Can we please kick Nate Silver to the curb

[boston bean]

His theories and calculations for election predictions are failing.

He’ll make up some bs about how he was right by excusing his errors.

[SouthernCal_Dem]

1. All he does is average the polls

It appears your issue is with the polls and not Nate Silver.

[boston bean]

3. Yeah and his "averaging" has proven useless.

[SouthernCal_Dem]

6. Let's see how the final numbers look

There are still millions of ballots that need to be counted, but it appears polls underestimated Trump support in the midewest yet again. However, the national popular vote could end up being accurate or at least within the margin of error.

[Tribetime]

58. Ohio was way way off. So was WI. MI and PA

[In It to Win It]

62. ... and FL

[In It to Win It]

64. It's a shit-in shit-out situation. He collects the public polling data from the "reputable" polling

firms and puts them in his models.

It turns out that the polling is SHIT!

[Jose Garcia]

12. It's a little more complicated than that

His model gives more weight to some polls based on how accurate those polls were in the past.

[SouthernCal_Dem]

22. You're correct

I just didn't feel like typing everything out.

[Towlie]

39. No assertion of probability can be called "wrong" unless it's either 0% or 100%.

That's why rain forecasts are given in probability percentages. If a forecast is 0% and it rains, or the forecast is 100% and it doesn't, you're free to conclude that the source is unreliable. But aside from that you'd need to randomly select and compile several of the source's past forecasts and their outcomes, and apply a sophisticated statistical test to determine the confidence level that the source deserves.

Note that Nate Silver's perceived inaccuracies can alternately be interpreted as evidence of "extraconstitutional shenanigans", as he calls them:

How FiveThirtyEight’s 2020 Presidential Forecast Works — And What’s Different Because Of COVID-19

Scroll 2/3 of the way down and find the third paragraph after "Step 3: Account for uncertainty, and simulate the election thousands of times".

Before we proceed further, one disclaimer about the scope of the model: It seeks to reflect the vote as cast on Election Day, assuming that there are reasonable efforts to allow eligible citizens to vote and to count all legal ballots, and that electors are awarded to the popular-vote winner in each state. It does not account for the possibility of extraconstitutional shenanigans by Trump or by anyone else, such as trying to prevent mail ballots from being counted.

more...

[WSHazel]

2. It's not Nate Silver's fault that roughly 70 million Americans voted for fascism

I didn't think it was possible for Trump to expand on his 2016 voter number either. I grossly overestimated my fellow Americans.

[sinkingfeeling]

4. No

[USALiberal]

5. FFS really? How about you whine about the pollsters?? nt

[IADEMO2004]

7. No one would ever lie to a pollster. Right?

[Claustrum]

8. While I agree with you, I don't think this is completely Nate Silver's fault.

He is aggregating polls so when the polls are wrong, his model is wrong. It isn't his fault that the input polls are wrong. Though, I will not believe in polls for a long time, so I wouldn't believe any of his models either.

[boston bean]

11. He is bringing nothing to the table.

Zero. Zip. Nada.

Except taking bullshit numbers and packaging them for the masses.

It doesn’t work. He is doing nothing. Easy to just blame the polls he uses to spout out garbage.

He doesn’t deserve the following he has.

[USALiberal]

16. So why don't you use the polling data and decide who will win? nt

[Claustrum]

27. Don't get me wrong. I stopped following him, 538 and any other pollsters right after Nov. 3.

I just think the fault isn't completely on him. He deserves some for sure. I think having a standard to view the differing range of polls are good to give them perspective. But when most polls are so wrong, they are completely useless. Thus, I personally will not follow any of them. I am sure a lot other people feels the same and his following/incoming will dramatically drop.

[Miguelito Loveless]

18. True, but I think polls are going to have a lote less credibility going forward

thus their income will dry up.

A LOT of candidate money is wasted on polling.

[BootinUp]

9. He didn't take the polls. Might be a clue. Nt

[boston bean]

13. So what is he adding except to give validity to these

friggin sham polls.

[USALiberal]

19. He said Trump had a 10% change of winning! You understand 10% happens correct? nt

[BootinUp]

61. His interpretation of the stats is pretty useful

In my experience. He has discussed at length in podcasts or blog posts that even if the polls missed this time like in ‘16, that there was enough polling advantage to say Biden was the clear favorite and he was right.

[LisaL]

10. He doesn't do his own polls. He uses polls done by pollsters.

It would appear that just like in 2016, polls were heavily skewed democratic.

[Buckeye_Democrat]

14. He's just a poll aggregator.

The pollsters have been struggling for YEARS because of technology that makes people less likely to answer phone calls from unknown numbers.

Maybe they'd do better if the name of the pollster appeared in smart phones or whatever? I never answer a phone call when I don't recognize the caller. That's a big difference from years ago when me and my family members would rush to answer the phone.

I don't know if pollsters hire a lot of minorities or not, but racist white people are probably more likely to lie to them after hearing their voices imo. Recordings might work better.

[Loki Liesmith]

15. You don't know what you're talking about

Nate’s algorithms compute the probability a lead can survive a polling error of a given magnitude. That’s all they do.

