suppose.
My problem with all this:
Nobody can predict the future. If they could then more people would be rich on Wall Street, sports betting et al. But they are not. There is a host of evidence that randomly guessing which stocks to invest in do better in the medium to long run than picking stocks on other criteria. Look up "A Random Walk Down Wall Street".
When people say they have a "system" and then say that they have made money with the system (investments, gambling) you do not hear from all the folks who also have a system that did not work. (Unless they trade on insider information or can count cards and all that).
Nate Silver has a system. But he won't tell you what it is. He creates these seemingly scientific numbers (President Obama has a 90% chance to win) but we have no idea what the system is. He says it is proprietary.
Silver pulled this same crap at Baseball Prospectus. After he left they scrapped his "system" because it was unworkable without making changes for individual players who defied his system.
What we don't know is whether or not Silver is lucky or good.
randome
(34,845 posts)I only hope he doesn't give in to temptation and see himself as a 'Democratic' pollster because we need unbiased views, not good news.
My hunch is that he will not do that.
Surya Gayatri
(15,445 posts)AngryAmish
(25,704 posts)Do you really want someone just to confirm your prejudices? Listen to MSNBC for that.
WinkyDink
(51,311 posts)tledford
(917 posts)AngryAmish
(25,704 posts)CJCRANE
(18,184 posts)there are several statisticians who did the same thing with fairly accurate results.
WinkyDink
(51,311 posts)take him seriously.
I WASN'T SURPRISED AT ALL BY THE ELECTORAL NUMBERS. WHY WERE THEY?
AngryAmish
(25,704 posts)Do you suppose, from their point of view, it would be good for them to say that they had no reasonable chance to win?
This is the problem: (and it may only be a problem because I have a bug in my ass about predictions and I'm stupid that way)
No one can predict the future. I wish people would stop acting like they can. For example, remember when the Administration said that they would have unemployment down in the 5% area if the stimulus went through? How in the hell are they supposed to know that? It is pure arrogance to say that they can make any sort of reasonable prediction about the future of the economy. A host of things can happen, good or bad, that can change the US economy, which is mind-bendingly complex.
I'm not bashing the administration - every politician does this. I just think it insults my intelligence when people say they can predict things I know they cannot. So I have a problem with Silver and his pseudo-scientific predictions.
ipfilter
(1,287 posts)but rather probabilities. There's a big distinction there. I can't predict with any accuracy that I will drive home this evening without getting into a car accident but statistically it is very probable that I won't.
upaloopa
(11,417 posts)If you really have a random sample, what is true about the sample is true about the whole within a margin of error. The larger the sample the lower the margin of error.
jenw2
(374 posts)Just look at the "heros" you see talked about on CNN and other "news" channels.
WinkyDink
(51,311 posts)OldDem2012
(3,526 posts)gravity
(4,157 posts)Individual polls are too noisy and the media has the tendency to cherry pick data to fit their narrative.
Silver just provides a way to filter out the noise and understand what is really going on in the polls. It isn't perfect, but it is much better than anything else in the media.
Vinnie From Indy
(10,820 posts)The OP's first paragraph:
"Nate Silver correctly predicted the outcome of the election in most states. That is good."
The OP's last paragraph:
"What we don't know is whether or not Silver is lucky or good."
AngryAmish
(25,704 posts)Nate Silver correctly predicted the outcome of the election in most states. That is a good thing.
then
What we don't know is how he did it, therefore he is either lucky or good at predictions.
Buddyblazon
(3,014 posts)CJCRANE
(18,184 posts)I'm not a statistician but it seems obvious that they aggregrate publicly available opinion polls.
Obamamama44
(98 posts)I relied in him because it was math. When all the pundits were calling it close and Romney up...it was spin..Nate was doing math. The end.
AngryAmish
(25,704 posts)Obamamama44
(98 posts)To show how he got his average. He did say I went by all the polls. If all the polls were wrong I would be wrong. But he said it was close mid October and started spreading from there. It will not go our way everytime but I feel like its just giving us the snapshot mathematically of our odds...it's not like we got 94 percent of tr vote we had 94 percent greater chance to win based on the polling and averages.
AngryAmish
(25,704 posts)He didn't show his work.
Obamamama44
(98 posts)To make sure you know how he got there...he was right...why question someone who got it right??
LondonReign2
(5,213 posts)Every single corporation with their intellectual capital.
That's their value -- Coke's secret recipe, Intel's chip designs, etc etc.
Why should Nate give up his intellectual capital for free? Hell, he even admits it isn't all that sophisticated, but there is no reason he should give away his work product for free anymore than Porsche should turn over their design specifications.
Lucinda
(31,170 posts)nadinbrzezinski
(154,021 posts)JoePhilly
(27,787 posts)Also, the team at Princeton was just as accurate.
Its called statistics ... and good statistics, good predictive statistics, are hard.
And you make a common mistake ... you are concerned about a single data point which is "unworkable" without making changes for "individual players who defied his system". That entire statement demonstrates a misunderstanding about statistics.
Statistics are about MANY data points ... not just ONE data point. For every data point that you find which does not "fit" the statistical model, there will be 10 to 20 data points that fit it perfectly ... statistics are about many data points. They predict the behavior of groups of people. And so the predictions may fit 10 or 20 specific people, and then not fit one or two.
