- 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
re: Why Nate Bronze is full of shite: Using his own data
Posted on 10/29/20 at 10:26 pm to mwade91383
Posted on 10/29/20 at 10:26 pm to mwade91383
quote:
That and their body of work (polls) speaks for itself. They’re right way more than they’re wrong
0
Posted on 10/29/20 at 10:27 pm to mwade91383
quote:Studies have been performed in the past that anywhere from 10-12% of Republican and Indy voters will lie, while ~5% of Democrats will lie.
I understand the idea of millions of Trump voters in hiding and “fooling pollsters” but some evidence would be nice.
Posted on 10/29/20 at 10:37 pm to The Boat
Nate bullshite- FIFY
Posted on 10/29/20 at 10:38 pm to Adam Banks
quote:Trafalgar has called 75% of their 48 races correctly. ACROSS all 453 posters and 9559 races in its database, over 80% are called correctly. So it actually calls fewer races correctly.
Explain to me then how he rates Trafalgar C- despite having a higher % races called correctly over some A+ like ny times sienna and less margin of error than some A rated pollsters
More importantly, though, Trafalgar's polling falls outside the margin of error 50% of the time, compared to 29% of all posters and races. So despite calling fewer races, it misses the races outside of random error, for more often.
And if you compare that to the top 100 pollsters, they call 83% of races correctly and only 23% of the races fall outside the MOE.
So Trafalgar's track record is not particularly good, but it's actually gotten worse as 2018 was particularly bad for them as they did fine in 2016.
But I suspect 2020 is going to be a bad year for them. Their polling way off from the overall average, and the overall average is the most accurate predictor, which doesn't bode well for them. More importantly, people have started diving more into their limited crosstabs, which has shown some major issues in their geographic sampling, I analyzed one of their Pennsylvania polls and found that their congressional district sampling, using 2016 margins, would have given Trump a 2-point larger win than he actually had. Plus their age and race sampling favored Trump as well.
Most notably though, there was a Google Drive file with a detailed breakdown of their sample. And they were nonsensical (Trump winning 30% of Ds; Biden winning 25% of Rs; Trump winning 50% of latinos and 30% of blacks). Their lead pollster (Cahaly) was initially defending it on twitter, calling it some ridiculous "50 cent effect," but the file was later deleted and he called it fake. So between the nonsensical crosstabs, and the responses from the pollster, I don't think they really know what they are doing, or worse, know what they are doing and don't care how nonsensical it is.
Posted on 10/29/20 at 10:38 pm to The Boat
Probably time for him to be demoted to
“Nate Tinfoil”!
“Nate Tinfoil”!
Posted on 10/29/20 at 10:42 pm to Scruffy
quote:I'm skeptical of this. Do you have a link for this? In fact, from what I've seen, there is little to no evidence of lying, but at least in 2016, the groups that really went for Trump were less likely to be polled because they are harder to poll and are less likely to respond. That's why pollsters started doing an adjustment for education in 2018. I mean who is going to take the time to respond to a pollster, lie about who they intended to vote for, then actually follow through with the lie in the voting booth?
Studies have been performed in the past that anywhere from 10-12% of Republican and Indy voters will lie
In addition, if this were true then why did the polling underestimate Obama's margins, and why have the polling errors been rather uniform on both sides overall?
This post was edited on 10/29/20 at 10:44 pm
Posted on 10/29/20 at 10:51 pm to buckeye_vol
I guess I’ll ask again.
quote:
Has there ever been a race of national importance when he predicted a conservative politician would win and they ended up losing?
Posted on 10/29/20 at 10:53 pm to Flats
quote:In 2012, the polling underestimated Obama's support pretty significantly. And Nate ended up predicting all 50 states and the exact amount of electoral college votes. Furthermore, only 2 of the 50 races fell outside the MOE (Iowa and WV) and one was wrong in Obama's direction and the other in Romney's.
Has he? If Silver is useful and accurate, then his error should fall on both sides of the reality line, shouldn’t it?
Posted on 10/29/20 at 10:55 pm to Scruffy
Peer reviewed? Credible? Link?
Not a flame, legitimately curious.
Not a flame, legitimately curious.
This post was edited on 10/29/20 at 10:56 pm
Posted on 10/29/20 at 10:57 pm to buckeye_vol
I guess that’s a “no”?
Posted on 10/29/20 at 11:03 pm to buckeye_vol
quote:
And if you compare that to the top 100 pollsters, they call 83% of races correctly and only 23% of the races fall outside the MOE.
You lie.
He rates abc news/wash post as A+. They call only 72% races correctly nbc news/wsj a- 73% I can go on and on and on.
