Friday, October 7, 2011

Q Day 6: Perry's Eggheads

Matt Glassman asks:
First: Any opinion on the future trajectory of political science work influencing practical politics. It seems like there is quite a buzz right now, but I have no sense of whether it is short-lived fad or possible SABRmetric style revolution.
Second: what are your thoughts on pitchers winning MVP awards?
The first question is (I assume) about Rick Perry's so-called Eggheads and the research into (and helpful for) electioneering. I don't know that I have much of an opinion here, but I'll speculate anyway, I suppose. My sense is that baseball should have been abut the best-case scenario for a sabermetric revolution: almost everything is measurable, and the goals are very clear -- and the gap between what people believed and what basic, simple sabermetric analysis knew was large. How does electioneering compare? It's true that elections results are measurable in votes, which helps, and that the goal of "more votes" is clear. But my impression -- and I know a lot more about the baseball situation -- is that the gap isn't as large in politics. Well, let's put it another way: my guess is that the gains from systematic analysis are probably less substantively important. It's a guess, though.

On the other hand, there's probably a lot less hostility from political practitioners towards political scientists than there was from baseball people to analysts, for a variety of reasons.

If you move beyond electioneering, I'm not sure that I see a lot of scope for game-changers. I do hope that the increased public presence of political scientists combined with the openness of folks such as Ezra Klein, Matt Yglesias, Andrew Sullivan, Greg Sargent, Jonathan Cohn, and others to what people at the Monkey Cage and the rest of us are saying makes me optimistic about the future of political reporting. But I'm not sure how that changes practitioners, exactly.

On the second question...if I was voting, I'd treat pitchers and non-pitchers the same way. They're eligible by rule, and I've never seen any reason to not include them based on whatever value they've had during the season.


  1. I think you may be underestimating the hostility towards political scientists from practitioners.

  2. When Bill James was hanging out his shingle back in the 80's, he made the rounds of major league cities, and let off a few stink bombs, to roust the masses and create buzz and swell his bank account.

    He came to Detroit to let off one of his stink bombs. According to this idiot James, the Tigers were a better team with John Wockenfuss at catcher, as opposed to the sometimes All Star Lance Parrish. His proof? The Tigers played about .500 with Parrish behind the plate, but about .800 with Wockenfuss in the lineup.

    Pretty straight up, right? I mean, winning percentage rules, doesn't it? Proof's in the pudding, heck, all us hardheaded realists know that, eh?

    Except, you have to remember, James is a moneygrubbing capitalist pig, same as you and I. He wanted to maximize his cash returns, and would twist the data to accomplish that, his real goal, and ignore what was really going on in order to do so.

    That Tigers ballclub was stacked with left handed hitters. Parrish was at times the ONLY right handed hitter in the lineup. It was a complete management failure, obviously, but that failure necessitated Parrish being in the lineup constantly, because the opposition was stacking their pitching staffs to pitch lefthanders against the Tigers' lefties. Teams were calling up lefthanders from the minors just to pitch a single game against the Tigers... that's how imbalanced things had got. If I checked, I bet there may have been a Single A pitcher or 2 who started against the Tigers.

    So on that RARE night when the Tigers faced a righthander, and Parrish was given a well deserved night off on his season long slog, the Tigers lefthanded bats FEASTED on that right hander, and pounded him to dust, filling up the short porch at Tiger Stadium... and giving that idiot James the "proof" he was using to make his bogus case. No prisoners on those nights, and the offspeed hitting righthanded Wockenfuss even helped, dumping a few curveballs into right field.

    But let's make no mistake, the Tigers would have been DOOMED if Wockenfuss had played every day. He was inferior behind the plate, and could never have been the power threat Parrish was against the lefthanders that were shutting down the Tigers lefthanded hitting predominance so regularly, leading them to the substandard performance on those nights Parrish was playing.

