May 10, 2006
So, it turns out that all those John Hughes movies that made me think of beautiful L.A. days were actually set in a fictional suburb of Chicago. I had no idea. I mean, would it have killed John to put someone in earmuffs at some point, or at Soldier Field (Bueller aside) or talking about the power of mini-Ditkas? I genuinely thought those movies--well, most of them, anyway--were set in L.A. and environs. E-Sheehan (lots).
One of the ongoing debates in sabermetrics is how to best deploy your top reliever. Most analysts dislike the strict restriction of a team's top reliever to save situations or tie games at home when a save is not possible. There are too many high-leverage situations that don't fall into that narrow band--and too many low-leverage situations that do--to make that the optimal usage pattern, but the idea that the ninth-inning is all-important has taken hold within the game.
This notion leads to some completely illogical choices, a particular subset of which I want to address here today. I'm not going to argue for radical changes to how teams assign relief innings, for a return to the ace-reliever model of the 1970-1985 period. All I want to see is a little thought going into the process, a little deviation from the established norm.
Assume, for the moment, that teams choose their closer and their set-up man in a manner that funnels the best pitcher to the former and the second-best to the latter. In practice, teams often don't align their talent this way, usually because they overvalue service time or previous closing experience. It is fair to say that they intend to have their best pitcher on the mound in save situations. We'll go with that for now.
What the rigid assignment of roles often does is create opportunities for the other team. Nearly every day, a manager uses his second-best (or worse) reliever to face the middle of the opposing lineup, while holding back his best reliever to face the bottom of the order. Joe Torre is probably going to the Hall of Fame, and his use of Mariano Rivera in the postseason has shown that he understands leverage. But last week in Arlington, he nearly gave the Rangers a game by allowing Aaron Small (gee, who saw this coming?) and Kyle Farnsworth to pitch through the top and middle of the Ranger order.
It's the last decision on which I want to focus. With the Yankees holding an 8-3 lead, two on and one out in the eighth, Torre went to the mound to relieve Small and bring in Farnsworth, with the #2 spot in the Rangers' order due up. Farnsworth would face three batters and leave a two-out, bases-loaded situation--now a save opportunity--to Rivera. Rivera eventually escaped with a lead and hung on in the ninth to preserve an 8-7 win.
The problem is the decision to use Farnsworth rather than Rivera against the Rangers' best hitters, Michael Young and Mark Teixeira. Allowing them to face an inferior pitcher, even in an 8-3 game, gives them a chance to further the rally. Rivera is being paid twice what Farnsworth is; shouldn't he be facing the best hitters late in the game, rather than the bottom of the lineup?
Phil Garner provides an even better example. Twice in three days last week, Garner used Dan Wheeler to pitch against the other team's best hitters in the eighth, then brought in Brad Lidge for three relatively easy outs in the ninth. Lidge may not be off to a great start, but he's still the Astros' closer, and Garner wasn't making this decision based on the two pitchers' recent performances. He was only considering the score and the inning, and by not looking at the opposing hitters, he aligned his talent in an incongruous fashion.
About nine months ago, I looked at this phenomenon in a more general fashion. I found that the difference in quality of batters faced between set-up men and closers was small; however, Tom Fontaine reported that the top and middle of the order were more likely to bat in the eighth inning, while the bottom of the order tended to bat in the ninth. The closer-centric bullpen means that the better reliever is going to face the worse hitters a significant portion of the time.
On a macro level, these decisions wash out a bit. On a micro level, when you're using your second-best reliever to get out Albert Pujols and Jim Edmonds, while using your best on the likes of Yadier Molina and Aaron Miles, you're costing yourself. Even if you think the ninth inning is somehow harder than the other eight--a dubious, but popular, perception--can you really argue that it's harder than the difference between 1000-OPS guys and 700-OPS guys? At what point do the soft factors yield to the hard ones?
This isn't about turning Lidge or Rivera or Billy Wagner into a Carter-era ace reliever who pitches 110 innings and often comes in as early as the seventh inning. It's just about thinking a little more deeply about who you want on the mound in the game's most important at-bats, and when those at-bats occur. Even in a framework of reasonably rigid roles, would it be that difficult to swap the innings of the set-up man and reliever based solely on where the other team's lineup falls? For example, instead of getting Wheeler up in the top of the eighth inning last week in St. Louis, so that he could face Pujols and company, Garner would get Lidge up. Same idea--he pitches one inning--but now he's facing the best hitters. Wheeler would pitch the ninth. It's actually a small change; you're asking the same work from your pitchers, just reversing the order in which they appear.
There has to be more to relief usage, even within a framework of fairly limited roles, than score and inning. There has to be more than mapping usage to a scoring rule that does little more than pump up one number on one guy's stat sheet.
Aligning the bullpen in the above manner makes sense, but one reason it won't happen is that relievers now get paid based on one number. It would help if "saves" were the next statistic to fall out of public favor. Over the past 25 years, the practitioners of performance analysis and the guys like me who stand on their shoulders have done a very good job of reducing the importance placed on batting average, on RBI and even on pitcher wins. However, rarely did those statistics ever drive the usage of players the way the save statistic has. In a span of 20 seasons, the save went from a way of properly crediting an important contributor to team success--the reliever who pitched well in preserving a close victory--to a set of criteria for using the team's best reliever. There can be no question that teams would not have arrived at this particular model for relief usage without the development of the rule.
The closer mindset--this whole mythology about the importance and the difficulty and the personal qualities required to pitch the ninth inning--is a farce. There were ninth innings for a hundred years before Bruce Sutter, and no one had any problems with letting the guy who had pitched the seventh and eighth preserve a one-run lead in the ninth. Check out some of the boxscores from the 1970s and early 1980s sometime. Again, I'm not advocating that kind of usage, but it's important to realize that the closer myth isn't even as old as the DH is.
What's particularly amusing is that there's a lot of commonality among the people who propagate the closer myth and those who would argue that starting pitchers are babied nowadays. It's not starters--who pitch in a completely different sport than their predecessors did 40 years ago--who have been coddled, it's relievers. Everyone has to have a "role" and know their role and pitch only in their role with sufficient warning and god forbid they have to warm up twice in a week without getting into the game. Workloads are way down, and working conditions are way up. I can give you a lot of good reasons why no one throws 15 complete games a year anymore, but I can't give you one good one why Trevor Hoffman is used the way he's used.