World Series time! Enjoy Premium-level access to most features through the end of the Series!
November 5, 2012
In Praise of the Modern Bullpen
The construction of the modern bullpen is silly. It starts with a junk stat (the save) and works backwards from there. There’s an anointed “closer,” his deputy (“the set-up guy”) who pitches the eighth inning, a couple of “match-up” relievers for the seventh inning, and some middle and long-relief guys who suck up innings four through six, as needed. In a close game, the relievers on the team with a lead are generally deployed in the (perceived) reverse order of their effectiveness as the innings unfold, with the apparent aim being to slowly choke off the other team’s chances of winning as the game goes further along. And to record a save.
The folly of this system has been pointed out over the years by many people (including me). There is even some precedent for how a bullpen can be constructed differently. It’s called the 1970s. Calls to get away from the inning-based system in favor of a better understanding of how to properly leverage relievers used to be groundbreaking. Now, if I wanted to write an article on the reasons why teams should dump the closer-set-up-match-up-middle-long model here at Baseball Prospectus, my editors would ask why I was re-hashing old material.
In general, the explanations that have been offered for why not even one team has abandoned the present model for a leverage model have focused on character flaws in the decision makers (read: GMs and managers) of teams. I know, because I’ve made a lot of these arguments. A few of my personal favorites:
· They must have failed math. I got a 5 on my AP Calculus test. And I have a Ph.D. Boom.
There might be some validity to all of the above (I really did get a 5 in AP Calc). It’s just that—like all arguments in which there are two sides—it’s more complicated than that. The other side might be filled with (gulp) reasonable and intelligent people who looked comprehensively at this issue and saw things differently. This week I gave myself a challenge, in the spirit of that election thing that’s happening. I’m going to genuinely imagine myself on the other side of the argument and try to make the case that the closer-set-up-match-up-middle-long model has benefits that, even if they don’t cancel out the advantages of a leverage-based model, make my position of a leverage-based bullpen a little less of a (pardon the mixed metaphor) slam dunk.
Argument #1: People perform better when they have well-defined roles
I honestly don’t know if anyone has done research on this, and I’m happy to have my mind turned by a well-conducted piece of research. (Note to self: place this in off-season queue.) But, before being pitchers, baseball players are human beings, and human beings are cyclical creatures, including the 24-hour cycle of day and night. One nice thing about an inning-based bullpen assignment is that a player generally comes into the game around the same time of day (give or take), and over time, his body will become accustomed to being ready for maximum effort around that time of day.
And then there’s the predictability aspect of it. A little while ago, I looked into the issue of whether closers pitched differently when they had ample time to see that they were about to go out to close a game vs. when their team took a lead late in the game and suddenly, they found themselves in a save situation. It turned out that closers did pitch differently, although I wasn’t able to find much in terms of their performance being affected. Maybe they weren’t worse off for the surprise, but some part of them noticed.
The problem with a bullpen based on leverage is that when a pitcher is called into the game might change from night to night, as possibly would what he’s asked to do (e.g., starting an inning vs. working out of a jam mid-inning). Modern bullpen construction might not be optimal in handling leverage, but in most cases, a pitcher can see when he’s coming into the game half an hour away. Sure, we have a cultural soft spot for the “ace” who yells “In omnia paratus” (a billion bonus points to whoever gets that reference), but that’s not the way most humans work. Why make it harder for a pitcher to do his job?
It’s tempting to look at my findings on “surprise” closing assignments and conclude that “Well then, there wouldn’t be any difference in performance, so we’re keeping the modern arrangement… to spare someone from feeling nervous?” The problem with that is that it’s based on a poor understanding of how the body reacts to episodic, rather than chronic stress. The “surprise” closing assignments that I studied are generally one-off events. They happen, but not very often. Game leverage, on the other hand, changes a lot more quickly than does the inning on the scoreboard.
Using the “relief ace” model, where the closer’s job is to pitch in a critical situation, even if it’s before the ninth inning, it means that he has to be prepared to get the call much earlier in the game. Sure, the manager probably would (and should) set some ground rules for which relievers he would go to in which situations, but the reality is that in a leverage-based bullpen, a closer/relief ace would have more “surprise” assignments handed to him. Plus, the closer/relief ace would have to get mentally involved on a lot more nights. The nice thing about being the ninth-inning guy is that you don’t have to worry until the bottom of the eighth, and by that point, the game may be put out of reach anyway. If the manager says “I’m going to you if it’s close and things get hairy starting in the seventh inning,” it means that the closer has to be mentally ready by the bottom of the sixth until he gets the call. And baseball games can get hairy very quickly.
In a key situation, is it better to use an inferior pitcher (all else equal) who knew this was coming or the superior pitcher (all else equal) who had short notice to prepare? I suppose that’s an empirical question that has not yet been formally answered.
