117th Congress: DEMS IN DISARRAY!!!111!!

[Kepler]

538 now predicts Dems will save the Senate. Still about 4:1 to lose the House.

 

[SonofSouthie]

Who doesn’t want to lose rights?

I used to just look at your drivel, smh and skate on by. You're on full on block now cuz.

 

[MissThundercat]

538 now predicts Dems will save the Senate. Still about 4:1 to lose the House.

I haven't had booze in 27 months, so I will take that with a large coffee and 27 shots of espresso.

As noted before, I will be voting Dem in the same way I support needle exchange programs.

 

[Jimjamesak]

538 now predicts Dems will save the Senate. Still about 4:1 to lose the House.

Saw some polling for the Alaska special election next week. There's three candidates (one of the finalists dropped out): Republicans Nick Begich and Sarah Palin and Democrat Mary Peltola.

Both polls had Peltola actually leading in the first round with Begich second and Palin third but Palin and Begich were within 1-2% of each other (41-30-29 in the first, 40-31-29 in the second).

The interesting bit is the ranked choice, Begich is beating Peltola in both if Palin is eliminated but Peltola is beating Palin if Begich is eliminated.

Pretty much confirms my "depends on who gets eliminated first" thoughts.

 

[Handyman]

I don’t understand how you can be for one and against the other. You’re either pro rights or you’re not.

Not all rights are created equal.

When people can act responsibly with guns we can change the laws. When gun companies start acting with even a modicum of empathy and respect for human life we can re-open the discussion. Until then take your guns and shove them.

 

[Handyman]

538 now predicts Dems will save the Senate. Still about 4:1 to lose the House.

So bet on the House and against the Senate because 538 fucking sucks.

(I am partially joking but only about the first part)

 

[Kepler]

So bet on the House and against the Senate because 538 fucking sucks.

(I am partially joking but only about the first part)

538 was fine, the interpretation of it by the pundits was what sucked.

Math is funny. It works.

 

[Handyman]

538 was fine, the interpretation of it by the pundits was what sucked.

Math is funny. It works.

It wasn't the pundits interpreting it that was the problem, it was Nate Silver and his ilk trying to be pundits that did it. The man got up his own butt and he hasn't stopped based on his COVID takes. YMMV.

I like what 538 does in general, but they are not the end all be all and there have been quite a few similar stats groups or math nerds that have shown the flaws in their methodology going back a while. It is all well over my head when it comes to the math but sometimes it is fun to watch a good old fashioned nerd fight!

 

[Spartanforlife4]

Saw some polling for the Alaska special election next week. There's three candidates (one of the finalists dropped out): Republicans Nick Begich and Sarah Palin and Democrat Mary Peltola.

Both polls had Peltola actually leading in the first round with Begich second and Palin third but Palin and Begich were within 1-2% of each other (41-30-29 in the first, 40-31-29 in the second).

The interesting bit is the ranked choice, Begich is beating Peltola in both if Palin is eliminated but Peltola is beating Palin if Begich is eliminated.

Pretty much confirms my "depends on who gets eliminated first" thoughts.

A Begich is a Republican in Alaska? That's like reading about a Republican Kennedy in MA.

 

[Kepler]

It wasn't the pundits interpreting it that was the problem, it was Nate Silver and his ilk trying to be pundits that did it.

No. Just no. You are wrong on the math. Nate said "there's a 15% chance," and that chance occurred. That is how math works. I'm sorry it doesn't fit your narrative.

 

[dxmnkd316]

No. Just no. You are wrong on the math. Nate said "there's a 15% chance," and that chance occurred. That is how math works. I'm sorry it doesn't fit your narrative.

In fact, I distinctly remember Nate in like September and October 2016 saying he really wanted to caution against the models saying 100%. Even his own model he was seriously warning that it was starting to show weakness towards November.

 

[Spartanforlife4]

Yeah, Nate's been unbearable about Covid, but 2016 is not something to ding him on.

