In 2018, which political prognosticators will you pay attention to?

Politics & Elections
41

Sure, I can believe he tilts Republican and this may color his outlook/predictions. I have little interest in his predictions prior to the 16 election cycle. It’s his analysis of the Trump phenomena(and all that surrounds it) that im interested in. As I said earlier these sort of pundits are hit and miss. The greatest of political minds in one election turn out to be chumps at subsequent elections. Im not silly enough to believe one correct prediction makes him a guru.

Adams did also predict Trump would win the nomination. I mean that has to count for something when most pundits were still viewing Trump as a laughing stock for much of the Primary campaign.

42

I don’t think it a copout at all. Had Clinton won we wouldn’t soon be getting a book by Scott Adams on his 2016 prediction success. All we’d get instead is a tail between the legs blog post explaining himself. There is a reason people feel his prediction of a Trump win is significant - because a Trump victory was his default position during most of the campaign. Had Clinton won no-one would currently be giving a flying fuck about Scott Adams right now. It would be far, far harder for Adams to claim anything he said on Trump, politics or the news media was now significant had Hillary won.

Im struggling to understand why posters in this thread think Nate Silver’s status has not been reduced after mostly getting it wrong this past election but Scott Adams’ reputation should not be significantly(even if temporarily) enhanced after getting it mostly correct.

43

Because Silver has a very good track record and actually uses lots of data and analysis (and a 30% chance for Trump is pretty reasonable looking back, based on the numbers), while Adams essentially made a lucky guess.

44

Scott Adams predicted that the team he wanted to win would win, but he did so not because of any real analysis but because that’s what he hoped for. We’re talking about people who make predictions for a living and go where the data analysis takes them. Anyone can get lucky. Just because someone who hardly understands basketball wins an NCAA bracket pool, that doesn’t mean I’d take their opinions seriously if we were betting on other sports or other NCAA games for that matter.

45

I see excuses and benefit of doubt are being wheeled out for Silver whilst the same excuses are not being given for Adams. Silver had Hillary with a 90% chance of victory at one point earlier in the campaign. Three weeks before election day he again had her at near 90%. Yet Adams’ earlier wobbles are not excused in any way. If anyone had Clinton with a 90% chance of victory three weeks out then something is very wrong with their model or analysis when she lost only three weeks later.

46

Silver has a track record of presenting chances of victory for a wide range of elections over several years with a model based on mathematics. When the model misses, that goes into refining the model.

Adams has a history of creating elaborate explanations for why what he desires to happen will happen, and when he misses he comes up with excuses for why he was right all along, or at least why it’s not his fault he was wrong.

Your belief that Adams would have posted a “leg between the tails”-post if Clinton won does not seem to fit the pattern I have observed with him. Do you have any examples of him ever doing anything other than doubling down or shifting the blame?

47

When I said *had Hillary won Adams would post a tail between the legs blog post *I meant he’d likely write a post in which he’d explain away or rationalize his wrong prediction. I did not mean Adams would post an unambiguous mea-culpa. I was responding to someone claiming a Hillary win would also have been a confirmation of Scott Adams’ predictive powers in 2016. This is not so. There is a reason Adams has made waves this election and it’s not because he was wrong or hedging. A Hillary win would have led Adams to become something of a laughing stock(see Dick Morris 2012). Adams may not be as good as he, or I, believe he might be but I have little doubt many on here are unjustly minimizing his accomplishments in the 16 election. All things considered he called it pretty damn accurately.

Im not so blinded by Adams as to think he is now a political guru. However, he did enough in 2016 to hold my interest with his take on the current political climate.

48

I can’t believe there’s a serious discussion Scott Adams. The guy is a crackpot. He’s been a crackpot for a long time, long before Trump. Yeah, he picked Trump because he wanted Trump to win. My boss is a Blue Jays fan, if I ask him who is going to win the next Blue Jays match he’s very likely to say “Blue Jays”.

49

Being a loudmouth and a fan of a particular politician isn’t an accomplishment.

50

Thanks for this post. It was insightful and informative.

51

I appreciate that sir. We may not agree politically but your openness to considering data, regardless of the degree to which it supports your point, is commendable.

52

I don’t even get the comparison.

For one thing, Adams was not correct; he predicted Trump would win in a landslide. He was “Running unopposed,” according to Adams. Trump did not win in a landslide.

For another, Silver didn’t start predicting things just with the 2016 race; he’s been doing this for years, across a multitude of elections, and his track record is excellent. Adams just got one prediction - well, actually, he didn’t get it entirely right. Sort of right. (I should point out he’d actually made quite a lot of definitive predictions when it comes to politics, and has been wrong about most of them, including his famed prediction that by August 2017 it would be embarrassing to not be a Trump supporter because Trump will have been amazingly successful.)

This is been gone over again and again, but no, it does not prove the model is wrong. I realize this sort of thing can be difficult to understand at first glance but if you do think about it it’s not THAT hard.

Suppose a good baseball player steps up to the bat, and you ask me, “Is this guy going to hot a home run?” As a FANTASTIC home run hitter will hit a homer 7, 8 percent of the time, so the correct answer is “Probably not. He’s likelier to hit one than most guys. A homer here is as likely as it ever is. Still, probably won’t happen.” IF he’s up against a shitty pitcher maybe it’s a 10 percent chance, so I tell you it’s 10 percent likely.