But even beyond tonight’s game, inning-based assignments minimize the amount of time that pitchers have to prepare for an appearance and deal with their own autonomic nervous system responses (that’s the part of you that does “fight or flight”). When that system is activated once in a while, it generally has no long-term effects. But research shows that when people live under constant threat (i.e. those autonomic processes are always turned on), it becomes harder to shut that system off. Long-term, that could shorten the shelf life of a good reliever. In going to a leverage-based bullpen, a team might win the game, but lose long term.
This argument is (like most of my stuff lately) based on conjecture from other fields and is certainly untested on a baseball diamond, but I’d like to think that it’s not unreasonable. It’s a side of the issue that has yet to be addressed by those who argue for a leverage-based bullpen.
Argument #2: The current bullpen model takes advantage of the natural male instinct to compete within a hierarchy.
Poor Jerome Holtzman, who by law must be cited in any sabermetric article relating to bullpen usage. I don’t know that everything that’s wrong about the modern bullpen can be traced back to his creation of the save statistic. It can explain the growth of the modern closer industry, but there are other things that have happened in bullpens over the last 20 years that have nothing to do with saves. Witness the creation of the “set-up man.” Teams now not only obsess over having one reliever who can shut the other team down in the ninth, but hope to find a guy who can do the same in the eighth. The Yankees paid Rafael Soriano many millions of dollars to do this, despite the fact that he would not be in line to get any saves (at least when they signed him).
Whether or not the set-up man is a sound strategy from a game and/or roster construction perspective is debatable. But there’s no denying now that the role exists, and that it is the role to which relievers aspire… except the closer. The set-up man is (don’t use a Star Trek reference, don’t use a Star Trek reference) the Will Riker to the closer Jean-Luc Picard. But he’s better than the match-up guys (Data, Worf, LaForge), who are better than the middle relief guys, who are better than the expendable and interchangeable long man (Ensign Redshirt). And the inning-by-inning bullpen arrangement makes it abundantly clear to everyone who is who. There’s a hierarchy here. How male.
Remember that bullpens are really a group of men, most of whom have been competing their whole lives. It’s how they got here. And now they are all in competition with each other. Ideally, the competition would be entirely professional, and most of the time, it probably is. But if you want to give Type-A personalities an incentive to get better at something, make it a competition. In contra-counter argument, I'd note that there will probably be a hierarchy within a leverage-based system (someone will get most of the really high-leverage situations), but at first, it will take some 'splainin' to the guys in the pen.
Argument #3: Exactly who are the pitchers to fill these roles?
One of the most commonly floated ideas for a leverage-based bullpen is to have the closer/relief ace go multiple innings, even if this means that he won't be available the next day. The argument is that you know you have a high-leverage situation here, and tomorrow, the game might be a 15-2 blowout. Closers nowadays do occasionally go multiple innings, although it's relatively rare. Guys who are being groomed to close are generally done so with the thought that, most nights, they will throw 15 pitches and exit, stage left.
It's easy to say "Well, tell him that he's going to go 30-40 pitches and get the next day off." Sure, you can tell him that, but if there's something that's been proven over and over again with pitchers, it's that you can't just stick a guy in a new role and assume that his performance will be the same. We've seen this in a more extreme form with guys who had been really good relievers (Daniel Bard, Joba Chamberlain) but flopped as starters. When starting, you can't throw every pitch with maximum effort. Plus, you have to think about things like turning over the lineup. For some guys, that takes away too much from their pitches and renders them ineffective.
There probably are players who would excel in locking down two-inning saves. But I would caution against the thought that they are exactly the same pitchers who do well in one-inning bursts. A team's current closer might do just fine in a two-inning role, but I think that people are a little too glib in assuming that this is the case. If a team really wanted to embrace this idea, it would be in a bit of a pickle in filling the role. It's not like other teams develop these sorts of pitchers, so there won't be any with a track record on the free-agent market. They can hope that they have someone on the staff who would do well in the two-inning role, which is a bit of a risk. Or they can grow their own, which will take time. This isn't to say that it can't be done, but it's not as easy as it looks on paper.
Going to a leverage-based bullpen scheme makes sense on paper. But, of course, games aren't played on paper, and paper-based models tend to oversimplify complex problems. I think that there are ways that the issues that I've raised above could be handled. Roles could be defined more in accordance with leverage, but still maintain the inning framework. (Easy one: Please use your best reliever in the ninth or extra innings of a tie game.) There are interventions to help manage chronic stress. Hierarchies will develop in any reasonably structured system, and soon guys would be competing for the spots that are clearly the choicest in the manager's head, although the de-emphasis of the save might be hard to swallow. There would have to be buy-in from everyone in the bullpen. And it might take time to really put this plan into proper action.
There are benefits of the leverage-based bullpen. They are hard to calculate fully, but all else being equal, it would represent an upgrade from the standard inning-based closer-set-up-match-up-middle-long model going now. The problem is that not all else is equal, and it's a long way to go to make it that way. The reason that alternate ideas haven't caught on might involve a little more than just teams being stubborn. There are real start-up barriers to these ideas. Some of them are silly, some can be overcome with time, but hopefully you can see that some of them would make you scratch your head and say "Yeah, I guess that would be a problem."