    That's where this sabermetrician nonsense breaks down... when the numbers geeks forget that baseball is about more than putting cash in numbers geeks' pockets... it's about doing what it takes to win... and that hasn't changed in the many decades I've been watching the game.

    So remember this, when you start thinking sabermetrician geeks are gonna get involved in politics and "improve" it. That's just laughable. Every one of these geeks has an agenda... and you better figure out what THAT is before you start worshipping at their shrine. I can tell you flat out what James' agenda was... he proved it for all in his first steps on the stage. You don't clamor into town tossing stink bombs, without getting called out for it.

  3. And of course pitchers should be eligible for the MVP. Verlander has been the perfect ace this season, and broke a number of perilous Tigers' losing streaks, just as your horse MVP is supposed to do. He took the mound and personally set their tired bullpen right on those nights, when they'd been shelled in games just previous. There were a a 1/2 dozen starts minimum that he both won and set them up for wins following, by righting their tired bullpen situation. That's what stud pitchers do, even if they only take the mound once every 5 days... they influence it all, before and after their starts. Managers know they can take a chance and exhaust their bullpen in games before their starts, and attempt to pull out those wins, because they know that stud is gonna be there to pull their irons out of the fire, when he takes the hill. He's going to find a way to go the distance, and make them "smart managers".

    Is he the league MVP? He better be top 5, that's all I can say.

  4. phat,

    Heh. On your first comment, I'm sure there is some hostility, but (1) political scientists are a *lot* better credentialed than Bill James other others were, and (2) there's a much stronger overlap in background. Doesn't knock out all of the hostility by any means, but it shouldn't, and I'd say isn't, as deep.

  5. FWIW, it seems to me there could be many similarities between Perry's Eggheads and the sabermetrics crowd. Both are/will be based on principles that are true. Both are/will be based on things the traditionalists hadn't noticed. Both are/will be advocated by folks (nerds) with a wee bit of baggage wrt the traditionalists (never played the game/never won the election). Importantly, both are/will become cults whose fundamental value will be overstated by adherents (perhaps in part due to the baggage mentioned just above).

    As I see it, there are two basic flaws with sabermetrics, particularly noticeable when the games count the most. The first is that sabermetrics deals with averages, when what we often care about - especially at money time - is variance.

    As an example, in last night's classic Cards/Phils game, the Cards got runners to first and second with none out in the eighth. John Jay bunted the runners over, which is a - the - "cardinal" sin in sabermetrics. The reason is that the mean number of runs in an inning with a sacrificed out is always less than in an inning without.

    But. When its late innings, with your entire season on the line, against a hurler at the height of his Cy-Young powers, playing with a slim lead, you're not looking to put up a 5-spot. A run is silver and gold there. Getting runners to second and third with one away and Pujols, Berkman and Holiday due up, you have to like your chances of getting that coveted run. Again, it isn't generic averages you care about in that situation, its the probability of putting up any run at all.

    The second flaw in sabermetrics is that it embraces disciplined hitters, who approach their craft with much more complexity than their free-swinging "see ball, hit ball" brethren. During an 11-game swing against inferior teams in late June, the more complex approach of the plate discipline crowd should, generally, translate into more success.

    But, again. When the season is on the line, when everything comes down to a game, or an inning, or an at-bat, would you rather have Scott Hattieberg up there, with his complicated approach to a plate appearance, or would you like a guy who sits dead red on a fastball, and is talented enough to adjust to an off-speed pitch? Might Hattieberg's more complex approach break down under pressure?

    As this crowd no doubt knows, Billy Beane has often poo-poohed his sabermetric teams' failings in the pressure of the postseason due to random bad luck in a short series. Well, maybe, Billy. Interesting, though, if we define "non-sabermetric Hall of Famer" as a guy who strikes out a lot more often than he takes walks, there are, AFAICT, two modern Hall of Fame/great hitters who struck out two times for every walk (the ultimate sin to Beane/Epstein and the sabermetric types):

    Roberto Clemente and Kirby Puckett.

