Those who seek to think clearly about the game of baseball can not only develop good ways of evaluating performance but also figure out how to treat the old, inferior ways.
Stat savants know, for example, that batting average and RBI are worse metrics than OBP and SLG. But they are more than just worse. They are redundant. They belong to a class of redundant stats: completely useless when better and equally simple stats are at hand, they deserve no place whatever in any kind of analysis.
Flawed stats are not necessarily redundant. Stolen bases is a flawed stat when it's deployed, as so often, to show a player's contributions to offense via the steal; stolen base percentage is vital to include because getting caught stealing hurts offensive production. Yet stolen bases is still a useful stat. It's raw data, for one, and together with caught stealing shows the magnitude of the baserunner's addition or subtraction to offense. Similarly, runs allowed per nine innings has long been seen as a flawed depiction of pitching performance because it punishes pitchers for fielding errors that lead to runs. Thus the "earned run" was invented and with it ERA, which exclude such run-causing errors. This was progress -- I suspect ERA is usually truer than RA. Still, runs allowed is worth attending to because better pitchers will record outs after errors instead of getting shelled. And so on: ERA is flawed for reasons that ERA+ addresses, but ERA remains somewhat useful (if only for being simpler). Any stat becomes a "flawed stat" when unjustified conclusions are drawn from it.
A few stats, however, are not only flawed but redundant, so bad they have no reason to exist. The ones that come to my mind happen to be the stats most often used to measure offensive production: batting average and RBI.
Batting average sort of, kind of, sometimes, hints at the thing OBP states precisely: how often a hitter gets on base instead of making outs. Now if walks, which batting average treats as non-events, were wholly the doing of the pitcher without regard to the batter's presence and actions, then batting average would be a fine stat. But walks are obviously the work of hitters too. Most hitters maintain stable walk rates over many years. That's because walking results from certain skills: the ability to tell balls from strikes, the ability to make contact to foul off close pitches, and the ability to hit for power, causing pitchers to nibble. If walks flowed solely from pitchers' mistakes, the difference between OBP and batting average would over time equalize for everyone, so it wouldn't much matter which stat is used. But some hitters consistently have large differentials between OBP and batting average, and others small.
Point is, batting average is redundant. If you want to know on-base ability, OBP is the stat to use. If you want to tell how often contact yields a hit, use batting average on balls hit in play. I can't see any purpose to which batting average should be put -- unless you're the agent of a hitter whose high contact rate produces a batting average more impressive than his OBP.
The redundancy of RBI is more obvious and less interesting. I'm not exactly sure which dimensions of offensive production RBI are supposed to show -- maybe clutchyish hitting plus slugging plus the crucial ability to hit a sac fly and exchange high-fives for making an out. At any rate, all those dimensions can be measured more effectively and still simply, without falling prey to RBI's glaring deficiency, its dependence on teammates getting on base. On the Big Red Machine, I might have had decent RBI totals, though only because I'm short enough to draw the occasional bases-loaded walk and humble enough about my baseball skills to know to kidnap members of the opposing pitcher's family, Jack Bauer style, in order to get pitches to hit. In retrospect, RBI is somewhat problematic as a redundant stat, since it's hard to tell what RBI even purports to signify and therefore which stats should replace it. In that way, stupidity is the RBI's own best defense. I hate the RBI.
Are there other redundant stats? I hereby put this question to Primarily Baseball's horde(s) of fans.* One important quality of redundant stats is that they can't be too much simpler to calculate than the stats that replace them. Otherwise very complex stats, like runs created, could be argued to render almost every offensive stat redundant, and furthermore there's value in simplicity.
*According to the OED, one fan can comprise a horde, if he maintains facial hair and habitually walks around wielding a hand-whittled club. So: stop shaving and start whittling. It's probably not a terrible tradeoff if you whittle fast.
P.S. Suzyn Waldman, August 9, 12:20 a.m., announcing a Yankees pitching chage: "And here comes Joe Torre . . ." She corrected herself after the commercial break, saying, "I'm a dope" and "I knew I would do that at least once this year." I agree with the former and share in the latter, but nice recovery, Suzyn.
P.P.S. I can't permit myself to type the name "Suzyn" without forswearing any intent to approve of the spelling.
