Nate Silver's 538 blog has a clear exposition of why advanced baseball statistics state that 20-year-old rookie phenom Mike Trout of the California Angels should win the American League Most Valuable Player award instead of the favorite, Miguel Cabrera of the Detroit Tigers, who is the first ballplayer since Yaz in 1967 to win the Triple Crown of batting average, homers, and RBIs.
Sabermetricians often preen themselves when sportswriters give the MVP award to the wrong guy because there have been some dumb votes in the past.
But it's not like nobody would have noticed that Trout was having an amazing season without advanced statistics. Trout absolutely electrified fans who watched him. He had all sorts of crazy feats like stealing four home runs from batters by leaping to catch balls over the fence. Cabrera, in contrast, had certain unglamorous deficiencies, such as a tendency to ground into double plays because he's a slow runner who hits the ball hard, which dedicated Tiger fans would notice.
It's not that hard to tell who's the most valuable player on a team if you watch the team every day. As Yogi Berra said, "You can observe a lot just by watching." But, when trying to pick the most valuable player in the entire league, sportswriters tended in the past to look to simplifying statistics, most notoriously the highly contextual runs-batted-in counter.
Still, picking between Trout and Cabrera is a complicated question. The difficulty raises questions about everybody's favorite panacea for fixing the schools: value-added testing of teachers.
As I've pointed out before, although the world has largely come around to the theory I propounded in the 1990s that teachers should be measured upon "value added,", almost nobody, including me, has much specific knowledge of how that ought to work in practice. And we're probably not going to get there all that soon, either.
For baseball statistics to reach the level of sophistication in Silver's posting has taken quite a few generations.
And, yet, there are still baseball conundrums that aren't clearly resolvable with statistics. For example, Silver deducts runs from Cabrera's overall performance for being a lousy third baseman. But Cabrera's fans write in to point out that he is a natural first baseman who volunteered during the offseason to lose weight and take lots of grounder so he could play third this season so the Tigers could sign the even fatter Prince Fielder to play first.
Playing a defensive position you are not cut out for can take a psychological toll on your hitting -- an old time example is Pedro Guerrero's slump in the first two months of 1985 due to the Dodgers insistence that he continue to humiliate himself at 3rd base. Finally, when they let him move to his natural left field, he responded to his liberation with one of the great hitting months in history, setting a record for most homers in one month.
(I have a general theory that sabermetric logic points baseball in an ugly direction: kind of like semi-pro slowpitch softball. Don't worry much about defense or baserunning, just get a bunch of hulks who can hit homers or get walks. But I don't think this kind of reductionism works in practice for a psychological reason: you put too many Dr. Strangegloves on the field at once, it cuts the heart out of a team. Sure, statistics might say that fielding isn't really that important, but botching too many balls embarrasses teams and causes bad feelings in the dugout. Errors depress teams, like they depressed Guerrero.)
Also, it's not clear how to account for the fact that Mike Trout spent the Angels' first 22 games in the minors (during which they went only 8-14). As soon as he came up he jumpstarted the Angels' offense.
It's easy to say that if he'd played the whole season instead of just 139 games, he would have scored even more than 129 runs and hit even more than 30 homers.
No doubt about it.
On the other hand, he likely would have regressed somewhat more toward the mean in on-base and slugging percentages -- after a superhuman start, his averages were slipping during the last two months of the season.
I suspect that it will turn out that there are all sorts of analogies to advanced teacher evaluations. If you, say, offer a $10,000 bonus for the school's Most Valuable Teacher based on value-added test scores, you will start to see all sorts of complex but not uninteresting arguments from interested parties for why more factors need to be taken into consideration.
While we have an extraordinary number of baseball statistics fans, we have almost no education statistics fans. So, we're still a long ways away from figuring out teacher statistics.
Yes, a long way. For instance, how would you like to be the 4th grade teacher who inherits a bunch of kids from a 3rd grade teacher who won the award for most value added the year before?
ReplyDeleteI have a general theory that sabermetric logic points baseball in an ugly direction: kind of like semi-pro slowpitch softball. Don't worry much about defense or baserunning, just get a bunch of hulks who can hit homers or get walks.
