- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
How Nate Silver and 538 Came to Be So Popular
Posted on 10/19/20 at 3:09 pm
Posted on 10/19/20 at 3:09 pm
Nate Silver is a data geek. He's very good at interpreting data. However, what he is doing now (and in 2016) is interpreting flawed data, and I'd argue making it more flawed. But why?
Nate started working for ESPN. He was excellent while working there. Want to know the chances of victory when your team is up 28-10 in the 2nd quarter? His analytics still live at ESPN today, and while there are of course improbable comebacks and crazy things that happen in sports, these analytics are pretty spot on - to the point that Vegas changed their systems for in game betting odds.
Right before the 2008 election, ESPN was at the beginning of their move into politics, and asked their new found advanced stats guy to make some political predictions. His system nailed it. He missed one state - Indiana, predicting that McCain would take Indiana but Obama took it in a historic victory.
The New York Times took notice and Nate moved from sports to politics. His 2012 forecast was perfect, and in 48 of the 50 states, he was in the margin of error.
So, Silver and 538 are a new star. 99 out of 50 states in the '08 and '12 elections - his word was golden. ESPN lured him back from the New York Times and doubled his staff. In an ill-advised attempt to grow the model, he tried to predict the Oscars and it was terrible.
Then came the 2016 election and the debut of the Polls-Plus model that he still uses. Not only did Donald Trump change the accuracy of the polls, Silver's additions made 538 even more inaccurate.
Silver gave Trump less than a 2% chance of winning the GOP nomination. He even said in an interview that 2% was too much and he was embarrassed.
When Trump won the nomination, his model predicted Clinton would have anywhere from an 85-99% chance of winning. We know what happened in 2016.
The same thing is going on this year with Nate. He is so confident in his model that he has doubled down on it. When he puts his controls in, it gives Biden Georgia and Texas. Rather than taking a step back and realizing something is wrong, he is confident in his model and puts out a ridiculous map like he did.
Right now, his model is saying a 12% chance of a Trump victory.
His model says the GOP has a 26% chance of keeping the Senate.
His model says the GOP has a 5% chance of keeping the House.
That equates to a 3.1% chance of keeping the White House and the Senate and a 0.2% chance of the GOP having all three. There is no way a real person can believe those odds. His model is flawed. He knows it but he knows that if he changes it and it gives Trump a decent shot he will lose his gravy train.
I am on the outside and don't have access to everything he does, but I feel like he is going to get embarrassed this year. I wish that in the name of science, he would drop his "Plus" and go back to his roots. It would probably be in the 55-45 range for Biden, but that's too much for his liberal masters.
Nate started working for ESPN. He was excellent while working there. Want to know the chances of victory when your team is up 28-10 in the 2nd quarter? His analytics still live at ESPN today, and while there are of course improbable comebacks and crazy things that happen in sports, these analytics are pretty spot on - to the point that Vegas changed their systems for in game betting odds.
Right before the 2008 election, ESPN was at the beginning of their move into politics, and asked their new found advanced stats guy to make some political predictions. His system nailed it. He missed one state - Indiana, predicting that McCain would take Indiana but Obama took it in a historic victory.
The New York Times took notice and Nate moved from sports to politics. His 2012 forecast was perfect, and in 48 of the 50 states, he was in the margin of error.
So, Silver and 538 are a new star. 99 out of 50 states in the '08 and '12 elections - his word was golden. ESPN lured him back from the New York Times and doubled his staff. In an ill-advised attempt to grow the model, he tried to predict the Oscars and it was terrible.
Then came the 2016 election and the debut of the Polls-Plus model that he still uses. Not only did Donald Trump change the accuracy of the polls, Silver's additions made 538 even more inaccurate.
Silver gave Trump less than a 2% chance of winning the GOP nomination. He even said in an interview that 2% was too much and he was embarrassed.
When Trump won the nomination, his model predicted Clinton would have anywhere from an 85-99% chance of winning. We know what happened in 2016.
