If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(Talking Points Memo)   Nate Silver's 7 most memorable predictions. Christ, he's just going to ESPN, not dead   (tpmdc.talkingpointsmemo.com) divider line 47
    More: Interesting, West Region, Football League Championship, number crunching, predictions, Wichita State, Senate Majority Leader Harry Reid, Majority Leader Harry Reid  
•       •       •

5140 clicks; posted to Main » on 23 Jul 2013 at 10:36 AM (38 weeks ago)   |  Favorite    |   share:  Share on Twitter share via Email Share on Facebook   more»



47 Comments   (+0 »)
   
View Voting Results: Smartest and Funniest
 
2013-07-23 10:27:54 AM
His most impression writing is still has writing about the 2008 primaries. Everything after that was actually pretty easy to predict.
 
2013-07-23 10:37:49 AM
No, *we're* dead. We're left with morons covering elections. "The vibrations are right".  SHOOT ME NOW.
 
2013-07-23 10:41:15 AM
So now he'll be able to tell us the winner in every game at the half?
 
2013-07-23 10:45:06 AM
Oh come on, you think he won't be pulling double duty at least for the presidential elections?
 
2013-07-23 10:47:02 AM
If he is behind their pay wall he is dead as far as I'm concerned.
 
2013-07-23 10:47:27 AM
And according to that list, he's not really that good at predictions.
 
2013-07-23 10:47:50 AM
I've got nothing against Nate Silver - he's a smart guy and he has made a name for himself by applying sound statistical analysis to broader culture.

But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)

Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.
 
2013-07-23 10:48:45 AM
The fact that professional football includes players does not preclude cheer leaders. Pundits still have a place to make the boring, mechanical process of executing a strategy entertaining to the willfully ignorant masses.
 
2013-07-23 10:49:39 AM
Why am I supposed to care about this guy?
 
2013-07-23 10:49:55 AM
Statisticians:
4.bp.blogspot.com
(and anyone in applied math really)
 
2013-07-23 10:50:23 AM

secularsage: Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.


This has already happened quite a few times. His concern shouldn't be getting predictions wrong - his only concern should be calling the Presidential election wrong. That's the only thing people will care about.
 
2013-07-23 10:50:32 AM

secularsage: I've got nothing against Nate Silver - he's a smart guy and he has made a name for himself by applying sound statistical analysis to broader culture.

But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)

Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.


What "people" do doesn't concern me.  If he posts something asserting a 60% confidence interval, and something in the 40% zone happens, I won't want him hung by his toes.  It'll still be better than the narrative versions we get from the media as a whole.
 
2013-07-23 10:50:40 AM
I thought Nate Silver was Ron Swanson's smooth alter-ego.
 
2013-07-23 10:51:59 AM
Prediction: Obama, the consistent favorite, will win in an electoral college landslide.

Outcome: Obama held consistent leads in the polls en route to an electoral college landslide.


I thought it was a horse-race the entire time that anybody could win?
 
2013-07-23 10:54:28 AM

abhorrent1: And according to that list, he's not really that good at predictions.


He doesn't do predictions. He forecasts probabilities.

secularsage: But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)


Yep.
 
2013-07-23 10:54:47 AM
Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.
 
2013-07-23 10:55:54 AM

impaler: Prediction: Obama, the consistent favorite, will win in an electoral college landslide.

Outcome: Obama held consistent leads in the polls en route to an electoral college landslide.

I thought it was a horse-race the entire time that anybody could win?


CNN finally became dead to me last summer with their headline stating "Obama and Romney statistically tied!" when Obama was up 3 with margin of error 3.  You could say they are tied, but you could equally claim Obama was up 6 at the time.  The most correct interpretation was "Obama up three points".  farking shysters.
 
2013-07-23 10:56:02 AM

secularsage: Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.


Funny thing is he doesn't like that, and the book I posted above talks about why not.
 
2013-07-23 10:56:20 AM
Prediction: Notre Dame will beat the spread against Alabama in the 2013 BCS National Championship Game.
 
2013-07-23 10:58:30 AM

abhorrent1: And according to that list, he's not really that good at sports predictions.


Good thing he doesn't have to focus on that at... where is he going again?
 
2013-07-23 11:00:13 AM
That snapshot of his correct and failed forecasts just shows what anyone should know. Sports are far too unpredictable to wager on. He's gonna be wasted at ESPN, which is already a toxic pit like most sports "journalism," because they spend all their time arguing about what's GOING to happen, instead of reporting on what's happening/has happened.

