Also this week: social media vs. everything, stories behind the papers, Evolution is going to double-blind review, and more. Oh, and Edward Scissorhands as a barista.
A couple of interesting story behind the paper interviews: Peter Chesson on Chesson 2000, and Jim Estes on Estes et al. 1998.
Evolution is transitioning to mandatory double-blind review in January. I think it’s a worthwhile experiment. I like that they’re making double-blind mandatory, because otherwise people likely would opt out non-randomly, making it harder to learn from the experiment. I remain unsure it’s going to be an improvement over the status quo and slightly worried it might be worse, given that the status quo at at least some ecology & evolution journals is gender-neutral outcomes without double-blinding, and that blinding gets seen through pretty often and (crucially) non-randomly.
Stephen Heard on why you shouldn’t publish the reviews your papers receive. I particularly like Stephen’s depressing but probably-correct point that it won’t actually increase trust in the peer review system. This is one of those issues for which there are good arguments on both sides (see Stephen’s links and the very good comment thread for these). I lean towards Stephen’s side but it’s not a strong lean. It’s hard to judge what would happen if reviews were published on a wide scale without, you know, actually publishing them on a wide scale. You can’t learn all that much from journal-level experiments on this because of self-selection. Anybody who doesn’t like Nature Communications’ new policy of publishing reviews can just not review for them.
Amy Parachnowitsch on whether social media is ruining everything. Good balanced post. A good example of why I don’t consider blogging to be social media, because I think blogging suffers less from some of the issues Amy identifies.
That viral tweet from 2014, predicting a 2016 Cubs-Indians World Series that would go to extra innings in game 7, is a prediction scam. Good fodder for an intro stats course or quantitative reasoning course. (UPDATE: The story has been corrected. Prediction scams are a thing. This particular example may have been a prediction scam, but that hasn’t been confirmed. I’d say the article is still useful fodder for an intro stats course. But now it’s also useful fodder for a journalism course. Thank you to a commenter for pointing out the correction.)
And finally, this week in Non-Scientific Humor Only Canadians Will Appreciate (ht @dandrezner):
I can only assume this coffee chain puts out its pumpkin spice lates around Christmas. 🙂
+1000 Internet Points for the best Tim Burtons/Tim Hortons joke in the comments. 🙂
Until I googled it I had absolutely no idea who/what “Tim Hortons” is, that’s a very Canada-centric reference! So in the sake of cross-Atlantic balance (and as you’re part of the Commonwealth) I’m going to post a link to the funniest thing that’s happened on British TV in a long time:
If that’s not worth 1000 internet points, nothing is! 🙂
There are only so many Internet Points, I can’t go giving them away just because people ask for them. Even though that is really funny.
Oh, wait, hang on a minute…[rummages down back of sofa] I found some. Here you go!
+1000 Internet Points. Don’t spend them all in one place. 🙂
You’re a generous man 🙂
🙂 This was my favorite British-themed tweet yesterday:
Oh yes, Tim Hortons. I’m ashamed for not having understood the joke right away. Having spent six months in Canada during my PhD, one would expect I’d remember it.
(Jeremy, I just remembered that I may have even met one of your students there… Either at a CSEE conference in Kelowna or in Churchill – Manitoba, in 2013. Not that this has any relation to this post or to anything else, for that matter. :D)
Harper opts out of a UN meeting to go to Tim Hortons. Oups that is not a joke… https://www.thestar.com/business/2009/09/24/doughnuts_over_diplomacy.html
minus 1000 points
The Vox story had to be corrected. They never actually had evidence that the tweeter spammed predictions in 2014 and it turned out he did not.
Might still serve as a good class example for students of journalism.
Cheers for this, will look into it and update the post if needed.
Yeah Deadspin had a good write up on it.
Just the product of garbage protocol. They might as well have linked to the “Tim Burton’s” tweet above and expounded on that as an example of journalists just making up quotes. After all, fabrication is a thing. This particular example of a NC Clinton voter may have been fictitious, but that hasn’t been confirmed…. 😉
Hmmm, that is indeed a pretty interesting find regarding the possible Twitter scam and questions of likelihood. And I definitely need something to take my mind off what happened in game seven.
I went to BR.com and looked at some numbers.
The linked Vox piece states the following:
“But as…Michael Schur pointed out, baseball pundits and gawkers have been predicting the Cubs championship for some time, with close observers noting their increasing quality and overall trajectory toward greatness. In other words, smart observers…knew that predicting a Cubs win, sooner rather than later, was one of the safest bets you could have made over the past few years.”
“So how do we explain why the rest of the tweet is so uncannily on target? There are two options: Gio’s tweet could be a remarkably lucky guess. But it could also be an example of a classic prediction scam.” Note that these options don’t include “GIO” making his prediction based on any actual baseball insight. Which I agree with because of the following data on recent performance and personnel moves, of the two teams:
2012: 61-101, 5th place NL Central, 36 games behind the Reds, 2nd worst record in MLB
2013: 66-96, last place NL Central, 31 games behind the Cardinals. Pirates on rapid rise.
2014: 73-89, last place NL Central, 17 games behind the Cardinals. Joe Maddon hired Nov. 2 or 3.