    What's different about Clemente/Puckett and every single hitter for the A's (or, this September, the Red Sox)? Clemente/Puckett were the consummate clutch hitters under pressure. Apparently, Beane's excuses notwithstanding, an excellent "see ball, hit ball" type of guy holds up under withering pressure much better than the complicated hitting approach of players Beane or Epstein loves. Funny how that works.

  6. (continued)

    So Perry's Eggheads will probably do some groundbreaking research that shows that 6.7 yard signs per square kilometer of sympathetic territory is correlated with electoral success, and perhaps some Epstein or Beane type in that campaign will focus disproportionate effort on ensuring that exactly 6.7 yard signs are put up. What a shame when it turns out their guy still ends up looking like a rank amateur in the klieglights, and the vaunted 6.7 principle turns out to be not nearly as solid as its cult adherents believe.

    In summary, I disagree with the conclusion that the Eggheads will have the same impact on politics that the sabermetric crowd had on baseball. This is due to the interests of the money behind the activity, baseball or politics.

    The money in baseball is made (lost) by owners, many of whom are happy to turn any sort of profit at all (they usually don't), even if they don't win. Baseball holds the national sporting consciousness during the dog days of summer, the period when sabermetrics should work best. If sabermetric hitters routinely break down in the pressure cooker of a pennant race, that's not as big a problem to the owners, since by then their sport has been left behind by casual viewers, typically in favor of football. (Note also that a large percentage of dedicated baseball fans, those who stick with the sport after football starts, buy the sabermetric stuff, which should also encourage owners to play along).

    By significant contrast, someone who has given a ton of money to the Perry campaign, and is currently feeling a lot of heartburn over their guy's terrible efforts on the public stage, should feel no comfort at all that some nerd has a data point saying that 6.7 yard signs per square kilometer has worked in the past.

    The money won't be motivated to play along with the Egghead/political science stat it probably won't have the same impact it does in baseball.

  7. Thanks for that, CSH. Very thoughtful.

    And I agree with you, that the "sitting dead red on a fastball" hitter is the guy you'd often want up at crunch time. Baseball is about mistakes, making them and NOT making them. The dead red fastball guy's eyes light up when that ball gets hung out over the plate. He was gonna be swinging at it wherever it was, but if it's suddenly up in his wheelhouse, there'll be nothing left of that pitch but a jetstream.

    That pitching mistake's results get amplified, in other words, because of the hitter's approach, as you imply when you speak of "variance". We get a spike in performance here, and that's what we're looking for in the clutch. And we set ourselves up to take advantage of the pitcher's performance valley... the mistake. It's the 2 in combination that lead to the end result, not just an individual's one-off analysis. It's more dynamic than static number crunching could ever model.

    Same in politics, no doubt, as you also imply. Electioneering might be able to take advantage of the sabermetric geeks, as Mr. Bernstein mentioned, although you have to watch it here, as guys like Silver twist up the data and have an agenda, much like James did in the case I mentioned above. That's where campaigns talk about "internal polls"... the stuff with true validity and that we plebes may never see.

    I think the root of this confusion is that we're foolishly referencing something as "political science". Politics ain't science. Political arts... maybe. Political history... absolutely. But referencing this as a "science" is the height of foolishness. Sorry if that ruffles any feathers, but I'd bet most political "scientists" couldn't science their way through junior high physics. And a semester of Statistics 101 and familiarity with Excel data charting ain't science... and that's basically all the sabermetrics guys et al are doing.

  8. CSH,

    Sabermetrics is not anti-bunting in all situations; in particular, the bunt with men on 1st and 2nd is (depending) usually a good move, IIRC.

    Sabermetrics also has no particular preference for different types of hitters. In the 1970s and 1980s, "disciplined" hitters tended to be strongly undervalued by baseball people, so analysts often pushed them as a generally undervalued type -- but it wasn't the only group of undervalued players, and it wasn't because analysts "preferred" that type of hitters.