The same thing is going on this year with Nate. He is so confident in his model that he has doubled down on it. When he puts his controls in, it gives Biden Georgia and Texas. Rather than taking a step back and realizing something is wrong, he is confident in his model and puts out a ridiculous map like he did.
Right now, his model is saying a 12% chance of a Trump victory.
His model says the GOP has a 26% chance of keeping the Senate.
His model says the GOP has a 5% chance of keeping the House.
That equates to a 3.1% chance of keeping the White House and the Senate and a 0.2% chance of the GOP having all three. There is no way a real person can believe those odds. His model is flawed. He knows it but he knows that if he changes it and it gives Trump a decent shot he will lose his gravy train.
I am on the outside and don't have access to everything he does, but I feel like he is going to get embarrassed this year. I wish that in the name of science, he would drop his "Plus" and go back to his roots. It would probably be in the 55-45 range for Biden, but that's too much for his liberal masters.
This post was edited on 10/19/20 at 3:16 pm
Posted on 10/19/20 at 3:16 pm to anc
People here at TigerDroppings laugh at Nate Silver, but out of many flawed pollsters and analysts in 2016, he was arguably the least-flawed. Although he gave Hillary a 71 percent chance of winning, that was already far more Trump-friendly than almost anyone else. Most other networks/sources were predicting Hillary with a 90-99% chance of winning. All the way up until Election Night there were still people criticizing Nate for being "too generous" to Trump.
This post was edited on 10/19/20 at 3:30 pm
Posted on 10/19/20 at 3:18 pm to AnyonebutSteelers
quote:He's not a pollster. And he had a 50-50 shot at picking the winner. You can't win 71% of an election.
People here at TD laugh at Nate, but he was actually one of the most accurate pollsters in 2016.
Posted on 10/19/20 at 3:18 pm to anc
Thanks for a coherent post with content. This post is like a sasquatch or the Loch Ness Monster: Rarely seen in real life. I appreciate the effort. Thumbs up.
Posted on 10/19/20 at 3:19 pm to AnyonebutSteelers
quote:He is not a pollster. He does not know enough to realize it is garbage in, garbage out with the polls he is using. He was wrong in 2016, end of story.
one of the most accurate pollsters in 2016
Posted on 10/19/20 at 3:21 pm to Jake88
He is a data nerd who still refuses to concede that his methods are wrong.
Posted on 10/19/20 at 3:24 pm to anc
quote:
He knows it but he knows that if he changes it and it gives Trump a decent shot he will lose his gravy train.
I am on the outside and don't have access to everything he does, but I feel like he is going to get embarrassed this year. I wish that in the name of science, he would drop his "Plus" and go back to his roots. It would probably be in the 55-45 range for Biden, but that's too much for his liberal masters.
Solid post, but this is where you lose me
That dude has taken an immeasurable amount of shite since 2016, especially considering he gave Trump a better shot than most.
No fricking chance he's deliberately tanking his model to satiate some kind of hidden agenda. The logic of that would be counter-intuitive in and of itself.
Posted on 10/19/20 at 3:25 pm to anc
quote:
How Nate Silver and 538 Came to Be So Popular
I'm surprised that anyone listens to him given how wrong he was in 2016.
Posted on 10/19/20 at 3:26 pm to Taxing Authority
quote:
You can't win 71% of an election.
Sure you can, in a sense. It's like saying LSU has a 71 percent chance to beat Alabama, or that there is a 71 percent chance of rain.
When Silver gave Trump a 3-out-of-10 odds of beating Clinton, that was already far more generous to Trump than most other analysts were. Most other analysts gave Trump essentially zero chance.
This post was edited on 10/19/20 at 3:27 pm
Posted on 10/19/20 at 3:46 pm to anc
quote:
When he puts his controls in, it gives Biden Georgia and Texas
His model absolutely does not show Biden winning Texas.
Posted on 10/19/20 at 3:47 pm to AnyonebutSteelers
quote:
When Silver gave Trump a 3-out-of-10 odds of beating Clinton, that was already far more generous to Trump than most other analysts were. Most other analysts gave Trump essentially zero chance.