2012: 68-94, 4th place AL Central, 20 games behind Tigers. Terry Francona hired after.
2013: 92-70, 2nd place AL Central, 1 game behind Tigers and 1st wild card team
2014: 85-77, 3rd place AL Central, 5 games behind Tigers. Corey Kluber wins Cy Young award but Royals emerge as another very significant foe, very nearly winning the WS.
Terry Francona was hired after the 2012 season and immediately turned the Indians around. So…one *might* have reasonably predicted the Indians making it to the WS in 2016 given their changes and rapid improvements. But even that would be crystal-ballish, given the continued dominance of the Tigers and the rapid rise of the Royals *even just within the division*.
Theo Epstein was hired in 2011 and there was moderate improvement from 2012-2014, but the Cubs were still a bad team in 2014. Joe Maddon’s hiring announcement was right before the GIO tweet, Nov. 3, 2014. But the case for the Cubs at the time, based on actual play, is entirely non-evidential. We now know they made some good amateur draft picks but sorry, that doesn’t count–teams do that all the time and it’s a roulette wheel as to how those picks will develop. Now add to this his prediction of a tie score at the end of nine innings in game seven.
Therefore my conclusion: likely a scam indeed. A wild lucky guess is possible, sure, but highly improbable given no team performance evidence to suggest Cubs in 2016 WS, circa November 2014.
The other possibility an innocent analogue to the prediction scam. Say lots of people each make a very few far-fetched predictions in tweets. Then whichever one turns out to be right goes massively viral. Whoever’s tweet that was looks either very lucky or like a genius. And of course it’s luck, because with enough predictions from whatever source, some of them will turn out to be right. The point being that we shouldn’t be surprised that *some farfetched prediction or other*, about *some event or other*, turns out to be right. Most everything has been predicted by somebody, at some point.
p.s. There’s a really interesting-looking book out I want to read called Superforecasting. Some economists asked lots of people to predict lots of world events–election outcomes, occurrences of wars or terrorist attacks within specified space-time windows, etc. Some people did better than others, of course, which you’d expect to happen just by chance. But then (from what I understand) those better forecasters continued to make correct predictions much more often than other people, in such a way that you can’t really explain it as just luck.
There are also studies of this sort of thing in the context of whether any stock brokers can consistently beat the market. As I understand it, the answer is either “no” or “maybe, but it’s really hard to identify those people”.
Yes that could happen, but I much doubt that it’s the case here: the prediction is just too specific (which likely contributes to why it “went viral” in the first place). Just a series going seven games at all, assuming two equally matched teams, is 1 – (29/32) = 3/32. If you also include p(any two particular teams being involved), and p(game seven tied after nine), you’re down near zero somewhere. It would take a LOT of random tweets to cover the possibilities.
Speaking of low p, note that the last WS to be tied at the end of nine, in game seven, also involved the Indians (1997, similarly devastating loss, to the Marlins). But that Indians team wasn’t as good as this one.
My number above is wrong: p(evenly matched teams going seven games) is 1 – (22/32), not 1 – (29/32), thus 10/32, not 3/32. [Neg. binomial p for 4 wins & 3 losses, for a single team, where p(win) = 0.5, = 5/32. Times 2 for the two teams.]. I gotta be more careful.
But even assuming equal chances of any team winning its league’s pennant, you’ve still only got a (1/15)^2 = 1/225 chance of Cubs vs Indians WS. But that 1/15 is being generous to the Cubs, circa Nov. 2014, when yes, they were improving under Epstein, but still 16 games under .500. You very roughly need to be about that many games *over* .500 to land a playoff spot, typically. Joe Maddon’s going to improve you by 32 games in two years? Well, he did well better than even that, but could anyone have known it? No, because for one thing, you’ve got four other teams in your division, against which you play about half your games, and you don’t know what they’re collectively going to do.
There is yet another possible explanation here: pre-target the Cubs and Indians, since it was well known by all that they had the longest WS title droughts, per league*. Forget about the other 28 teams. Then script out a bunch of tweets, one for each of the next 25 years or whatever, where the two end up deadlocked at the end of game seven. With a 10/32 = 0.31 per-year chance of the WS going seven games, it doesn’t take long before you get one success, e.g. after five years you’ve already up to p = (1 – 0.31)^5 = 0.85, and at ten years it’s up to 0.98. You do still need to factor in the p(game seven tied after nine) and that will lower the total prob., but keep in mind that low scoring games have a higher prob. of being tied than do high scoring games. Strangely though, this one was a high scoring game.
* note that this is per-city, not per-franchise. How do you compare these two with franchises with shorter (but not insgnificant) histories who’ve still *never* won one, or even been in one for that matter. Cases in point: Expos/Nationals, Pilots/Brewers, Padres, and Mariners (all with 0 WS titles in 48 years), and Senators/Rangers (0 titles in 56 years).
Speaking of being lucky, there are actually many interesting stat/prob/likelihood/prediction questions in baseball, especially given the recent “sabermetric” statistical revolution. I wrote a post dealing with one of these this week, the “Pythagorean” winning percentage expectation and the concept of “luck”:
See the “Luck” column in this BR.com standings table: http://www.baseball-reference.com/leagues/MLB/2016-standings.shtml for reference.