ReplyDeleteThat may have been true twelve years ago, when "Moneyball" was written, but it certainly isn't anymore. Billy Beanne himself in the last few years seems to have made defense a key to his team-building plans, evidently determining that it is the more modern market inefficiency (him having popularized and over exposed on-base percentage around the time of "Moneyball", of course).
In fact, stat-heads have come up with increasingly subtler and more accurate ways to work defense and baserunning into their formulas. That's part of the reason why statheads are pretty much united in saying that Trout had the superior year in 2012 over Carbera - and it wasn't close (the metric at the other end of that link (fangraphs' measure of Wins Over Replacement) is, I think, the top single measure of player value accepted in the stathead world - and, you may notice, Cabrera wasn't especially close to second most valuable player in the AL last year - mostly on account of weak baserunning and really weak fielding).
(also - being a passable first baseman is no more valuable than being a poor third baseman. First base is the easiest position on a baseball field - most third basemen could, with some training, play a decent first, but very few dedicated first basemen could ever play a decent third).
As I've pointed out before, although the world has largely come around to the theory I propounded in the 1990s that teachers should be measured upon "value added,", almost nobody, including me, has much specific knowledge of how that ought to work in practice. And we're probably not going to get there all that soon, either.
ReplyDeleteThe main problem is that teachers (within the range commonly found in the US) don't actually make a lot of difference. So it is hard to tell the good ones from the bad ones given all the noise. And trying to so can create all sorts of bad incentives and other problems.
... If you, say, offer a $10,000 bonus for the school's Most Valuable Teacher based on value-added test scores, you will start to see all sorts of complex but not uninteresting arguments from interested parties for why more factors need to be taken into consideration. ...
You would also see a lot of cheating if the scores are based on tests the teachers administer themselves (as is usual).
It should be easier to get rid of the very worst teachers but in general teachers aren't the problem.
I think you view that sabermetric logic drives baseball towards slow pitch softball is pretty outdated. If you confine "moneyball" to the idea that walks/power were underrated (and the negative value of strikeouts overrated -- 2000-2007/8), I'd say you're right. As defensive metrics have improved, teams have placed far more value on run prevention, especially as slugging has diminished, presumably through a combo of tighter drug testing, strike zone expansion and changes in the baseball.
ReplyDeleteThe whole Cabrera vs Trout discussion has stats guys and traditionalists in opposite roles. A decade ago, we'd have had stats-oriented guys pointing out that Cabrera was a better hitter and that defense was too hard to measure, while the scouts/traditionalists would be noting how Trout was five tool,all-around great player. The difference today is that the old school guys are pointing to old school stats and ignoring any new school stats (even the ones that would confirm Trout's greatness as an all-around player). It seems ridiculous, but a good part of the resistance to Trout's candidacy is based on spite of the saber guy's. Perhaps not a surprise given their sanctimonious attitudes.
I don't understand why a hard-core sabermetician would even CARE who won the MVP. Isn't the point of sabermetrics that the numbers speak for themselves? If the numbers say Trout was better than Miggy, then that's that, right?
ReplyDeleteAwards like MVP, CYA, ROY, and HoF membership all involve voting and therefore assume an element of subjectivity. Sabermeticians should want no part of that.
" While we have an extraordinary number of baseball statistics fans, we have almost no education statistics fans. So, we're still a long ways away from figuring out teacher statistics"
ReplyDeleteIt's not so much that we don't have education stat fans as that we have a league staffed by Hal Chases and Chick Gandils supervised by Arnold Rothsteins. When fans feel the fix is in they pay scant attention to stats. A timely example from the great state of Michigan
http://watchdogwire.com/michigan/2012/11/14/failing-school-has-every-teacher-and-principal-rated-highly-effective/
A list of similar perversions could be extended to any desired length.
Guerrero played more games at 1b, 3b, and RF than LF. He had played 8 career games in LF when he was moved there in 85 (242 in RF at that point, for example).
ReplyDeleteI'm guessing this example is one of those things SABRE people point out is post-event rationalization nonsense.
of course the media helped silver's predictions come true
ReplyDeleteYour theory about where sabermetrics is pointing baseball is out of date. The first meaningful contribution of sabermetrics was pointing out how valuable walks are and that they were undervalued. Thus teams stack their rosters with guys with high OBP because they could pay them less and still get good results. This is no longer the case and guys who draw walks are now payed closer to their values. The next wave of sabermetrics had to do with defensive and baserunning stats and teams started to invest more in great athletes.