Subscribe to:
Post Comments (Atom)
4 comments:
Nice job. Your argument for the redundancy of batting average is very good, and the things that would have to be true about walks for batting average to be a good stat. In fact, there was a time early in baseball history when BA was computed as OBP, but they stopped doing that because, in fact, Henry Chadwick thought that the walk was entirely the pitcher's fault! In fact, of course, not only is walking a skill, it is MORE of a skill than hitting for average, in the sense that BB rates show far more year-on-year correlation than batting average.
There is actually another, more precise sense in which batting average is redundant. Let's say you take an average hitting team, one that hits .260/.330/.420. Now, take away a bunch of home runs and walk, and add in a bunch of singles; you can do this so that the OBP and the SLG do not change. Now, the team as a whole might hit .300/.330/420, a high contact but low power and discipline team. Conversely, you could take out a bunch of singles and add in walks and home runs and make the team hit .220/.330/.420, sort of a poor man's Adam Dunn.
What happens when you simulate these line ups? Does the team that hits .300 score a lot more than the team that hits .220? No. They score almost exactly the same number of run. In fact, the .220 team is SLIGHTLY better, but the effect is so small that it is not really worth considering.
I'm not sure that I entirely agree with you about ERA vs RA. As with the OBP vs BA issue, it would be legitimate to switch from RA to ERA if pitchers had no control over the rate at which they give up unearned runs. I haven't seen a study on this, but I think that they do. Look, for instance, at Curt Schilling. Throughout his career, he has displayed a very consistent tendency to give up scant unearned runs. He hasn't pitched on noticeable good fielding teams, and he seems to give up unearned runs at a much lower rate than the other pitchers on his team. Maybe it's luck, but it seems to be a pretty persistent trait. Maybe it's because he's God on Earth and really gets tough when the tough get going to buckle down and pick up his teammates when they are down. Or maybe he has always struck out a ton of people and been an extreme flyball pitcher (there are fewer errors on fly balls). If it is indeed an ability, then ERA and ERA+ would both underrate Curt Schilling.
You obviously should try to account for the quality of the defense for the pitcher, but I am not sure that the RA to ERA adjustment is a very illuminating way of doing so.
The "normal" stat that I hate the most is definitely the save. Not only is it completely redundant--good closers all have conversion %'s of around 90, so the raw save total is overwhelmingly simply a reflection of opportunities--but it is actually pernicious. Because it is not a "save", teams with a one run lead in the bottom of the seventh, with the bases loaded, nobody out, and Barry Bonds coming up, will send out some crappy reliever to give up the inevitable grand slam or bases loaded walk; while a three run lead with nobody out in the bottom of the ninth and Adam Everett, Brad Ausmus, and the pitcher due up will call forth the Dropkick Murphy's, masturbatory gesticulations, and maybe a Riverdance. This is managerial malfeasance of the highest order. I honestly believe that eliminating the save will, given time, have a salutatory effect on bullpen usage.
Good comments. Saves are indeed horrible stats -- and not just horrible but pernicious, since they have directly worsened the way baseball is played, by no less than creating a whole position. Like the RBI, the save is redundant, but redundant because it's stupid, founded on dubious assumptions about actions that constitute and reflect skills. We can't say the same for batting average. Batting average is on its face redundant (though it also features a faulty assumption that walks aren't skills). RBI and saves do give information other stats don't, so they become redundant only if one believes the extra information they give has nothing to do with the quality of performance. As, methinks, one should.
I'm inclined to agree with you about RA versus ERA, though I'm not quite sure how to resolve this question. In the abstract, there appear to be times when unearned runs shouldn't be charged to a pitcher and times when unearned runs should be charged. When ERA was invented, it probably should have excused pitchers only from errors that directly led to runs on the same play. At any rate, the Schilling example is a powerful one. Certain types of pitchers -- like those log high strikeouts and flyballs and low walks -- deserve to be valued more highly than ERA or ERA+ suggests because they don't put their defense in a position to make errors.
Also, shouldn't Dustin Pedroia walk more? You need an HDTV to see any separation between his knees and his belt.
I am developing an increasingly burning hatred for Bill James. The man is incapable of saying anything without irritating the hell out of me.
Have you thought of initiating a dispute with your credit card company over the charge for Bill James Online? Reason: "James purported to be smart. He's not really." Then again, a mere refund doesn't make up for the lost time and energy. Punitive damages of $1,000 for every "catcher pride point" would seem fair.
Post a Comment