 

[Kepler]

John Stewart just beat the GOP like a conservative's step child.

 

[ScoobyDoo]

John Stewart just beat the GOP like a conservative's step child.

This exchange is incredibly sweet.

https://twitter.com/SenatorTester/status/1554572014660321280

 

[Kepler]

Just think, one out of work liberal comedian just did more for American veterans than all Republicans, combined, for the past fifty years.

Everyone in the military should vote blue. Red only wants you to bleed for their profits, which they will then never share with you and in fact screw you in the tax code to protect.

 

[Handyman]

No. Just no. You are wrong on the math. Nate said "there's a 15% chance," and that chance occurred. That is how math works. I'm sorry it doesn't fit your narrative.

Speaking as someone who followed Nate religiously on more than just his web site, I will just say you don't understand what I am getting at and that is on me because I explained it terribly thanks to lack of sleep the last 5 days which is on me. His 2020 "15%" prediction was spot on (I posted it here almost in real time) so that isn't the problem either. (dx I am not dinging him at all for 2016 he had no way to know that people were lying to pollsters about who they would vote for or that Comey would tank Hillary)

Nate's issue is he doesn't speak about enough that the math only works if the basic fundamentals of the numbers are correct. They almost never are, and certainly aren't even close anymore. Just because he has a formula that in theory works, doesn't mean the answer he gets is right. It just means it should be correct. He, like many of his ilk, are trying to quantify things that cannot be completely quantified and will almost certainly not be understood on its face by the average person watching CNN. Like college hockey and their algorithm...sure it is based on math but that doesn't mean it is giving us the best field or even the correct one. It only works if all the data is correct. Polling is a joke these days and no matter how much you try and tweak it to offset it there is just variables that cannot be predicted. (many of the problems are exasperated in local or state elections where polling is even less so one weird poll can gum up the works like it did say in Minnesota when a BS poll showed Tina Smith tied with Jason Lewis...it can get just too volatile) In 2016 it was the silent Trump voter, in 2020 it was quite a few different ones, in 2022 who knows what it will be. We often don't learn what the issue is until the post mortem. That is what he tried explaining in his Election Eve blog in 2020 but by then the narrative was already out there that his numbers showed a likely Biden victory and the Senate being better than 50/50.

Now in and of itself that is fine, but the problem is he knew that his numbers were being used to prove things they weren't proving and he never corrected those people for it. (and even promoted it at times especially on Twitter) Plus after 2016 he either didn't make corrections to his model or the corrections he made weren't enough to offset what has become apparent to many out there; that polls are just fundamentally flawed at this point so weighting them and aggregating them too will be flawed in the same way. He hedged late with his predictions (which again I have no problem with because his "15%" scenario played out almost to a T) but he doubled down on how that proved his system worked and then really gave himself a self colonoscopy. He never seemed to accept that the real world outside the numbers were showing that the numbers can't predict things in the way he thinks they can and how most people believe they are. When other stats nerds (people way smarter than I could even pretend to be) started pointing that out to him asking questions he went defensive and tripled down which flies in the face of what he was trying to do in the first place. He hid that well on his site but it was all over Twitter for weeks. Then, when the people who specifically follow elections started pointing out the flaws in his system and logic it got way to technical for people like me but needless to say his reputation is not sterling. (even worse with his COVID nonsense)

If you need it broken down better put it this way...the average person has very little idea what probability is or how it works. They believe that if you flip a coin 10 times it should hit 5 each way because that is what basic math tells them. Then there is the next level of person who gets the basics but can't see the forest from the trees. These are the people who play roulette and have math strategies to try and beat the game. They see 4 reds in a row and bet black hoping to hit it based on the numbers. (my buddy does this it can be infuriating to watch) More often than not they will win if they do it right, but they get super upset when things go sideways, which sooner or later they do because probability is not a guarantee. You can flip a coin 100 times and get 100 heads even if the odds are stupidly long. These two groups believe when the longshot hits it has to be a fraud because the math says it shouldnt happen! (you would think the word probability would be the tell) Nate used to factor that in to his commentary and somehow that got lost along the way. Now if you want to just be rando Twitter guy doing this as a hobby that is cool. There are lots of them and they are fun reads and know their stuff. If you are going to act as the face of Election Analytics (whether he meant to or not he became that after I think 2012) and write countless pieces and have your work referenced by pretty much all the news/media/pundits then you should really always remember to speak to the lowest common denominator.

