A tale of two roles

While there are many relievers who were starters in a past life, most make the move because they couldn’t cut it in the rotation. Maybe they had only one or two quality pitches, maybe they tired too quickly, or maybe they just weren’t very good and the team had invested a lot of money in them. For whatever reason, the pitcher was deemed a failure as a starter and was shipped off to the bullpen—with hopes of being reborn as a reliever.

Much smaller is the number of effective starters who are then moved to the bullpen—usually to closer. This seems to happen for one of two reasons. Either the team has a lot of depth in the rotation but no one to close (Brett Myers in 2007), or because injury concerns disqualify the pitcher from starting every day (John Smoltz in 2001).

The two most famous transitions, and arguably the most effective, are Dennis Eckersley and Smoltz. Both were above-average starters for at least a decade before moving into the bullpen and finding success there as well.

Smoltz made the move due to injury problems. He was coming off major elbow surgery that caused him to miss the entire 2000 season. Not knowing whether his arm could hold up under the strain of the full workload of a starter, the Braves moved him to the closer role once he was healthy. He thrived in that role, saving 10 games down the stretch in 2001, and became the full-time closer for the next three seasons.

Injury concerns also prompted his return to the rotation in 2005. Smoltz believed he’d hold up better with the predictable workload of a starter rather than the erratic usage of a closer. He successfully completed the transition back to starter, throwing more than 200 innings in each season between 2005 and 2007. However, injuries have crippled him the last few seasons and he’s working his way back in the Red Sox organization.

The reasons for Eckersley’s transitions are less clear. His ERA had shot up in 1986 while with the Cubs, from 3.08 in 1985, to 4.57. In the mid-1980s, 4.57 was good for only an 88 ERA+. Many were concerned he was done. Tony LaRussa thought differently, however. He decided to move Eckersley to the bullpen, where the experiment surpassed everyone’s wildest expectations.

In 1988, LaRussa went even further, and began limiting Eckersley’s innings. Eckersley began to pitch only when he had the chance for a save; and generally only one inning at a time. Today, this is common usage for closers, but at the time it was revolutionary. And Eckersley made it pay off big time, with five consecutive tremendous seasons, highlighted for statheads by his 606 ERA+ in 1990, and for other fans by his Cy Young/MVP season in 1992.

So these moves were obviously successful, but were they the right moves? Should the Braves and the Athletics have kept our duo in the rotations (assuming, of course, they could have held up to the wear and tear)?

Let’s look at how each performed in the two different roles. We’re going to focus on the defense independent pitching stats because we’re interested in understanding the pitcher’s value, not the rest of the team’s.

Both Eckersley and Smoltz performed much better as relievers than as closers. That’s a fairly normal phenomenon. In The Book, Tom Tango, Mitchel Lichtman and Andy Dolphin found that the average pitcher improved his ERA by about 0.8 runs when pitching in the bullpen, which is right about where both Eckersley and Smoltz ended up.

Looking at these tables, it seems obvious that making these pitchers relievers was the right decision. But I’ve left off one very important column.

Pitcher value is composed of two major components; performance and durability. You could have the best pitcher in the world—someone who’s completely unhittable—but if he could throw only one inning a year he wouldn’t be very valuable (well, maybe if you were the Mets). So we need to consider how many innings Smoltz and Eckersley pitched in each role before we get a good understanding of how valuable they were.

Let’s take the same tables as before, but add a column showing each pitcher’s average innings pitched per season.

Those are some pretty big drops from starting to relieving. Eckersley saw his workload decrease by 140 innings a season, while Smoltz was pitching 120 fewer innings per year.

How do we compare these situations? On the one hand, we have lesser performance in a lot of innings. On the other, we have phenomenal results, but in a handful of innings. There are lots of different ways we can go in trying to make sense of these results. Some are quick and dirty (multiplying ERA+ by innings pitched, for example). Others are more involved (determining the statistics of a hypothetical pitcher that would make the two performances equally valuable). I’m going to use an overall measure called Wins Above Replacement (WAR).

The WAR concept has been around for awhile in different incarnations. Basically, it measures how much more valuable a given player is than a replacement the team could pick up at any time (often called freely available talent). If you’re interested in a couple of different ways to calculate WAR, check out the links in the references section. I’m going to use Sean Smith’s historical WAR database for my comparisons.

Let’s look at a new table. This one just measures the WAR value for our pitchers in each of their roles.

Even though Smoltz and Eckersley were more valuable on a per-inning basis as closers, the small number of innings they pitched out of the bullpen really takes away from their overall value.

Some graphs put together by Erik Manning at Beyond the Box Score illustrate this as well.

Both pitchers had many seasons as starters better than their best relief season and it’s simply due to the number of innings they pitched.

Did Bobby Cox and Tony LaRussa make the right decisions in moving their starters into the closer role? It depends. Neither pitcher was anywhere near as valuable while closing as he was in the rotation. But as we touched on above, perhaps neither pitcher could survive in the rotation.

With Smoltz, we have three seasons after he returned to his starter role that suggest he was up to the task of starting. He probably should have returned to the rotation earlier—perhaps in 2002. Eckersley is harder to figure. After 1987, he never threw more than 80 innings in a season. But when he was moved to the bullpen he was coming off a season in which he racked up 201 innings pitched. To the casual observer, he looked to be on the downswing as his ERA jumped. A closer look at his fielding independent stats suggest that he may have been unlucky in 1986, and perhaps could have survived as a starter for at least a few more seasons.

It’s hard to argue with the success both Smoltz and Eckersley had as closers. It’s certainly the reason Eckersley was elected to the Hall of Fame in his first showing on the ballot. But it’s also hard to argue that their respective teams wouldn’t have been better off keeping them in the rotation and garnering the same level of performance each had proved capable of. The A’s would have been roughly two wins better per season (1.2 for the seasons when Eckersley was at his best as a closer), and the Braves 1.5 wins.

Now it might not have mattered that much, or worked out that way, had history gone differently, but the evidence suggests moving quality starters to the bullpen is a bad idea. So all the Yankees fans and writers who want to move Joba Chamberlain to the bullpen may want to note how even the most successful of those transitions turned out for the teams.

References & Resources
WAR data courtesy of Sean Smith’s Historical WAR database. I can’t say enough about how great a resource this is.

John Smoltz and Dennis Eckersley WAR graphs courtesy of Erik Manning at Beyond the Box Score.

For more information on various ways to calculate Wins Above Replacement (WAR), check out the following series:

• From Colin Wyers here at The Hardball Times:
  How to measure a player’s value
  Part 1, Part 2, Part 3
• From David Cameron at Fangraphs:
  Win Values Explained
  Part 1, Part 2, Part 3, Part 4, Part 5, Part 6, Part 7