And it looks like the algorithm performed well.

[USALiberal]

17. +1000! nt

[gab13by13]

20. Who cares?

[cwydro]

21. Same with the DU prognosticators.

I cringed every time I saw a predictions for a landslide, a blue wave, a rout, a humiliation. But probably not as much as they’re all cringing now.

I never understand why they do it. Pretty much the same ones who were wrong in 2016. But still they continue.

[boston bean]

23. I believed in 2016. Learned my lesson well.

[LisaL]

26. I actually felt better this time than in 2016.

In 2016, polls were close. But it would appear polls were even more skewed this time.
I guess we just can't believe most polls.

[LisaL]

25. Yep.

But polls were better this year than in 2016. Which would suggest that democratic lean was even bigger than in 2016.
Polls were 6, 7 point off in our favor.

[USALiberal]

31. +1000! nt

[MontanaFarmer]

24. Nate, in one of his last posts,

gave his model's protection in the event of a polling error exactly like that in 2016. It's going to wind up being the electoral map, except i think ultimately Biden wins Georgia. Modeling works, it just requires the input data to be accurate. Nate had Biden winning. Biden is going to win.

[treestar]

28. Sometimes I think we should just wait to count the votes

all the predicting is useless. Human nature while waiting, I suppose.

The campaigns might use polls to help determine where to expend resources. For us, it's just blather until the real poll.

[Dem2]

29. Every cycle

Someone has to be kicked to the curb.

I think this is an emotional reaction. He was one of many aggregators who said the same things - how were they to know that the issue in the polls hadn't been resolved by the pollsters?

[hlthe2b]

30. Frustration is at the limitations of statistics-- and polls suffer the same issues

As we see in all of science, whether in conducting clinical trials or trying to differentiate populations and predict outcomes in polling. And that is for multiple reasons, but especially in the fact that choice of inputs is everything. By that I mean the common refrain: "garbage in, garbage out" in terms of findings.

Inputs in polling are not unlike inputs in the methods governing epidemiological studies and clinical trials. If you do not adequately account for bias in your sampling then you cannot ever hope to achieve a valid outcome.

More to the point, however, polling like such clinical trials are statistic-dependent. By that, I mean that there is always going to be the limitation of being able to reliably differentiate outcomes when the comparison groups are close. By that, I mean that a clinical trial attempting to show very small differences in outcome between two drugs will likely fail because the "power" (ability) of any study to show close differences is insufficient without enormous sample sizes, while a much smaller sample of participants can readly detect LARGE differences. Similarly, a poll can readily and accurately predict differences between very disparate populations sampled, while when the percentage of people voting differently is very close can fail to be accurately quantified. So a population that is so close to 50:50 is nearly impossible to poll reliably.

I understand the frustration with the polls seemingly failing us. Just as I've bemoaned the failure of a clinical trial to accurately show what I thought would be a significant difference in a new medical treatment. But this is the limitations of statistical methods, not a given person nor necessarily a given pollster.

[muriel_volestrangler]

32. Fuck - do you mean Trump just won?

Or that everyone is now saying Trump is bound to win?

Because the presidency is the vote most people are talking about.

[milestogo]

33. Maybe he's not taking into account the cheating by Republicans.

[Towlie]

34. He's not, and he wrote that himself. Here's the link:

How FiveThirtyEight’s 2020 Presidential Forecast Works — And What’s Different Because Of COVID-19

Scroll 2/3 of the way down and find the third paragraph after "Step 3: Account for uncertainty, and simulate the election thousands of times".

Before we proceed further, one disclaimer about the scope of the model: It seeks to reflect the vote as cast on Election Day, assuming that there are reasonable efforts to allow eligible citizens to vote and to count all legal ballots, and that electors are awarded to the popular-vote winner in each state. It does not account for the possibility of extraconstitutional shenanigans by Trump or by anyone else, such as trying to prevent mail ballots from being counted.

[JCMach1]

36. Hard to account for USPS and hacking that leaves little to no trace

I am pretty certain the RGV was hacked in Texas and no one will ever investigate because of where it happened and the level of pandemic there

ES&S with those hackable modems were also used in Miami-Dade

[uponit7771]

42. +1, none of them are ... They minimize the affect of voter suppression

[Codeine]

35. I'll certainly give the aggregate model less credence

if only because I’m giving the polls less credence. That isn’t on Silver, or honestly even the pollsters. They thought they had adjusted the model appropriately, but I just think we may be at a point where polling is fundamentally broken; technological and social changes have rendered it a moribund science.

[ProfessorGAC]

57. Back To The Old Saw...

..."Garbage in, garbage out!"
The data he used was flawed. He can't correct for flaws he didn't know existed until after the fact.
If he tried, his model wouldn't be valid.

[cbdo2007]

37. It is actually playing out EXACTLY as he has been saying for the past month.

[Azathoth]

38. This isn't Silver's fault. Washington Post had a poll showing Biden up 16 points

in WI last week. There was profound, systemic error in much of the polling.

Silver's model is only as good as the data fed into it. If all the reputable polling has major bias, it's going to skew his predictions.