The problem isn't the math, the problem is that math cannot tell you enough to get the full picture...especially when you can't even trust the data being used in the math. I like what he is trying to get at but much like sports analytics the nerds seem to forget there are factors no math will be able to deal with and that should be addressed. YMMV.

(we probably should not continue this here and I apologize for veering the thread this way)

 

[Handyman]

Just think, one out of work liberal comedian just did more for American veterans than all Republicans, combined, for the past fifty years.

Everyone in the military should vote blue. Red only wants you to bleed for their profits, which they will then never share with you and in fact screw you in the tax code to protect.

When Stewart has a bone to pick he does it better than most. He did the same for 9/11 responders when the GOP held up their health care as well.

Why Stewart relegated himself to a how on Apple I will never know. If he didn't want to be on TDS anymore I get that but he carries more weight with the average voter under 50 than anyone.

 

[Kepler]

I get that but he carries more weight with the average voter under 50 than anyone.

Jon's audience is over 50 now. Remember, he started at TDS a quarter of a century ago.

 

[Kepler]

Nate's issue is he doesn't speak about enough that the math only works if the basic fundamentals of the numbers are correct.

I do not agree with you. Granted I have not read 538 much since 2020, but in my experience both Nate and the guy who was his mini-me for a while whose name I cannot recall (Cohen maybe?) went out of their way to try to educate the Great Unwashed about probability and statistics and the environmental assumptions that make the math have some semblance of reflection of reality. The connection between the target, the concept, and the metric is fraught philosophically, and yes of course the average dolt will never grasp the nuance, just as believes in streaks or clutch hitting or the gambler's fallacy. But Nate tries, hard.

I don't care about his rock star image or whatever it was, I only care about the methodology, and in a kingdom of the blind he was the only one-eyed man for a long, long time.

 

[dxmnkd316]

Speaking as someone who followed Nate religiously on more than just his web site, I will just say you don't understand what I am getting at and that is on me because I explained it terribly thanks to lack of sleep the last 5 days which is on me. His 2020 "15%" prediction was spot on (I posted it here almost in real time) so that isn't the problem either. (dx I am not dinging him at all for 2016 he had no way to know that people were lying to pollsters about who they would vote for or that Comey would tank Hillary)

Nate's issue is he doesn't speak about enough that the math only works if the basic fundamentals of the numbers are correct. They almost never are, and certainly aren't even close anymore. Just because he has a formula that in theory works, doesn't mean the answer he gets is right. It just means it should be correct. He, like many of his ilk, are trying to quantify things that cannot be completely quantified and will almost certainly not be understood on its face by the average person watching CNN. Like college hockey and their algorithm...sure it is based on math but that doesn't mean it is giving us the best field or even the correct one. It only works if all the data is correct. Polling is a joke these days and no matter how much you try and tweak it to offset it there is just variables that cannot be predicted. (many of the problems are exasperated in local or state elections where polling is even less so one weird poll can gum up the works like it did say in Minnesota when a BS poll showed Tina Smith tied with Jason Lewis...it can get just too volatile) In 2016 it was the silent Trump voter, in 2020 it was quite a few different ones, in 2022 who knows what it will be. We often don't learn what the issue is until the post mortem. That is what he tried explaining in his Election Eve blog in 2020 but by then the narrative was already out there that his numbers showed a likely Biden victory and the Senate being better than 50/50.

