Picture it. You’re the manager of a Major League Baseball team. Through eight innings, your starting pitcher has had a great game. Unfortunately, your offense hasn’t been able to put as much distance as you’d like between your team and the opposition. You’re winning, but not by much. The batters due up in the ninth inning have already faced your pitcher two or three times today. Your closer is available for use in the bullpen. Who do you send out to face the opposing batters in the ninth inning: your starter or your closer?

A situation like this arose in Game Five of the National League Division Series between the Philadelphia Phillies and St. Louis Cardinals. Chris Carpenter had a shutout going through eight innings and was brought back in the ninth inning to face Chase Utley, Hunter Pence and Ryan Howard. Considering that the St. Louis Cardinals are in the World Series, anyone who doesn’t remember what happened can likely form an accurate guess as to what happened.

Tony La Russa’s decision to stick with Carpenter seems to be the type of thing at which statistically inclined baseball analysts would shake their collective heads. Surely, a manager is better off setting up a series of match ups involving a fresh version of the team’s best reliever rather than a pitcher who has already been seen multiple times by the opposing hitters. After all, a quality reliever once through the order must be better than a starter the fourth time through the order. Right?

Well, it seems that even those who carry around an abacus can jump to conclusions occasionally. Looking at the actual numbers of starters versus relievers in the ninth inning shows us that the decision isn’t nearly as cut and dry as anyone would have predicted.

After initially criticizing LaRussa’s Game Five tactics, Mitchel Lichtman, a co-author with Tom Tango of The Book, compared elite relievers to elite starters when pitching the sixth to ninth innings. Strangely enough, when the top ten pitchers from 2007-2010 went late into a game, their eighth inning performance was close to what their overall numbers would lead one to expect, but their ninth inning, despite facing a similar pool of batters, is decidedly, almost 40 opposition wOBA points better (and more than 20 opposition wOBA points better than an elite reliever), causing Lichtman to conclude:

So unless I am missing something, it looks to me like there is something special about a starter pitching the 9th, at least an elite starter.  It is looking more and more like their stuff is “on” that day and/or their pitch count is low, or something else that the manager and pitching coach can identify. So we don’t seem to see the 3rd (or 4th) time through the order penalty that we expect.

Max Marchi, writing for The Hardball Times expands the sample a little bit, selecting all the games in the past 20 years in which the starting pitcher completed eight innings while giving up only a single run at most. For leverage purposes, he then dropped any game in which the pitcher’s team provided him with more than three runs. His findings looked like this:

  • No runs allowed: 76% of relievers; 74% of starters.
  • One run allowed: 14% of relievers; 16% of starters.
  • Two runs allowed: 7% of relievers; 5% of starters.
  • Three runs allowed: 2% of relievers; 3% of starters.
  • Four or more runs allowed: 0% of relievers; 1% of starters.

In other words, 90% of starters and relievers allow one run or less in tight, low scoring games, as defined by Marchi. The outcomes of these games makes it near impossible for observers to criticize a manager no matter what he decides in such an important situation.

However, unsatisfied with his original findings, Lichtman expanded his sample to include all pitchers and relievers from 1993 to 2010, and considered the differences between close games and not so close games.

If we look at aggregate data for starters and relievers pitching the 9th, without controlling for the score differential (and park factors), we get the illusion that starters pitch much better than expected and much better than relievers, relative to their overall true talent, only because starters tend to pitch when their team is ahead by 2 or more runs and relievers tend to pitch when the game is close or their team is losing. When the pitching team is ahead by 2 runs or more, for some reason, the batting team’s wOBA is substantially lower, for starters and relievers, than when the game is close or the batting team is winning.

He comes to the conclusion that a reliever with a lower wOBA against than the starter is a better option even before you consider the leverage in which the appearance is occurring. Lichtman theorizes that “putting in a reliever who is 10 to 20 points in wOBA better than the starter can save a team from .5% to 1% in win expectancy.”

So, it’s clear now, put a reliever in for the ninth inning. Right?

Not so fast. Lichtman talks about the 10 to 20 point gap in wOBA, but on average the gap between relievers and starters is seven points. Going by his own scale, that would mean a 0.35% win expectancy sway. Even the most stringent haters of intangibles would call that minuscule, and something that could be as easily caused by something that isn’t necessarily quantifiable.

What I think Lichtman is actually describing here is a situation where a mediocre starter is being pulled in favour of a better than mediocre or a dominant reliever. However, the original situation that started this whole discussion was the case of an above average starter being pulled for an above average reliever.

Like many things, what’s right and what’s wrong isn’t nearly as clear as any side of the debate would lead us to believe. Several times on this blog we’ve discussed the role of the manager and how difficult it is to properly judge his impact outside of when he plays the percentages or doesn’t. When numbers don’t indicate a very clear advantage one way or the other, perhaps it’s one of those rare times that we have to trust in the manager’s expertise to a degree.

Perhaps in this situation, with the numbers not guiding us as they usually do, we can look at the outcome of his decision, and, as hard as it may be, say that Tony LaRussa made the right call. I feel like I need a Silkwood style shower now.