    But at any rate, here's regular season/postseason OBP+SLG for the three players you mentioned:

    Clemente: 834/803
    Puckett: 837/897
    Hatteberg: 772/839

    The theory doesn't in this case pan out; Hatteberg and Puckett were better in the postseason, while Clemente was worse. Hey, it's not a definitive study; I can think of lots of reasons not to accept it at face value. Similarly, assuming anon's story about James and Wockenfuss is true (source? couldn't find it anywhere), it's entirely irrelevant to whether sabermetric analysis in general is useful or not. Of course James made mistakes! So what? That has zilch to do with whether analytic approaches to baseball knowledge are worthwhile or not.

  9. My source is my memory, Mr. Bernstein. And while that source is not always reliable, in the memorable case of a hardworking Detroit Tiger, name of Lance Parrish, who had his name besmirched by some hack, and I jumped all over it at the time, I'd say the source is pretty good. If you have access to Detroit News or Detroit Free Press archives, you may be able to access some stories. It was a bit of buzz here. I wish I could recall Sparky Anderson's comments as a quote, but I really can't. He generally scoffed at or ignored stuff, as you may know, but I have no specific recall.

    I'm not saying the saber geeks are useless, they're not. I'm saying data validation is the most critical step in the process. And the worst thing you can get into is the infamous "data swamp". You have everything, but you really have nothing, because you're either unable or unwilling to apply proper validation and judgement to the swamp of data collected, wherein it's truly implementation ready. It's just a swamp. That's where the saber geeks can leave you, if you're not careful. James did that with Parrish's case. He didn't use any real judgement. I say he was out to throw stink bombs and create buzz and cash income. YMMV.

    The most important quality you can have in a productive hitter is "eye of the tiger". The geeks can't quantify it, but now that I've referenced the term, I'm guessing as a baseball guy you know exactly what it is I'm talking about. Clemente and Puckett had it in (unmeasurable) volume, as CSH mentions (and we can mention the pitching era Clemente hit in, a point I'm sure you knew was coming, to explain why his numbers were seemingly deficient in some way).

    The geeks can't measure EotT, but you and I can, with our own eyes.

    Delmon Young sat out last night with injury. You think the Tigers wanted to play Raburn in his place? I don't. I bet if I checked with the sabers, they'd point me to some stuff saying those 2 guys are close. I don't buy it, no matter what the numbers say. Young has the EotT, and Raburn clearly doesn't. (Now watch, Raburn will probably homer twice tonight. ;-)

    Hey, I'm an engineer. I love data. But I've been around data analysis long enough to understand the critical importance of data validation and proper judgement in its use. And then there's the ever present agenda issues. I mentioned Silver only because he's clearly presenting data that tingles lefty legs, and he was even attempting to tingle them into believing the House wasn't gonna flip last November, right up 'til election day. And why wouldn't he? Like the James/Parrish issue, he's only executing his agenda of cash flow. Proper lefty data analysis? I go with Bernstein, Sabato and my boy Charlie. Historically, I know they spend time on proper data validation and judgement, no matter if they tilt left a bit.

    The baseball saber geeks work fine for rotisserie baseball stuff, but it's harder to transfer it to comprehensive analysis of what happens on the field.

    Years ago, Pistons GM Jack McCloskey, an NBA talent guy if there ever was one, used to gather the wise heads around the table to discuss prospects, and at the end of the discussion, he'd have everybody close their notebooks and he'd ask one question: "Can this guy play in the NBA?" And they'd have a go around the table and each guy had to man up and answer it, not with numbers and stats, but with reasoned analysis, based on what they saw in the guy. Trader Jack said that was one of the most clarifying and important parts of their decision-making process. I think it always is, and should be.

  10. Jonathan,

    Thanks for taking up my argument. I realized after I posted that I should have looked up Hatteberg (I used him because of his role in Moneyball), and you're certainly right that his OPS is better in the postseason.