/I've bet on exactly one college football game in my life, and won, so you should listen to me.
 
2013-07-23 11:08:19 AM

Carn: Oh come on, you think he won't be pulling double duty at least for the presidential elections?


He's taking FiveThirtyEight back from NYT so he'll be doing election coverage on his own site.
 
2013-07-23 11:09:11 AM

Glenford: Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.


If Silver forecasts 10 games, with one team having a 60% chance to win, and they lose 4 of those 10, people will point to those 4 losses as a failure.

To a statistician, if they won all 10, that would be a failure.

I give a 0% chance on getting the general population to ever understand that concept.

/Now excuse me, I have to play on a slot machine someone abandoned after playing 100 times without a win. It's pretty much a sure thing.
 
2013-07-23 11:09:46 AM

odinsposse: Carn: Oh come on, you think he won't be pulling double duty at least for the presidential elections?

He's taking FiveThirtyEight back from NYT so he'll be doing election coverage on his own site.


Is that right? I assumed it would be hosted at Grantland.
 
2013-07-23 11:13:18 AM

secularsage: I've got nothing against Nate Silver - he's a smart guy and he has made a name for himself by applying sound statistical analysis to broader culture.

But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)

Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.


They kind of missed one of the major points in his Congressional predictions. Silver picked 54-55 seats at a time when that was one of the highest predictions. Yet he was "wrong". That's like someone predicting the Cleveland will score 54-55 points against Baltimore, and then having people tell you you were wrong because they scored 63.

But, yes, I wish they would concentrate on the fact that science is what's winning here, and that Silver is just really good at science.
 
2013-07-23 11:15:42 AM

impaler: Prediction: Obama, the consistent favorite, will win in an electoral college landslide.

Outcome: Obama held consistent leads in the polls en route to an electoral college landslide.

I thought it was a horse-race the entire time that anybody could win?


anyone named Barack Obama.
 
2013-07-23 11:16:31 AM

Glenford: Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.


That's really the problem - people don't seem to understand the difference between a prediction and a probability. Saying there's a 5% chance Sanchez will lead the Jets to the Superbowl this year doesn't mean you're wrong if it happens. It WILL mean the apocalypse is upon us.
 
2013-07-23 11:17:07 AM

ManateeGag: impaler: Prediction: Obama, the consistent favorite, will win in an electoral college landslide.

Outcome: Obama held consistent leads in the polls en route to an electoral college landslide.

I thought it was a horse-race the entire time that anybody could win?

anyone named Barack Obama.


You need some serious unskewin' son.
 
2013-07-23 11:21:52 AM

impaler: Glenford: Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.

If Silver forecasts 10 games, with one team having a 60% chance to win, and they lose 4 of those 10, people will point to those 4 losses as a failure.

To a statistician, if they won all 10, that would be a failure.

I give a 0% chance on getting the general population to ever understand that concept.

/Now excuse me, I have to play on a slot machine someone abandoned after playing 100 times without a win. It's pretty much a sure thing.


What's your confidence interval on this?
 
2013-07-23 11:24:25 AM

DamnYankees: odinsposse: Carn: Oh come on, you think he won't be pulling double duty at least for the presidential elections?

He's taking FiveThirtyEight back from NYT so he'll be doing election coverage on his own site.

Is that right? I assumed it would be hosted at Grantland.


Nope. It will be it's own thing.
 
2013-07-23 11:26:39 AM

vygramul: Glenford: Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.

That's really the problem - people don't seem to understand the difference between a prediction and a probability. Saying there's a 5% chance Sanchez will lead the Jets to the Superbowl this year doesn't mean you're wrong if it happens. It WILL mean the apocalypse is upon us.


I always use his weather example from The Signal and the Noise when trying to explaining it. If there's a 90% chance of rain that means that historically when there has been the  exact same atmospheric conditions, it has rained on 90% of those days.

/paraphrasing, the book is at home
 
2013-07-23 11:28:22 AM

Glenford: What's your confidence interval on this?


+- 0% with 100% confidence.

/is my contempt for the mathematically illiterate plebes showing? I'm kind of an ass when it comes to math.
 
2013-07-23 11:33:19 AM

secularsage: I've got nothing against Nate Silver - he's a smart guy and he has made a name for himself by applying sound statistical analysis to broader culture.

But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)

Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.