Posted on 10/29/20 at 11:05 pm to Adam Banks
quote:72% and 73% are A+’s in the progressive education system.
They call only 72% races correctly nbc news/wsj a- 73% I can go on and on and on.
Posted on 10/29/20 at 11:05 pm to The Boat
Pretty reasonable IMO.
Posted on 10/29/20 at 11:06 pm to Lima Whiskey
quote:
Dunning-Kruger effect?
He’s trying to intimidate critics (and make himself look clever) with a big word
His whole statement about Dunning-Kruger...exemplifies Dunning-Kruger.
Posted on 10/29/20 at 11:11 pm to Flats
quote:Well he's been doing this since 2008, so only 3 presidential elections. Now he's been doing Senate elections since 2008. I know he got 35 out of 35 right in 2008, it looks like he missed 2 out of 35 in 2018 (Florida and Indiana, although Indiana's was an incumbent).
Has there ever been a race of national importance when he predicted a conservative politician would win and they ended up losing?
Overall though, 2016 was really his only big miss. And most of his misses were close states/races anyways. As it pertains to Senate elections, other than 2008 and 2018, thus far the Rs have performed better, and this is the year that there are enough Rs favored but not favored by a lot (e.g., Perdue, Daines) to see what happens.
Posted on 10/29/20 at 11:35 pm to Adam Banks
quote:I didn't lie. I downloaded the dataset from his webpage in excel. Added up all of the races by the top 100 pollsters (5019), all of the races called correctly (4159 or 82.87%, and all of the races outside the MOE (1163 or 23.17%).
You lie.
quote:Well you're pointing out 2 pollsters, and I said TOP 100. So how do you get lying out of that?
He rates abc news/wash post as A+. They call only 72% races correctly nbc news/wsj a- 73% I can go on and on and on.
But only 7% of ABC/Washington Post and 10% of NBC News/WSJ polls fall outside the MOE. And their ratings are largely based on the error of the margin, how their margins compare to the expected MOE, and how well they perform relative to posters polling the same race. For example, ABC/WAPO's average error is 2.8 points, which is 2.0 points better than the expected MOE and 1.7 points better than other pollsters.
NBC/WSJ has an average error of 4.0 points, which is 1.2 points better than the expected MOE and 0.7 points better than other pollsters.
Trafalgar has an average error of 5.6 points, which is 0.9 points WORSE than the expected MOE and 0.9 points better than other pollsters.
So calling races correctly is not that useful because if you poll closer races then you're more likely to miss than if you polled blowouts. Margins are more useful for that reason. And Trafalgar has the largest error, performs the worst relative to the expected MOE, and performs the worst relative to other pollsters.
And to highlight an example. In the 2018 governors race, Kemp won by 1.4 points. The local Atlanta Fox poll, ending on 10/29 had Abrams up by 1, so it missed by 2.4 points. Trafalgar's poll, ending 5 days later (11/03), had Kemp winning by 12 points, so it missed by 10.6 points, despite being even closer to the election (so it should be more accurate).
So what is more impressive, being off by 2.4 points 8 days before a close election, but ultimately "calling it wrong" or being off by 10.6 points 3 days before the election but "calling it right?"
This post was edited on 10/29/20 at 11:37 pm
Posted on 10/30/20 at 12:43 am to buckeye_vol
quote:
So what is more impressive, being off by 2.4 points 8 days before a close election, but ultimately "calling it wrong" or being off by 10.6 points 3 days before the election but "calling it right?"
Given that the outcome is binary correctly calling the race is clearly more important. close only works in hand shoes and hand grenades and coming really close means Stacey abrams still isn’t governor.
Posted on 10/30/20 at 12:45 am to The Boat
quote:You’re reading and talking about his site. That’s all he cares about.
It shows how full of shite Nate Bronze is. shite in shite out. Look at the actual voting data.
Posted on 10/30/20 at 1:35 am to The Boat
His model has Biden with a 57% chance of winning Georgia? A 67% chance of winning Arizona? A 64% chance to win North Carolina? A 65% chance to win Florida?
We're supposed to take him seriously?
We're supposed to take him seriously?
Posted on 10/30/20 at 6:56 am to CGSC Lobotomy
quote:
His model has Biden with a 57% chance of winning Georgia? A 67% chance of winning Arizona? A 64% chance to win North Carolina? A 65% chance to win Florida?
We're supposed to take him seriously?
Either this board or the national pollsters will never recover in the next few weeks.
Page 4 of 5
Popular
Back to top
Follow TigerDroppings for LSU Football News