    Just wanted to add an aside: one of the best things to come out of Sabermetrics, IMHO, is the widespread adoption of the OPS metric. I tend to think that adds a lot of value, with Puckett's 1991 WS a classic illustration. Puckett hit .250 in that WS (pretty mediocre), with an OPS of .950 (very good). We remember Puckett as quite good in that series, which OPS captures much better than BA.

    For me, there are basically two distinct challenges to winning a WS: being better than opponents in the amorphous slog of the lengthy regular season, and being better than your opponent when it counts in the clutch. Sabermetrics has added a lot of value on the first one (mastering the amorphous slog), but it is pretty weak on the second (the clutch), at least in my uneducated opinion, relying somewhat on Beane's famous dismissal that there is anything unique about playing in October vs. playing in July - when he argued that the postseason was basically random.

    Returning to Hatteberg, it seems to me he is a pretty decent illustration of this. Hatteberg had one outstanding postseason series (2002 LDS, in which his 103-win A's team lost to the inferior Twins), 2 series in which he was quite bad (2003 LDS and 1998 LDS), and 2 in which he hardly played (1999 LDS and 1999 LCS).

    If the object of the postseason is to find a way to win a short series, then Hatteberg's teams only did that once in five tries, a series in which Hatteberg had a measly one plate appearance (1999 LDS). Hatteberg's team lost the other four, including the one other one he hardly played, two in which he was bad, and one where he was outstanding, but his also outstanding team lost nonetheless (to inferior opposition).

    So is Hatteberg a good postseason player? By blended stats, yes. But here's the rub: if you had to pick guys from history to be on a team that had to win one game, and you could only choose guys with a postseason OPS 50 points higher than regular, and there was a fair, golf-like handicap system adjusting for baseline differences...

    ...are ya still picking Scott Hatteberg? I'm not, and I honestly don't think that's just because I'm pushing this argument, though I could be deceiving myself.

    Oh, and as a PS, one of the fun aspects of this discussion is that I also went back and looked up Clemente's post-season stats, where I stumbled upon a delightful factoid which had somehow slipped my attention all these years. I'll leave it as a tease since anyone interested can easily look it up:

    Who was the winning pitcher in Game 7 of the 1960 World Series?

  11. OK, but who really thinks that Scott Hatteberg was a great player? The questions is whether you would rather have Hatteberg than a similar player with a different "shape" of production...or, indeed, the real question in the 1970s and 1980s was whether you would want, say, Mickey Tettleton vs. someone with an empty .285 BA.

    BTW, for those who want to know what it was really all about, see the the great David Grabiner's Sabermetric Manifesto.

  12. One of the first laws of statistical inference is that all sampled data comes from a homogeneous population. And its true, of course, that a game in the postseason is the same as one in June based on 54 outs, and the bats/gloves/balls, etc. But that's not the aspiration of sabermetrics; they're trying to tell us how to win. At that level, a postseason/late season and early season game couldn't be more different.

    Some of it is personnel; pitchers in particular are used differently in the postseason. A postseason series is also usually a bit longer than a regular season series. Mainly, though, the intensity of the pressure is vastly different, which even the most casual fan is well aware of, so one wonders how this critical difference slipped the attention of the sabermetric stat experts.

    What's different about pressure? A lot of things, but one of the main ones is to simplify as stress levels rise. The Tettleton type (poor average, good power, lots of walks) is not a particularly simple approach to hitting; it follows that under the pressure of the playoffs a Tettleton should lose a lot of effectiveness. He only made it to the postseason once, but what a disaster that was.

    I just did a quick look at baseballreference. com of the top 100 OPS guys, trying to find the Tettleton types, that is "poor BA, good power, decent walks". Kinda hard to tell over there, but there are three obvious ones: Killebrew at #84, Thome at #17 and McGwire at #10.

    Killebrew's brief postseason career did have a higher OPS than his regular season totals. Thome was bad in his postseason effort; McGwire was even worse after the regular season ended.