Then the guy hits a homer. Now, if you start crowing how wrong I was, I have to tell you, I’m going to think you’re a pretty big idiot. I didn’t day he WASN’T going to hit a homer - and had I said that and he grounded out, it still would have been a stupid prediction. The correct answer to a question like that is always “Maybe, and here is the approximate chance.”

I’ll tell you what; from now on let’s bet money. You bet in accordance with Scott Adams’s predictions and I’ll bet according to 538, and we’ll see who makes money.

I think Adams made some good points about elements of political marketing; his observation that Trump’s simple, repetitive, absolutely relentless approach was a much better one than people were giving him credit for was absolutely correct. But he’s quite wrong about a lot of things, and it’s plain in the text of his blog, which is becoming increasingly partisan and divorced from an objective assessment of reality.

53

The landslide prediction by Adams: this is partly why I said Adams was only largely correct. His prediction was not completely accurate. However, I will say that the landslide prediction was oh so close to coming true. Clinton was one stabilising bollard away from becoming almost unelectable at the 9/11 memorial service, one secret service agents strong arm from falling flat on her face. Very, very few were predicting that Trump would be virtually unopposed on election day. I would say Adams was closer to being correct on this than most of his critics now admit.

My misunderstanding of Silver’s model and number crunching: yes, perhaps I am misunderstanding. I realise that a 90% chance of victory is less than certain. However, im not completely convinced that a pundit/analyst who tells me at the height of an election campaign that candidate A has a 90% chance of winning and candidate A loses has a right to be giving us those sort of percentages, ever. I feel Silver did not take into account the element of risk in his prediction of a Clinton win; the risk of a health event or the risk of a Comey type letter. These potentialities were missing from Silver and his ilk.

A bet on on Adams’ predictions versus those of 538: no, I won’t do that. I readily admitted earlier that such predictions are a mugs game. That some predictive guru at one election becomes a laughing stock at the next. I wouldn’t be at all surprised if this is the case with Scott Adams. I also agree with you that he has become more partisan in recent months. He is still a more worthwhile read on today’s politics than 95% of pundits.

54

I’m sure Adams would say this too – he also made up a sock account on message boards to crow about what a genius Scott Adams was (a couple of years ago, IIRC). And then when caught, explained that it really wasn’t a big deal at all.

His predictions were largely inaccurate. In addition to predicting Trump would win by a landslide (which was not even close), he also predicted that Hillary would win on another occasion (which did not occur, but was much closer to occurring than a landslide for Trump). At no point that I’m aware of did he predict a very close election with Trump losing the popular vote but winning the electoral college, which is what occurred.

55

Looking at your table (and thank you very much, by the way), it’s clear that the predictions were clustered and similar and failed in the same way - which makes perfect sense as they were based on the same underlying poll data. So it’s really GIGO, if the polls aren’t working, it doesn’t matter which predictor/aggregator you follow.

Which brings us back to the error margins: by estimating a 30% chance of Trump victory (instead of 10%, or 1%, or 0.1%), Nate Silver did a better job at showing the true uncertainty in the data than anyone else. (Among the predictors who presented their uncertainty level - I don’t think RCP did?)

56

Neither Silver nor Adams were entirely correct in 2016, but Silver was closer to correct than Adams was. Adams predicted a Trump landslide of unprecedented magnitude, that wouldn’t even be close. Silver predicted a close race that Clinton had a better chance of winning, but that Trump had a significant chance of pulling out a close win. What actually happened (a close Trump win) was something that Silver gave a 30% chance of happening, and that Adams gave a 0% chance.

That said, Silver did still get something colossally wrong in the 2016 race: During the primaries, he said repeatedly that Trump had a less than 1% chance. He said this, not based on his math, but on his gut, and his gut was horribly wrong. I have a great deal of trust in Silver’s math, but none at all in his gut.

I believe that he’s learned his lesson from that, and now does not trust his own gut, either. Presumably, he won’t make gut predictions in 2018 or 2020, and if he does, I’ll ignore them. But as long as he bases his predictions on the math, the track record shows that he’s better at the math than anyone else out there.

57

RCP doesn’t offer a % probability of victory like Wang, Silver, the NYT, and some others do, but they do offer two separate maps:

RCP Electoral Map, which showed a whopping 14 states worth a combined 171 ECV as “toss up”.

RCP No Toss Up States Map, which just assigns the state to whichever candidate they were leaning, regardless of how slight the lean.

The overall effect, to me at least, was basically saying ‘there’s a bunch of uncertainty here’.

58

Ah, that’s right. To me, calling, say, Florida a “toss-up” versus saying it’s 55/45 (Silver’s final numbers) is mostly a matter of preference - since we can’t re-run the experiment to see if 55/45 is false precision, or if calling AZ and GA toss-ups is far too timid. Especially since Silver’s written analyses were basically “there’s a bunch of uncertainty here.”

59

I’d still love to do RMSE analysis for other prognosticators and also for 2008 and 2012, but I don’t have time at present to track down the data. If someone, or a collection of someones, can post links to said data to this thread, I’ll do it in the next couple days. Even better yet, if someone wants to do the calculations themselves, feel free to go nuts. The actual math is pretty straightforward.

60

I think the key to doing this is extending the calculations to weigh by their own error bars (of course requiring even more data collection…)