This. People are predisposed to think that 30% is nothing. I had someone put it to me this way: "If you had 3 bullets in a gun with 10 chambers with it pointed at your head and one shot, how comfortable would you feel with 30%?"
Posted on 10/19/20 at 3:48 pm to PentagonTiger
Or put another way, when an MLB all star gets a base hit are you shocked?
If Trump wins this year he will have egg on his face, but his model have Trump a true chance in a way that no other major model did in 2016.
If Trump wins this year he will have egg on his face, but his model have Trump a true chance in a way that no other major model did in 2016.
Posted on 10/19/20 at 3:55 pm to AnyonebutSteelers
A lot of people also don't understand that 29 percent for Trump doesn't mean it was some impossible insurmountable thing either. It just meant that Hillary had a 2/3-almost a 4/5 shot and Trump had a 1/5 to almost 1/3 shot.
It was an upset, but it was people ignoring how much Hillary was really persona non grata where it mattered that time.
It was an upset, but it was people ignoring how much Hillary was really persona non grata where it mattered that time.
Posted on 10/19/20 at 3:57 pm to anc
quote:
How Nate Silver and 538 Came to Be So Popular
..from Nate Silver to Nate Bronze to Nate Pewter?
He is headed in that direction.
Posted on 10/19/20 at 4:03 pm to ZappBrannigan
quote:
A lot of people also don't understand that 29 percent for Trump doesn't mean it was some impossible insurmountable thing either. It just meant that Hillary had a 2/3-almost a 4/5 shot and Trump had a 1/5 to almost 1/3 shot.
Exactly. If you have three balls in a hat - two blue, one red - and pick one of them out blindfolded, it should be perfectly unsurprising if you get the red ball.
No more surprising than, say, an NFL kicker missing a 50-yard field goal attempt.
Posted on 10/19/20 at 4:25 pm to ValDawgsta
quote:
Or put another way, when an MLB all star gets a base hit are you shocked?
If Trump wins this year he will have egg on his face, but his model have Trump a true chance in a way that no other major model did in 2016.
So if Hillary would have won, would you be saying he was less correct than the people who gave her a 90% chance? Of course not, everybody would be saying he's a genius again.
This is a slight oversimplification, but probabilities can typically only be verified over a large number of iterations, which you can't do with an election. If there was a multiverse and we looked at 1,000 2016 election results and Hillary won 70 times and Trump won 30, Silver's model would be the GOAT. In the real world we have no way of knowing how accurate his probabilities are. As long as an outcome is covered by a non-zero percentage, that outcome is "predicted".
Posted on 10/19/20 at 4:26 pm to AnyonebutSteelers
(no message)
This post was edited on 1/9/21 at 7:25 am
Posted on 10/19/20 at 4:28 pm to AnyonebutSteelers
quote:
Exactly. If you have three balls in a hat - two blue, one red - and pick one of them out blindfolded, it should be perfectly unsurprising if you get the red ball.
That's a deeply flawed analogy to what Silver is doing.
Posted on 10/19/20 at 4:31 pm to Flats
quote:Exactly.
This is a slight oversimplification, but probabilities can typically only be verified over a large number of iterations, which you can't do with an election. If there was a multiverse and we looked at 1,000 2016 election results and Hillary won 70 times and Trump won 30, Silver's model would be the GOAT. In the real world we have no way of knowing how accurate his probabilities are. As long as an outcome is covered by a non-zero percentage, that outcome is "predicted".
Had Silver stated that Trump had a 5% chance of winning in 2016, he would be as right as he was with his 30% chance.
Both have a non-0% chance to Trump, therefore Silver cannot he wrong.
His modeling around a single non-reproducible event cannot he verified.
Same goes for 2020.
He is giving Trump even less of a chance to win this time around.
I think, somewhere around 12-13%.
Guess what.
If Trump wins, Nate Silver will be as right this time as he was last time.
This post was edited on 10/19/20 at 4:34 pm
Back to top
Follow TigerDroppings for LSU Football News