[Captain Stern]

40. How so? Did I miss something...has trump won?

Nate Silver doesn't conduct polls. He's a math guy. He takes the poll results that are available, and uses the results of those polls to weigh the probabilities of what will happen. That's it.

Where there any particular polls that he used, that you felt he shouldn't have? Did he way some more heavily than you feel he should have. If so, which ones?

[Klaralven]

41. Polling needs to be replaced by an analysis of what is happening on social media

I'm sure that Facebook and Twitter have a better handle on the sentiments of Americans than do the polling organizations.

[muriel_volestrangler]

44. Heavy sarcasm? (nt)

[Klaralven]

45. Not at all. I'm completely serious.

Pollsters only get 6% of people they call to respond to their questionnaire. So the respondents in polls are the most sociable and least busy segment of the population.

Thus, there is huge self-selection error in the sampling.

Instead, it is far more accurate to do lexical analysis on the total universe of tweets and use big data analysis to identify issues, attitudes and trends.

Of course this also assumes that the social media companies aren't pressured to drive significant segments of the populace off their platforms and onto alternate services or to the dark web.

I'd be surprised if Facebook and twitter didn't already have the analysis infrastructure already in place for marketing and advertising purposes.

[muriel_volestrangler]

46. Oh yeah, social media is oh-so-reliable. No one has ever detected any fake accounts

or coordinated networks of bots or ultra-partisan groups who try to influence others. It's completely immune to commercial pressure too. Plus you can guarantee that everyone's behaviour on it is truthful, not at all linked to how they want to be seen by their peers or employers, and everyone shows their political feelings in exactly the same proportion as everyone else.

What's not to like about it? "But I saw it on Facebook" is a byword for reliability. It's right up there with "people say".

[Klaralven]

53. Social media knows a lot about posters, whether they are fake or not, bots or not.

They have the IP addresses that the posts are coming from. Probably know whether poster is using a VPN. They can construct graphs of poster relationships that tell a lot about the nature of groups.

Besides, in politics, truth doesn't really matter. It's mostly an intellectual fantasy being played out by the emotional based on rhetorical presentations of partial falsehoods.

What you want to know is what influences are being effective in shaping opinions of voters. Whether they are being shaped by rational discourse of facts versus illogical conspiracy theories spun from thin air doesn't really matter.

[muriel_volestrangler]

55. Oh, there was me thinking this thread was about how to predict how people will vote

rather than becoming Cambridge Analytica and trying to manipulate them.

Go ahead. Stare into the abyss, and try to become it. We won't join you.

"Truth doesn't really matter". Can I quote you in future discussions about voting?

[Klaralven]

59. "in politics, truth doesn't really matter" -- not specifically about voting.

Voting is only an approximate process.

[muriel_volestrangler]

60. This thread is about "election predictions". ie "voting predictions".

As I said, if you want to be the new Cambridge Analytica, go ahead. Don't expect Democrats to join you.

[Bonn1997]

43. Statisticians must help to get answers

Pre-election and exit polls have matched vote counts much better historically than recently, and even now, they appear to still match often. People like Nate will be the ones to figure out why they sometimes don't match. The "shy Trump voter" and "rigged voting machines" are hypotheses that must be tested. Who else do you propose test them?

[BannonsLiver]

47. I think the best bet moving forward

Is to assume all our general elections will be extremely close. Of the last 5, 3 of them have been decided by extremely close margins. It took a once in a lifetime political talent like Obama for the 2 that weren’t.

[MoonlitKnight]

48. Cook Political Report was a failure

Worse than 538.

[BannonsLiver]

49. Yeah a total disaster

I think my respect for Charlie is gone for good.

[Rstrstx]

52. Just about everyone was worse

Though to his credit Dave did tweet this on Oct 31:

Link to tweet

[Rstrstx]

50. For the presidency, so far he's only missed on 2

Florida and ME-2. North Carolina may also be off, we’ll see what happens with the absentee ballots, and if he did miscall GA it was only by the thinnest of margins but I’m hoping it ends up in the D column.

Florida obviously is the big miss, anyone watching the Miami-Dade results on Florida’s website saw a huge red flag there once early voting started. Perhaps his model should start taking early voting into account.

I’d argue the problem isn’t with Nate, it’s with the pollsters. His predictions are only as good as the data he’s given. Remember that poll that showed WI with a +18 Biden lead? It was from a reputable pollster too.

[Alhena]

51. He gives good catnip to political junkies like me, I'm glad he's around

[Tarc]

54. Far as I'm concerned, polling is dead

They'll have to come up with some other metric to keep the whackadoos like Tulsi out of the later primary debates. Other than that I do not see a downside to scrapping it all.

[BusyBeingBest]

56. Nate Silver gives you stats and shows you probabilities

based on the most reputable polls. That's it. You read him to draw your own conclusions. He is not a fortune teller.

[BGBD]

63. This is dumb

Silver said a little while back that of Biden lost in Florida, which he was only a little better than even odds to win in the model, his odds would drop to basically even to in the WH.

Florida polls were off and we lost Florida.....the rest of this election sure felt like a 50/50 race didnt it?