Correcting Bias (Bill's June 9th Poll)

June 25, 2019

This is NOT about Bill’s Twitter polls, or one in particular, about political candidates, though it may shed some light on what’s ultimately driving those polls, but rather about a poll question concerning which MLB player, Bill Madlock or Harvey Kuenn, was "greater." https://twitter.com/billjamesonline/status/1137867406259490816 . To murder the suspense, Madlock romped, getting 56% to Kuenn’s 24%, which in turn barely eclipsed the third-place vote getter, "Never heard of either one," with 20%.

As Bill’s intro explains, they had superficially similar offensive stats. I voted for Kuenn, though he largely played before my time. In the second year I followed baseball at all, 1962, he played left field for the pennant-winning Giants, platooning sometimes with Willie McCovey. The next year, as a Giant and then in 1965 as a Cub pinch-hitter, he was the final victim in two of Sandy Koufax’s no-hitters but I missed his stardom almost entirely. In the years from 1973 through 1987, on the other hand, I was a fanatical watcher of MLB and followed Madlock’s career closely, beginning to end. I voted for Kuenn, though, because he started his career as a shortstop, implying that he might have been a good defensive player in the decade before I saw him, and because in his prime he was considered to be the equal of Rocky Colavito by at least one nominally-sane MLB GM. Also he won a pennant as the Brewers’ manager, which maybe tells me he was a pretty smart ballplayer.

Mostly, though, I voted for Kuenn because I knew he would lose the poll, and I suspected he would get creamed. I wanted to minimize the cremation.

How did I know Kuenn would lose? Recency bias.

At the outset, I want to make clear that I have no clue why Bill runs polls at all, much less why he runs the particular ones he runs, only that I share the widespread curiosity about his motives and his aims. In this case, I believe that running a poll about Madlock and Kuenn could serve as a part of a longish series of polls whose aim is to see how much of the results in Bill’s Twitter polls about baseball may be attributed to recency bias.

Forget about Kuenn and Madlock for a minute. If you were designing a test for recency bias, wouldn’t you run a series of polls asking who was greater, Player A or Player B, if A and B are nearly identical in stats but played a generation apart? Just looking through MLB history and identifying twins born decades apart is a fun project. That was one of the coolest parts of the Historical Abstract (the first one, I think), where Bill told us which old-time player (whom no living person ever saw play ball) resembled which contemporary player, statistically and often physically. To this day, when I see the name "Ed Delahanty," I flash back to Bill’s comparison of him to Joe DiMaggio (whom I never saw play either, but who has made a powerful visual impression on me nonetheless) and Delahanty comes alive in my mind in a way he never had before.

So in a series of poll questions juxtaposing very similar players of different generations, where the more recent player would always win (which I think he would), Bill could confirm not only that there IS a recency bias, but quantify how strong it is. At least among Bill’s Twitter-followers, we could then discount other poll questions, and understand how much to discount for recency. That would be valuable information, and far more interesting (to me, at least) than just knowing who was greater, Kuenn or Madlock.

In a real sense, of course, we know that answer before we get results: neither is greater, at least not so’s you’d notice. It’s just the perception of which one was greater that we would learn, and that perception, I believe, would be overwhelmingly due to recency bias.

Since pairs of twins cannot be identical, and certainly not in terms of precise dates of their careers, we could also learn the degree to which their career-years affect the recency-bias question. Take the twins I’ve written a few articles about, the Willies McCovey and Stargell. Their career totals are incredibly close to identical, but they played not-quite contemporarily, Stretch debuting in 1959 and Pops breaking in in 1962. This was so close I’m totally comfortable describing them as contemporaries, but I’m sure there are a few participants in Bill’s polls who started following baseball in 1979, catching Stargell’s MVP season but missing virtually all of McCovey’s career. That would be a very small recency-bias, probably IMO enough to swing the results in Stargell’s favor, though not necessarily.

If you were to ask about an early 21th century player and his early 20th-century counterpart, however, I would virtually guarantee that the results would make Madlock/Kuenn seem tighter than Gore/Bush. Run enough of these polls and you can perhaps learn how much to discount the results of other polls on issues that are susceptible to recency bias in baseball, at least among Bill’s set of Twitter followers.

I’m not suggesting that that’s what Bill was up to on June 9th. "Recency bias" is much more my issue than Bill’s—I see it all over the place, whereas Bill tends to minimize its importance when I have suggested it as causing the poll results it has. But if Bill’s up to something entirely other than establishing a base here, as I suspect he is (up to something entirely other), it would still work as an effective tool in this regard.

Where this suspicion of mine overlaps with the political polls is in the matter of the polls not measuring what we might think they’re measuring (greatness in ballplayers, or political support among Presidential candidates) but rather in establishing bias in those being polled.

This is very useful, but only in the long run. Say Bill has 10,000 people who regularly respond to his polls. As has been pointed out numerous times on this website, and on Twitter, the polls are totally not indicative of the wider population, being a self-selected group, wildly unrepresentative of the population as a whole, being more male, more middle-aged and older, less conservative, higher-income, less lactose-intolerant, etc. and therefore utterly invalid in showing what the U.S. voting public thinks.

But down the road, let’s assume the Democratic field will winnow itself down to one candidate, and the Republican field will winnow itself down to one candidate, which they both will in about a year. At that point we will know, because it’s been poll-tested hundreds of times, what Bill’s 10,000 responders think of these candidates. There will be very few wide swings in how they poll head-to-head, in part because the polls will be taken collectively over many many different points in the past few months and in part because Bill’s responders are, overwhelmingly, the same exact people in June that they were in April.