ReplyDeleteThis is why the sabermetrically inclined favor Trout over Cabrera, even though Cabrera is clearly a slightly better hitter. Trout is much better at defense and baserunning.
It reminds me of how old timers used to say Dimaggio was better than Williams even though Williams had better stats.
Should be Andrew McCutchen.
ReplyDeleteSpeaking of Pedro Guerrero, he had the greatest sports quote ever.
ReplyDeleteInt: "Pedro, what do you think about out there at 3rd base?"
PG: "First, I pray they don't hit the ball to me. Second, I pray they don't hit it to Sax."
I'm rooting for MIKE TROUT! Go California Angels!
ReplyDeleteGames are far simpler to model and forecast statistically than real world activity, as the rules limit the scope of possible actions, AND baseball is by far the simplest of all games to model statistically, since it involves a straightforward repeated action between two individuals (trying to throw a ball past a batter) and not much complex interaction. So you should expect baseball statistics to be far more helpful and accurate than statistics in other areas of life and even other games. (Compare basketball and football statistics -- aggregate stats tell you little, the most sophisticated practitioners use complex mapping of specific actions and interactions on the playing field). Bottom line: if there are any problems left in baseball stats after 20 years, be very afraid of resting too much on education stats.
ReplyDeleteSince we can't seem to measure teacher performance, how do we even know that teacher performance matters much? Maybe any idiot can teach if the classroom has smart students, but nobody can teach if the classroom doesn't have smart students.
ReplyDelete"I'm rooting for MIKE TROUT! Go California Angels!"
ReplyDeletehttp://www.youtube.com/watch?v=AdSYhwLENvM&feature=related
But didn't Cabrera play 3B earlier in his career? I think he only started playing 1B recently. His inexperience at 1B came up during the controversy over the last play of Armando Galarraga's almost perfect game. Cabrera awkwardly went for a ball well to his right instead of letting the second baseman get it. Galarraga had to rush to cover first, where there was a close play and the ump blew the call.
ReplyDeleteSpeaking of Nate Silver, Was Nate Silver the Most Accurate 2012 Election Pundit? Lots to chew on here.
ReplyDeleteOh, and Trout is easily the AL MVP. His huge defensive and baserunning advantages make up for any comparative loss in SLG.
"Your theory about where sabermetrics is pointing baseball is out of date."
ReplyDeleteBut that's just a slight mid-course adjustment after PED testing started to crack down on the basic discovery of sabermetrics: muscleheads rule in baseball when all you care about is wins.
Look at the Tigers, for example. Why did 240 pound Cabrera switch to 3rd? So they could sign 275 pound Prince Fielder to play 1st.
Why couldn't Fielder DH?
They had 240 pound Delmon Young at DH.
This is not an elegant team, but it's an effective one.
"...muscleheads rule in baseball when all you care about is wins."
ReplyDeleteMeanwhile the Giants have won 2 World Series with a bunch of guys who extend at-bats and hit line-drives. The homerun derby style of the American League has now lost 4 out of the last 5 years.
Also, how do you measure value-added in teachers when there is no equivalent to "wins" in education? For certain students, the goal is merely to graduate with baseline math and English skills and no criminal record. For other students, often in the same school, the goal is to get accepted to Harvard.
In evaluating baseball players, or any zero-sum game participants, you start with the objective (to win) and work down from there. In education, there is no single metric to work down from.
While reading Steve's and others' profuse commentary on that ridiculous web site, my mind wandered and I actually started reading something not about HBD regressions and steroid statistics... There was an old article in the sidebar about Wycliffe A. Hill, the man who actually tried to create a script-writing robot. Talk about your Moneyball At The Movies...!
ReplyDelete>Cabrera's fans write in to point out that he is a natural first baseman who volunteered during the offseason to lose weight and take lots of grounder so he could play third this season so the Tigers could sign the even fatter Prince Fielder to play first.
ReplyDeleteUmm, in Cabrera's first 5 years in the majors with Florda Marlins--before arriving at Detroit--he never played 1B, while playing 363 games at 3B. Whatever the story/reason for Cabrera playing 1B in Detroit, and subsequently moving to 3B--Cabrera's fans' explanation ain't it.