clock menu more-arrow no yes mobile

Filed under:

How to fail spectacularly at building a bullpen

Trying to create a role-based bullpen is usually a recipe for disaster.

David Manning-USA TODAY Sports

For years, fans have been asking the question, "How is it that Dave Dombrowski can pull off wizard-level trades on a regular basis, but can't seem to put together a decent bullpen?" Some have answered that question by saying that he just doesn't try, or treats the bullpen as an afterthought. A quick review of the 2014 season should quickly put that notion to rest.

A man who spends some $20 million for two years to lock up one of the (at the time) best relievers in baseball is not a man who just doesn't care.

A man who trades away his top two pitching prospects mid-season to acquire the (at the time) best reliever on the market is not a man who is dealing in afterthoughts.

And yet, neither the Joe Nathan nor the Joakim Soria acquisitions really worked out. So what's the problem? Does Dombrowski just not know how to put together a bullpen?

Precisely. But the problem is not with Dombrowski, per se, it's a problem with the way the baseball world thinks.

Several changes over the past few decades have conspired to create a mindset that makes building a good bullpen more of a matter of luck than a matter of skill.

One change was the creation of the "save" statistic. Another change was the amount of attention being paid to pitch counts. Even well after the save statistic was introduced in the late 1960's, managers pushed their starters deep into games, rendering bullpen depth unnecessary.

Sparky Anderson won a World Championship in 1984 behind three starters who threw a combined 667 innings, and two relievers who threw 278 innings. (Yes, he had a fourth starter and a handful of lesser relievers, but this is the basic sketch.) Sparky would frequently solve the problem of short starts by just calling on one of his relievers to come in and pitch two or sometimes three innings in relief.

When managers started watching pitch counts more closely, both starters and relievers began throwing fewer innings, which meant that bullpens needed more depth, and (apparently) more clearly defined roles. A good bullpen, however, doesn't need roles so much as it needs crisis pitchers — a very different concept from that of a "closer." A crisis pitcher enters the game in crisis situations, no matter what inning that might be, whereas a closer is very much tied to the ninth inning.

The following chart shows a random selection of some of the best relievers in baseball from both "yesterday and today." Notice that guys like Rollie Fingers, Goose Gossage, and Dan Quisenberry appeared almost as much in the eighth inning as they did in the ninth, while more recent relievers are almost exclusively ninth inning specialists:

Relief Pitchers by 8th and 9th innings

Much of the baseball world has bought into a flawed premise: that the ninth inning is the most important inning in the game. It might be, depending on the score and who is due up, but it might also be the easiest inning in the game. In fact, one study found that teams with a lead going into the ninth win upwards of 85 percent of the time. It almost doesn't matter who pitches that final inning.

Yet managers today identify their best reliever, apply the "closer" label, assign him to pitch the ninth, then see their strategy validated as he accumulates a bunch of saves.

Filling the role of a ninth inning specialist can mask a pitcher's weaknesses, however, precisely because closers are often given the relatively easy task of entering the game with a clean slate, only needing to record three outs.

Joakim Soria looked like a dominant pitcher when he was closing games for Texas earlier in 2014, but perhaps part of that dominance was a reflection of the fact that in 33 innings, he only had to deal with one inherited runner. When he came to Detroit, he was used in slightly higher-leverage situations, given six inherited runners to deal with in only 11 innings, and under those conditions, he allowed 33 percent of his runners to score.

Many fans were frustrated with Brad Ausmus' bullpen management, which was almost entirely role-based and not flexible enough to adapt to new information. Joba Chamberlain was labeled as "the eighth inning guy," and did a great job in that role prior to the All Star Game. In the second half, however, his performance deteriorated rapidly, but Ausmus continued to reflexively push the "Joba Button" once the scoreboard rolled over to the magic number eight.

The same can be said of Ausmus' use of Joe Nathan, continually using him in the ninth inning despite his repeated struggles there, or of the way he under-utilized one of his best pitchers in Al Alburquerque simply because "when we use Albie, it's earlier in the game, sixth inning, occasionally seventh inning."

However, it must be said that Dombrowski uses the exact same philosophy when building the bullpen, and this is part of why his bullpens have been shaky. He looks for individuals to fill roles — a closer, a set-up man, etc. — instead of creating a flexible group of crisis pitchers.

(Real-life example: Dombrowski filled a role by signing Joe Nathan; meanwhile, "eighth-inning guy" Joaquin Benoit wrapped up his season in San Diego with a 1.49 ERA, a 0.773 WHIP, a K/9 rate of 10.6, and a BB/9 rate of 2.3)

The inherent risk is that relievers are volatile, and very few of them can sustain lengthy consecutive seasons of dominance. Building a role-based bullpen puts all the eggs in one basket, so to speak. If your closer can't close, then what? If your eighth inning guy stops pitching well in his assigned inning, where do you go from there?

This is why building a bullpen these days seems to be a matter of luck. If your role-pitchers all magically pitch well at the same time, you're golden. Even Ned Yost has a hard time screwing it up. If they don't, however, then you're stuck in the position Dombrowski is all too familiar with: facing the reality of a weak bullpen year after year, and searching for help at the trade deadline.