Big deal. The model creates a high probability of making a good choice. And yes, occasionally, there will be some data point that does not fit.
Here is an interesting chart from Slate ... play with it. Its shows the various predictions by pundits. Nate and some others used statistics, others used their "experience" ... see how well they did.
http://www.slate.com/articles/news_and_politics/politics/2012/11/pundit_scorecard_checking_pundits_predictions_against_the_actual_results.html
AngryAmish
(25,704 posts)I have some education in statistics. Sports-wise I'm definitely in the Sabermetricians camp. I don't even have a problem with the way Silver made his predictions in theory - poll of polls and all that. But we don't know how he did it because he says it is a secret.
In the scheme of things this does not matter. But it irks me.
Obamamama44
(98 posts)Don't have to know how, he gets paid to do it now, so why would he share his secret weapon?
AngryAmish
(25,704 posts)Unless he shows his secret weapon we don't know if he was lucky or good.
I listen to more sports radio than I should. Every so often there are commercials for sports betting hotlines. They say they will give you the Monday Night Football winner against the spread for free. So lets say that 100,000 folks call up for the winning bet.
50,000 get Team A. 50,000 get Team B.
Team B wins.
People who got Team B are happy and think that this tout has inside information, a system etc. People who got Team A are sad and think the tout is a jerk.
So 20,000 people who bet Team B call the hotline the next week for the winner. But the prediction costs $20 this time. The tout has converted a phone and a commercial into $400,000
Repeat the process.
If we don't know Silver's algorithm, then we don't know if he is doing it with luck so far or with good methodology.
Obamamama44
(98 posts)Sickly accurate..why so irritated!? I'd be irritated if I followed one that was wrong but not one that got it right...he's a numbers guy...
JoePhilly
(27,787 posts)I also have a Stats background. And the way you do stats has to do with determining the probability that your prediction is wrong.
Generally, you design the methodology such that your theory will FAIL unless the probability of agreement is better than 95%, or even more. 99% is not uncommon.
In advanced statistics, you do this by evaluating the probability of many smaller probabilities. If you have ranges of probabilities of various outcomes, you can run repeated models against those smaller probabilities. And from that you can calculate the probabilities of various outcomes.
This is what Nate Silver and the Princeton folks did. They took the polls, and then used those to build probability models, by state and even by county in some cases. The poll ranges vary. And so, if you run simulations that draw from those ranges, you get different predictions on the outcome. But, if you run those simulations 100 times, and Obama wins 99 times, you get an idea of reality.
And this is what Nate and the other statisticians did. They built models, based on actual polls, and then ran simulations to test the outcomes.
Meanwhile, Rove and his gut were on TV screaming foul, and we are all laughing at him now.
SidDithers
(44,228 posts)Correction. Nate Silver correctly predicted the outcome of all states.
Sid
Obamamama44
(98 posts)It was close, as we see how long it took them to count, but it was blue in the end!
CJCRANE
(18,184 posts)WilliamPitt
(58,179 posts)and was four tenths of a percentage off in the popular vote this year.
"Yeah, yeah, I know about the results, but why should I trust him?"
Cynic, heal thyself.
Jennicut
(25,415 posts)It is probability based on averages.
Which is why it was hilarious that the Repubs doubted him.
11 Bravo
(23,926 posts)Jim__
(14,077 posts)As to your claim:
He gives a pretty good overview of his methodology: here ( http://fivethirtyeight.blogs.nytimes.com/methodology/ ). That's not every thing; but it is better than no idea what the system is.
FightForMichigan
(232 posts)CJCRANE
(18,184 posts)Never a miscommunication. No one can explain that.
LondonReign2
(5,213 posts)we won't know if they are magnetic or just really sticky
FightForMichigan
(232 posts)liberal_at_heart
(12,081 posts)it's just nice to know someone is using actual real life facts and numbers and not manipulating them to say what they want them to say. I say good for him.
truebrit71
(20,805 posts)arthritisR_US
(7,288 posts)gollygee
(22,336 posts)joeybee12
(56,177 posts)Logical
(22,457 posts)Hamlette
(15,412 posts)he has also written a book on why some predictions are right and some are wrong. And your allegations about his baseball model is not supported by any facts I can find or have read about.
What I don't get is what is the problem around here about Nate. He didn't "correctly predicted the outcome of the election in most states" he predicted the outcome in ALL of the states. In 2008 he got all but Indiana right. He also called the 2010 house races correctly.
And Gallup? What was once considered to be the gold standard? Not so much.
Nate considers everything and discounts or weighs more heavily sources depending on how those polls/sources have done historically. He also considers other "stuff". One day a month or so ago he had a post about why he had increased Obama's odd based on the consumer confidence level going up. Apparently consumer confidence has an influence on politics (duh) so he factors it in.
He posts nearly every day and his posts are fillled with information about his model. Check him out next time. Until someone else predicts results better than he does, I'll read him.
he's not lucky, IMHO, he's good.
Sophiegirl
(2,338 posts)Can I hire Nate to predict the next massive jackpot lotto numbers? I swear, I'll share half of it with him. Maybe more than half!! I'm not greedy....and I'll even pay taxes on the windfall!!!