Let’s assume for a second that in Bill’s final polls leading up to Election Day, Elizabeth Warren is leading Donald J. Trump by a ridiculous margin—let’s say, 76% to 24% head to head. As a predictor of the election results, Bill’s poll is utterly without meaning, but when those results come in, we can allow for exactly how biased Bill’s polling is. If Warren beats him in the popular vote by 52% to 48%, we would know that’s Bill’s responders were biased towards Warren by 24%, and we would then know how to correct for that bias in future polling on political issues where party affiliation is meaningful. Bill would presumably have a stable population with a well-known bias, and we would no longer have to take wild guesses as to how much they are biased and in which direction.  If, after the election, Trump were to challenge the results (which he will, of course, unless he wins), Bill could ask "How many think Trump actually won the election?" or "How many want Warren to be arrested for election fraud?" and know how to adjust his 10,000 Twitter responders to get an almost-instant response to how the nation as a whole feels about this, or any other, politically charged issue.

Again, I don’t have the slightest clue that this is Bill’s ultimate aim in all this poll-taking. Going on my previous decades of trying to think along with Bill, I’d guess that if I were right, or even close to right, this would be the second or third time I’ve achieved that lofty goal. And I certainly haven’t got a shred of confirmation, from Bill or anyone else, that this is what he’s doing with his polls. I’m out on a limb here, and with this article, I’ve succeeded in sawing halfway through it.

BUT: this is a hell of a lot of trouble that Bill is putting himself through with this series of Twitter polls, and it doesn’t mean a thing unless he follows through on it for almost two years. It messes up his "Articles" page, IMO, for one thing. 9 out of the 15 articles currently on the first page are poll-results articles, which presents difficulty for those of us who could care less about Bill’s political polls, and I’ll estimate that it represents a sizable chunk of Bill’s daily effort to keep good records of the polls, the trends, the patterns, the anomalies, the adjustments, the news that might explain the current poll results, etc. This is not a casual decision, in other words, made simply out of mild curiosity as to the leanings of his Twitter followers.  

What Bill realized, I think, is that he has a valuable resource here, a sizable number of Twitter followers, with one major drawback, an unaccountable bias on every conceivable issue, but I think he’s figured out a way to account for that bias. In head-to-head "Who’s greater?" questions about baseball, he now is getting the tools to validate his poll results without worry about recency bias, and in head-to-head political questions he now is eliminating the partisan bias that his followers bring to political issues.

If I’m right, or even close to right, I hope I’m not giving away the game prematurely. I don’t think I am—this just seems like WAY too much effort to expend casually. I don’t think I could be screwing things up for him, in that BJOL subscribers can’t constitute much more than a blip of his Twitter followers, and I’m making this article inaccessible to non-BJOL subscribers. And even if Bill’s reaction to this piece is "What the hell is that lunatic raving about this time?" that wouldn’t invalidate entirely the possible uses of his polls I’m discussing here. At the very least, I hope he runs some more head-to-head polls of MLB players whose chief difference is the period in which they played, which would help us quantify the recency bias  among Bill’s Twitter following.

 

LATE NOTE: I just read a question posted somewhere online (forget where—sorry) asking who was greater, Joe DiMaggio or Derek Jeter, and it wasn’t even close. Jeter kicked DiMaggio’s ass around the block.  I don’t agree with that response, but that’s not important: the important part is who else could Jeter beat in a head-to-head poll of baseball fans. Gehrig? Ruth? Even coming in a distant second to Ruth (or Musial, or someone of that stature) is interesting to me: what kind of people vote for the player they have seen and rooted for (and admired, adored, worshipped at his shrine, etc.) over a clearly better player? In other words, is recency bias just egotism? "I saw him, therefore he’s better than someone I didn’t see." I’d been assuming that recency bias was simply an error of perception, a lack of perspective on history, but maybe not. Maybe it’s a branch of a mild pathology.

I think there is a valid basis for voting for Jeter over DiMaggio, btw: length of career, fan-friendliness, the beauty of the women each one dated and married, and most of all, the difference of the quality of the opposition each one played against, in the 1930s vs. those of the 2000s. I just don’t think all that adds up to an overwhelmingly clear victory for Jeter, but maybe that’s just me.

 

 

 

 
 

COMMENTS (3 Comments, most recent shown first)

Steven Goldleaf
Not for nothing. but so far most of Bill's HoF polling seems to result in the most recent player winning, the second most recent coming in second, etc. It's just a few polls so far, and I haven't looked very carefully at the 1-4 rankings in all of them, but that's my impression.
1:00 PM Jun 26th
 
Steven Goldleaf
Matt—I agree that recency bias is more extreme in young people than in older people, but how much worse is it, and what are the metrics in old people? We don’t know. We don’t know nothing, do we? That’s what we could learn.

It certainly doesn’t stop with older people, does it? I have an image of DiMaggio in my head, but I missed seeing him play by a decade, so I certainly have no image for his less-famous teammates. Do I give Joe Gordon and Bobby Brown and Johnny Lindell their full due, the same that I give to players I actually saw? Probably not. Couldn’t pick them out of a police lineup.

There’s got be SOME recency bias there, maybe as bad as it is with younger folks, only pushed back a few decades. At some point, the 1910s, say, it all becomes equal—there’s no reason you or I should be more knowledgeable than a younger person about players from a half-century before we were born. What could it possibly matter that we missed them by 50 years but someone else missed them by 70 or 80 years? After a point, it all becomes “the distant, unknowable past,” doesn’t it?

6:01 AM Jun 26th
 
MattGoodrich
If recency bias exists, I would think it's in young people. Us old people are the opposite. Everything was better in the past, but especially baseball players. Joe DiMaggio was in a timeless song. Who puts "Derek Jeter" in a song?
3:41 AM Jun 26th
 
 
©2021 Be Jolly, Inc. All Rights Reserved.|Powered by Sports Info Solutions|Terms & Conditions|Privacy Policy