December 10, 2012
Do Closers Age Differently Than Other Relievers?
It must be tough being a closer. Most nights that you pitch, you're going into a tight situation. If you do your job, well... you just did your job. If you mess up, the next day, the papers will talk about what a horrible human being you are. My father told me when I was growing up that a thousand "attaboys" is worth one "uh oh." He was right.
I've never been a major-league closer, but I'd have to imagine that it's a nerve-wracking experience. Most of the time, when we think of the pressure placed on closers, we think about the effects of the situation right there in the moment. Is the enormity of having the whole game riding on your right arm (because lefties can't be closers—sorry, Zachary Levine) too much to bear? There's evidence that closers aren't indifferent to the situation, but what about a different type of pressure effect: the effect of constantly being under pressure over the course of time.
People who are often under stress, even low levels of stress, display a response to that stress over time. The body is made with a fight-or-flight response to handle stress, but turn those circuits on too much, and they start to burn out. You can blame the hormone cortisol for that. Eventually, the bloodstream keeps a large supply of cortisol on hand, and that can activate adrenaline glands (what it's supposed to do in a crisis) even when you don't want it to. I got to wondering whether closers show this effect, at least in that their performance suffers for it.
As always, if you are afraid of numbers, please replace the next part with "Russell picked up his magic 8-ball and asked the question of whether closers age differently. The 8-ball rolled around and said...
Warning! Gory Mathematical Details Ahead!
I selected all pitcher-seasons in which the pitcher debuted after 1993, threw more than 50 innings, and made 75 percent or more of his appearances as a reliever. I defined a closer as a reliever who either notched 75 percent of his team's saves or saved more than 30 games in a season (the latter to control for pitchers who were traded mid-season and got 15 here and 20 there).
For each season, I calculated each pitcher's age on April 1 (Opening Day), and his K%, BB%, HBP%, and HR% for each year. I looked at the raw change (15% in 2011, 16% in 2012 equals a delta of +1.0%) in each stat in consecutive-year pairs, assuming that he pitched primarily in relief (and had 50 innings to his name) in both years. These deltas were my primary dependent variable. This left me with just shy of 1000 paired seasons.
I ran a mixed-linear model. For the initiated, I used an AR(1) covariance matrix to control for individual level variance (some pitchers don't change as much from year to year) and used a simple fixed-effect model. I entered last year's strikeout (or walk or HBP or HR) rate as a covariate to control for the fact that extreme values in either direction may run up against floor/ceiling effects or be more prone to regression to the mean. I entered age as a factor, rather than as a covariate. This models each step in the age progression as a separate event, rather than as a member of a regressed line. There are problems with both approaches, but experience has taught me that this is the less-bad method. Finally, I entered whether the pitcher was a closer in the season in question (coded 0/1) and whether he had been a closer in the previous season.
In this way, we've controlled for individual variation in performance, age, and previous level of performance. What's left should give us a decent estimate of what the effects of being a closer are. You can read the results that follow as "Given an average reliever, and controlling for his age and previous performance, we would expect X out of him."
The answer was that when it came to strikeouts, closers actually gained more/lost less than their non-closing brethren. New (or re-instated) closers showed deltas that were about two percentage points higher than non-closers, meaning that a closer is more likely to strike out two extra hitters per 100 faced. Closers who were reprising their role from the previous year gave back about half of that, but were still ahead of the game from other relievers.
Walk rates were a little less dramatic. New closers were ahead (that is they had lower predicted walk rates) of non-closers by about half a percentage point, while returning closers were about two-tenths of a percentage ahead. Changes in HBP and HR rates were also relatively small, on the order of a couple tenths of a percent, although in the good direction as well for both.
If anything, closers age better than their non-closing counterparts.
What it Means
Except maybe not.
Someone out there hopefully took a good research methodology class and can see a giant flaw in that assumption. There is evidence that some relief pitchers age differently than others, and that being a closer is a good marker of this difference. But, where is the causality here? Is there something about being a closer that makes pitchers better? Maybe the stress and strain is actually protective in some way?
A pitcher who time and time again rises to the occasion in the face of adversity learns valuable life lessons that help him to be there for his team for a long time even when others might fail. That's grit. (Hey, who let that sportswriter into my basement?)
Maybe we've got something else at work here. Closers are not chosen at random. In fact, they are chosen specifically because they are good pitchers. Leaving aside the usual arguments about the inefficiency of the modern closer role, managers do usually try to pick their best reliever for the job, ideally one who's not going to fall apart very quickly. Maybe these results are an indicator that managers are simply good at picking that guy out. They are baseball subject matter experts, after all. We know that for hitters, there are different developmental trajectories, with hitters who debut earlier peaking later. Perhaps there is a similar effect for relievers. The good ones get to be closers, and they also follow a nicer career trajectory.
That said, there's something to be said for investigations of aging patterns (for all players) beyond those that simply look at age, and knowing that the pitcher is a closer is a piece of information that has predictive power and should be incorporated into models. The question is where that effect comes from.
So now that I've convinced you that the idea of there being a psychological drain that comes from closing is silly, perhaps I might argue that the effect might still be there. We don't have a lot of examples of good relievers who aren't forced into the role of closer, so we don't have a good control group to study. Maybe these closers would have been even better if relieved of the difficult duties of the ninth inning (which of course, would be a waste of their talents...) Then again, like a lot of narratives that are projected into the heads of relievers (and baseball players in general) we start thinking about the pressure that's involved only when he runs out of the bullpen. The closer himself can construct his life around knowing that he will appear in the ninth inning of a close game, and prepare himself accordingly.
What we have here is a simple finding and a complex and ultimately muddled explanation that doesn't resolve neatly into a nice little package. According to my 8-ball, "Outlook is hazy."