Nate does not not not make predictions. He analyzes the data and gives the probablity. Take Indiana in 2008. He didn't "predict" it for McCain per se. He said that McCain had a 52% or whatever chance of carrying Indiana. And it went to Obama against small odds. These things happen in human events. But Nate's analysis was flawless in all the other states, including the improbable carrying of Omaha in Nebraska's not-winner-take all.
 
2013-07-23 11:33:42 AM
It is much much easier to analyze the probability of political outcomes than it is to predict anything in sports. They are two completely different beasts.
 
2013-07-23 11:36:32 AM
Picking the presidential winner is more analogous to picking who Americas favorite NFL team is (and you get the whole of the Internet to do your research).
 
2013-07-23 11:39:36 AM

impaler: Glenford: What's your confidence interval on this?

+- 0% with 100% confidence.

/is my contempt for the mathematically illiterate innumerate plebes showing? I'm kind of an ass when it comes to math.


If you're going to mock people for being unfamiliar with numeracy, you should probably have a decent handle on literacy.
 
2013-07-23 11:39:38 AM

ikanreed: secularsage: I've got nothing against Nate Silver - he's a smart guy and he has made a name for himself by applying sound statistical analysis to broader culture.

But I hate it when articles are written about him like he's some sort of prophet, sage or wizard. He's using the same tools that researchers in all social sciences use, and he's not doing anything that's amazing or remarkable - he's just using predictive analytics to conduct meta-analyses of existing data. (Silver himself has said as much; he's enjoying the ride, but doesn't yet seem to believe the hype surrounding him.)

Unfortunately, Silver is also being set up for a big fall at some point, because his analyses are based upon predictable circumstances and probabilistic outcomes. He's going to make a call at some point based on probability and the longshot alternative is going to occur instead. Even though Silver will have said nothing wrong (he's just preaching probabilities) people will turn on him as a false prophet.

What "people" do doesn't concern me.  If he posts something asserting a 60% confidence interval, and something in the 40% zone happens, I won't want him hung by his toes.  It'll still be better than the narrative versions we get from the media as a whole.


No kidding.  The article mentions "predicted all 50 states" in 2013, but fails to note that he basically predicted Florida having a 51% chance for Obama or less.  Not only that, his prediction for the split were so tight that if he actually had a pro-Obama bias (he underestimated Obama's win) and his error flipped, I think the only state he would have gotten wrong was Florida, and Obama would still have a comfortable electoral lead.
 
2013-07-23 11:49:21 AM

Glenford: impaler: Glenford: What's your confidence interval on this?

+- 0% with 100% confidence.

/is my contempt for the mathematically illiterate innumerate plebes showing? I'm kind of an ass when it comes to math.

If you're going to mock people for being unfamiliar with numeracy, you should probably have a decent handle on literacy.


Uh.

il·lit·er·ate  [ih-lit-er-it]adjective1.unable to read and write:an i lliterate group.2.having or demonstrating very little or no education. 3.showing lack of,  especially in and literature.4.displaying a marked lack of knowledge  in a particular field:He is musically illiterate.
 
2013-07-23 11:49:51 AM

Glenford: If you're going to mock people for being unfamiliar with numeracy, you should probably have a decent handle on literacy.


'Mathematically illiterate' is a cromulent phrase.
 
2013-07-23 11:55:39 AM

Carn: Glenford: impaler: Glenford: What's your confidence interval on this?

+- 0% with 100% confidence.

/is my contempt for the mathematically illiterate innumerate plebes showing? I'm kind of an ass when it comes to math.

If you're going to mock people for being unfamiliar with numeracy, you should probably have a decent handle on literacy.

Uh.

il·lit·er·ate  [ih-lit-er-it]adjective1.unable to read and write:an i lliterate group.2.having or demonstrating very little or no education. 3.showing lack of,  especially in and literature.4.displaying a marked lack of knowledge  in a particular field:He is musically illiterate.



in·nu·mer·ate: Without a basic knowledge of mathematics and arithmetic

A less redundant and more precise phrase.
 
2013-07-23 11:57:56 AM

Glenford: A less redundant and more precise phrase.


It's more precise, but no less redundant.
 
2013-07-23 12:01:28 PM

Glenford: vygramul: Glenford: Nate Silver doesn't make predictions - he calculates the probability of an event happening. Even if Obama had lost the election, Silver would still have been correct - there was a 10% chance that Romney would win.