  13. But the point is that you could tell a story the other way, too; you could say that the Tettleton approach is a smarter one, and that it'll hold up better against good pitching and high stress. Or it could be that for the players the postseason isn't actually higher stress; actual high stress is when your job is on the line (and if so, it's possible that those who can't handle it are weeded out long before anyone gets close to the postseason).

    The point is that these are all hypotheses that can be tested -- but that as soon as you look at it that way, as hypotheses that can be tested, the you're in the realm of sabermetrics. Believe it or not, pre-James that mostly (and there were always exceptions) just wasn't done. Baseball people just knew certain things (such as that minor league performance was meaningless for predicting major league performance -- yes, they really did believe that), and that was that.

    Sabermetrics is not about particular players or even particular statistics. It's about, as James said, the "search for objective knowledge about baseball." That has plenty of limitations, both in what properly comes under its scope and how much can actually be learned, but that's what it is.

  14. Well, it might be an interesting conversation in light of sabermetrics to talk about performance under pressure. Interesting, if the sabermetrics crowd wanted to, assuming they didn't (a la Grabiner and Beane) implicitly reject such ideas via the notion that 162+ games are objectively similar enough that we can safely study them as an amorphous mass.

    As a general rule, its my understanding that for a non-expert type (e.g. Hatteberg or Tettleton), more complex systems break down under pressure more easily than simpler ones. We could argue about this, but who's the most expert "complex plate discipline" guy in history? Teddy Ballgame? He only made one World Series, and in seven games he was worse than anyone else we've discussed, notably terrible in games 6 and 7, which his Sox entered with a 3-2 lead and left with a Series loss.

    When you suggested that players are not as concerned about the pressure of the playoffs as they are their paycheck, you were talking to a sympathetic audience in me, as I am a selfish conservative bastard who tends not to buy that "I only want to win a championship!" tripe.

    But then I remembered Posnanski's recent excellent reminiscence about the 25th anniversary of the crazy 1986 sports year, and how while Buckner has always handled himself with class, its surely clear to everyone that one error is different for that guy than all others, even though he had a very good, and very lucrative, career.

    And as bad as that was, a week earlier: Donnie Moore and Dave Henderson. If you find yourself tempted to think that players cynically care about their contracts more than winning, stop by and check out Moore's WHIP from '85 to '88: 1985: 1.087, top ten in Cy Young AND MVP; 1986: a still quite-good 1.128; 1987: a near-career-worst 1.538; 1988: a career-worst (full season) 1.697; 1989: took his own life.

    RIP, Donnie Moore. We're pretty certain that something happened to Donnie Moore in 1986, and we have a pretty good idea what it was, and that it had nothing at all to do with his contract.

  15. Just to finish on a lighter note, a while ago I mentioned that the Jay bunt in the 8th inning against the Phils made sense because "2nd and 3rd, one out, Pujols, Berkman and Holiday due up", you have to like your chances of scoring a coveted run.

    Predictably, Roy Halladay walked Pujols, loading the bases for Berkman. Berkman is a great hitter, he's 5th among active players in OPS, and 19th all-time. But he looked awful in that at-bat, he struck out on a couple of filthy breaking balls in the dirt; on the third strike he looked like one of those new immigrant students from a non-baseball country taking his first cuts in the grad student summer softball league.

    In my imaginary dissociative world, Bill James suddenly appeared on my couch for that at-bat, and when Berkman went down, James chortled "Ha! That didn't even look like Lance Berkman!"

    I got thoughtful. I agree, I said, but then how can we study an at-bat like that together with his other 587 PA in the regular season, which did look like Lance Berkman?

    Imaginary Bill James grew very stern and silent. Then he awkwardly gave me the finger and disappeared again.

  16. Mr. Bernstein,

    I had a post in here, portion of which followed up on your request for source on the James/Parrish incident. What happened to that post?

  17. Spam filter got it. No idea why (although it does sometimes filter for length). Should be restored now.


Note: Only a member of this blog may post a comment.

Who links to my website?