Now in and of itself that is fine, but the problem is he knew that his numbers were being used to prove things they weren't proving and he never corrected those people for it. (and even promoted it at times especially on Twitter) Plus after 2016 he either didn't make corrections to his model or the corrections he made weren't enough to offset what has become apparent to many out there; that polls are just fundamentally flawed at this point so weighting them and aggregating them too will be flawed in the same way. He hedged late with his predictions (which again I have no problem with because his "15%" scenario played out almost to a T) but he doubled down on how that proved his system worked and then really gave himself a self colonoscopy. He never seemed to accept that the real world outside the numbers were showing that the numbers can't predict things in the way he thinks they can and how most people believe they are. When other stats nerds (people way smarter than I could even pretend to be) started pointing that out to him asking questions he went defensive and tripled down which flies in the face of what he was trying to do in the first place. He hid that well on his site but it was all over Twitter for weeks. Then, when the people who specifically follow elections started pointing out the flaws in his system and logic it got way to technical for people like me but needless to say his reputation is not sterling. (even worse with his COVID nonsense)

If you need it broken down better put it this way...the average person has very little idea what probability is or how it works. They believe that if you flip a coin 10 times it should hit 5 each way because that is what basic math tells them. Then there is the next level of person who gets the basics but can't see the forest from the trees. These are the people who play roulette and have math strategies to try and beat the game. They see 4 reds in a row and bet black hoping to hit it based on the numbers. (my buddy does this it can be infuriating to watch) More often than not they will win if they do it right, but they get super upset when things go sideways, which sooner or later they do because probability is not a guarantee. You can flip a coin 100 times and get 100 heads even if the odds are stupidly long. These two groups believe when the longshot hits it has to be a fraud because the math says it shouldnt happen! (you would think the word probability would be the tell) Nate used to factor that in to his commentary and somehow that got lost along the way. Now if you want to just be rando Twitter guy doing this as a hobby that is cool. There are lots of them and they are fun reads and know their stuff. If you are going to act as the face of Election Analytics (whether he meant to or not he became that after I think 2012) and write countless pieces and have your work referenced by pretty much all the news/media/pundits then you should really always remember to speak to the lowest common denominator.

The problem isn't the math, the problem is that math cannot tell you enough to get the full picture...especially when you can't even trust the data being used in the math. I like what he is trying to get at but much like sports analytics the nerds seem to forget there are factors no math will be able to deal with and that should be addressed. YMMV.

(we probably should not continue this here and I apologize for veering the thread this way)

The polls weren't wrong in 2016. They weren't wrong in 2020 either. Most were within the margins for error. The popular vote was something like a percent or two off estimates. THe problem is that fat tail error and the all-or-nothing system we have in 96% of our states' elections combine to create extremely dramatic shifts with only small changes in voting totals. Something like 160,000 people decided the 2016 election.


In the Tina Smith election, the poll you're referring to said Smith was +1 on Lewis (43-42) with a 5-point MOE. The next poll from them was 45-42 (+3) with a 5-point MOE as well. The election result was 48.7-43.5 (+5.2). I'd say they got it right. Only a handful of polls landed closer to the pin. https://projects.fivethirtyeight.com/polls/senate/minnesota/

If you look at the firms with A- or better, they were all very close to +5.2.

Fivethirtyeight projected a 12.3-point margin for Smith. What's interesting is the algorithm was beat by the polls in this case

• Lite projection (polls only): +8.3
• Classic (polls, fundraising, past voting patterns and more): +9.2
• Deluxe (Classic+Expert Opinions): +12.3

The Math Wins (this race)

Looking at the presidential election in MN, the polling average was 51.8-42.7 (+9.1). The result was 52.4-45.3 (+7.1).

Digging deeper into the data, of the states with more than 25 polls conducted in October or later, only two of the 21 had errors greater than 3.5%: Montana at 11.9% (polls underestimated Trump) and Wisconsin at 5.4% (underestimated trump). The median error was only 3%.

If you sort the states by error, the trumpiest states had the highest errors. Only RI, WI, and IL broke the top 20.

(All numbers are for the 2020 election)