That's really the problem - people don't seem to understand the difference between a prediction and a probability. Saying there's a 5% chance Sanchez will lead the Jets to the Superbowl this year doesn't mean you're wrong if it happens. It WILL mean the apocalypse is upon us.

I always use his weather example from The Signal and the Noise when trying to explaining it. If there's a 90% chance of rain that means that historically when there has been the  exact same atmospheric conditions, it has rained on 90% of those days.

/paraphrasing, the book is at home


I thought that the percentage of precipitation was the actual area expecting rain. So 90% chance of rain would mean that 90% of a given area would receive some non-zero amount of precipitation. Not sure where I heard that though, so grain of salt and all that.
 
2013-07-23 12:10:50 PM

Carn: CNN finally became dead to me last summer with their headline stating "Obama and Romney statistically tied!" when Obama was up 3 with margin of error 3.  You could say they are tied, but you could equally claim Obama was up 6 at the time.  The most correct interpretation was "Obama up three points".


Sorry, I'm going with CNN on that one.  If the results are at or within the margin of error, it's more responsible to characterize it as too close to call than to assert one candidate definitively has a lead.

Of course, the best course of action would have been to report no interpretation of the numbers at all, as they were too inconclusive to actually mean anything.  But CNN has a lot of air time, and Twitter polls can only fill so much of it.
 
2013-07-23 12:41:08 PM
He's dead to me.

I know I'm not the first to say this.

His book looks interesting but where am I going to find time or room for another book at the moment?
 
2013-07-23 12:48:12 PM

poot_rootbeer: Carn: CNN finally became dead to me last summer with their headline stating "Obama and Romney statistically tied!" when Obama was up 3 with margin of error 3.  You could say they are tied, but you could equally claim Obama was up 6 at the time.  The most correct interpretation was "Obama up three points".

Sorry, I'm going with CNN on that one.  If the results are at or within the margin of error, it's more responsible to characterize it as too close to call than to assert one candidate definitively has a lead.

Of course, the best course of action would have been to report no interpretation of the numbers at all, as they were too inconclusive to actually mean anything.  But CNN has a lot of air time, and Twitter polls can only fill so much of it.


Why shouldn't they have reported Obama +6 then?  A true dead heat would have been 0 +- whatever.  I'm sticking with my hatred.  They chose the interpretation which gave them the headline they wanted "Dead heat!!!", which was misleading, without explaining that that was only one possible interpretation.  It also proved to be incorrect.

/end threadjack
 
2013-07-23 03:29:36 PM

poot_rootbeer: Carn: CNN finally became dead to me last summer with their headline stating "Obama and Romney statistically tied!" when Obama was up 3 with margin of error 3.  You could say they are tied, but you could equally claim Obama was up 6 at the time.  The most correct interpretation was "Obama up three points".

Sorry, I'm going with CNN on that one.  If the results are at or within the margin of error, it's more responsible to characterize it as too close to call than to assert one candidate definitively has a lead.


The "margin of error" boundary is not a binary one. If one is leading .01% above the margin of error, there is a 95% chance they will win, if they are leading .01% below the margin of error, there is a 95% chance they will win (assuming a 95% confidence interval for said margin of error).
 
2013-07-23 06:57:09 PM

impaler: poot_rootbeer: Carn: CNN finally became dead to me last summer with their headline stating "Obama and Romney statistically tied!" when Obama was up 3 with margin of error 3.  You could say they are tied, but you could equally claim Obama was up 6 at the time.  The most correct interpretation was "Obama up three points".

Sorry, I'm going with CNN on that one.  If the results are at or within the margin of error, it's more responsible to characterize it as too close to call than to assert one candidate definitively has a lead.

The "margin of error" boundary is not a binary one. If one is leading .01% above the margin of error, there is a 95% chance they will win, if they are leading .01% below the margin of error, there is a 95% chance they will win (assuming a 95% confidence interval for said margin of error).


Yeah, you can't get overly wedded to significance levels. The best way to report it is probably that the poll shows a ~95% chance that Obama has some lead, most likely about 3%. People have this weird idea that a confidence level of 95.1% is incontrovertible proof, while a confidence level of 94.9% is utterly irrelevant. Statistics is all about getting away from inflexible yes/no predictions, so your interpretation of statistics shouldn't be inflexible either.
 
Displayed 47 of 47 comments

View Voting Results: Smartest and Funniest


This thread is closed to new comments.

Continue Farking
Submit a Link »






Report