THT's Toolbox

PITCHf/x Archive
Updated Marcel projections
Win Shares
PrOPS
All-time best sluggers
WPAB calculator

All content on this site (including text, graphs, and any other original works), unless otherwise noted, is licensed under a Creative Commons License.

Print this Article

E-mail this Article

The Case for a Quicker Hook

by Jeff Sackmann
May 25, 2007

Whenever a manager has a conspicuously slow hook, especially if he's been burned a few times by stretching out his starters, I always wonder why. After all, most teams these days have a seven-man bullpen, often including at least one guy who can go multiple innings. With that kind of backup, just how many innings does the starter need to throw?

Of course, there's more to it than that: You can't randomly mix and match relief innings. But if a manager was intelligent about it, tried to spread around the work and used his best relievers mainly in the highest-leverage situations, he could probably get more innings from the guys who end up sitting for a week (think the closer in a week of losses, or the mop-up man in a week of wins).

This is a much easier conversation to have with some actual numbers involved, so let's give it a try. Using the current arrangement of the Brewers bullpen as an example, here's how many innings I think are reasonable to get from each member of the pen on a per-week basis:

Francisco Cordero (closer), 3-4
Derrick Turnbow (set-up No. 1), 3-4
Matt Wise (set-up No. 2), 3-4
Brian Shouse (LOOGY), 3
Chris Spurling, 3-4
Elmer Dessens (mop-up), 4
Carlos Villanueva (long man), 6
Total, 26-29

A few notes about those numbers. I'm not exactly saying that Dessens and Villanueva should throw 10 innings every week. If the starters do go six or seven innings every night, it wouldn't be necessary. But it's absolutely reasonable to say that they could. Also, for a reference point, three innings per seven games is about 69 innings over the course of a season—a very easy total for a reliever to reach. Four innings per seven games is 92 innings, which is pretty high, but still reasonable if you think in terms of roster spots rather than pitchers. Few relievers make it to 90+ innings these days, but few relievers are healthy and on a major league roster for 162 games.

Let's go with the higher number first. If you can get 29 innings from your bullpen in a week, that's a bit more than four innings per game. It's close to five if there's a day off that week. If all of those games are at home and/or all of those games are ties or wins going into the ninth, that means you need exactly five innings every night from your starter. No more. Sometimes, the manager could finish it off by giving the long man the ball for the last four innings. Other times he could go through four guys for one inning each. It's certainly nice to get your starter through the sixth and seventh inning occasionally, but my point is that it's not necessary, at least if your goal is to "save the bullpen."

Objections

Most discussions of optimal bullpen usage are largely a waste of effort when it comes to practical applications. Most managers have a firm mindset about who ought to pitch when. If there are no save situations in a week, the closer may not pitch. If every game is a close one, the mop-up man may never pitch. In some cases, that might be the wise decision (just ask Brewers fans about how often they'd like to see Dessens in the game); in others, it's a waste of a roster spot.

So, to say that we can simply assign a number of innings to each pitcher on a per-week (or per-month, or whatever) basis is folly. It's interesting from a theoretical perspective, but no matter how convincing the argument, managers are going to try to ride their mid-rotation starters into the seventh inning for the foreseeable future. Then again, we are talking theoretically here. Just because managers aren't going to do it doesn't mean it's wrong.

The better objection, in my mind, is that the guys who would pitch the sixth or seventh inning in place of that starter are not as good. The Brewers are an excellent example: Often, the sixth inning guy has been Spurling or the since-DL'd Greg Aquino. Neither is the sort of guy you want to rush into the ballgame when, say, Jeff Suppan has racked up 84 pitches and is cruising to victory. That's fair (and often correct), but it raises a better counter-argument: If you don't want to go to that pitcher, what's he doing on the roster?

If major league baseball needs 210 relievers, some of them are going to be pretty dreadful. But do managers really need that many insurance policies? Perhaps they do, but if so, maybe teams should be more honest about that role. If someone like Spurling is the worst-case scenario for the sixth inning, why would he ever pitch in a remotely close game? Here we veer into the same arguments that statheads always make about bullpen management. Why use your worst pitcher (or even your third-worst pitcher) in the seventh inning of a game in which you're trailing by one or two runs? If you're not willing to use him as much as you would a slightly better reliever, why use him in an important situation at all?

The manager algorithm

In these types of arguments, we always end up talking about the way in which managers make decisions. They want to avoid second-guessing, so they religiously go to their closer in a save situation. They use their designated set-up guy in the eighth. They use their lefty to get the opponent's toughest lefty out in the seventh or eighth. If their team is tied or trailing by a run or two, they go to the "second team," consisting of everybody else, also in their own sort of hierarchy. This is adjusted occasionally by usage patterns; sometimes a pitcher needs a day off or another pitcher needs to get some work in.

Fine. I'm perfectly happy to accept that managers ought to be absolutely predictable in the moves they make. In any given situation, there's a "best move" that they could make. We might not know what it is at the time, but it exists. Because we have the tools to analyze these situations in real time, it's fair to say that managers can't "win" games, they can only "not lose" games. (This is why most fans think their team's manager is the worst. Pushing buttons and having those moves work out doesn't reflect on his genius, while doing something risky and having it not work out reflects on his stupidity. Exceptions are rare.)

I'm hardly the first person to suggest that this algorithm needs to be tweaked. There are better ways to use a seven-man bullpen; heck, that better way might involve getting rid of two of those guys and calling up a couple of pinch-hitters or platoon partners in their place. That's another debate for another day. But while we're adusting that algorithm, let's think harder about when the bullpen ought to come into play. If the seven-man bullpen is here to stay, that's 25-30 innings a week ready to be deployed. Unless it's Johan Santana on the mound or a relief corps made up of Jose Limas, I suspect managers would be better served turning to the pen earlier and taking advantage of those innings.

Or, perhaps more realistically, they should all get together at the next winter meetings and decide to build that into the algorithm. It'll still be boring, fans will still hate their managers, but the system will be a little bit better.

Jeff Sackmann consults for several major league teams. He runs the websites MinorLeagueSplits.com, CollegeSplits.com, BrewCrewBall.com and contributes to the fantasy magazine Heater. He welcomes comments via e-mail.

Do you have a general question or comment for one of THT's writers? Send it in to our weekly mailbag We also welcome unsolicited op-ed pieces of approximately 500 words for consideration. We reserve the right to edit for length, clarity and consistency of style. Please include your whole name and location to be considered. If you have a comment about this specific article, please email the writer.

The best online source for major league baseball tickets is Ticket City.

Next Article: Rick Ankiel's Comeback>> <<Previous Article: This Annotated Week in Baseball History: May 20-26, 1963

Now shipping

Search THT:

THT's Toolbox

The Case for a Quicker Hook

Objections

The manager algorithm