In Defense of the Three-Man Rotation

I used to be a fan of the four-man rotation in the post-season, but I have changed my mind. One of the key events that altered my opinion was last year when Joe Girardi successfully rode the three-man playoff rotation of C.C. Sabathia, A.J. Burnett, and Andy Pettitte all the way to a World Series title. At the time, I was skeptical that this was the right move, but I was impressed at how well it worked, which is why I am surprised that the Yankees are going with a four-man rotation in the ALCS.

Sometimes bad decisions turn out just fine, so I wasn’t completely convinced that the Yankees did the right thing in 2009. I was soon persuaded that the three-man rotation was the way to go in the post-season after a conducting a study with Sean Forman on the impact of pitches thrown and rest days on performance. In our analysis of games from 1988 — 2009, estimates showed that days of rest had very little impact on performance. On average, every rest day lowered a pitcher’s ERA by about 0.015; however, the estimate was not statistically different from there being no effect. It seems that most of the recovery benefits that pitchers receive occur in the three days of rest between starts. And though we found the impact of pitches thrown was small, one day of rest was worth throwing about two fewer pitches less than average in the previous game. If you want to get improved performance from managing pitcher workloads, fewer pitches is better than more rest. And the broader lesson is that small deviations in pitches thrown and rest days don’t seem to have much effect.

In the case of Sabathia and Hughes, they should be good to go in Games 4 and 5. In Game 1, Sabathia threw 93 pitches, for the season he averaged 105 pitches per start. Twelve pitches fewer than average translates to an improved ERA by 0.08 runs. And in his previous five and ten starts (where we found stronger effects than the previous game) his average pitches thrown was right on his season average.

In Game 2, Hughes threw 88 pitches, for the season he averaged 102 pitches per start. Like Sabathia, he pitched about the same number of pitches in his last five and ten starts, which included two one-inning relief stints on regular starters rest. According to our estimates, the reduction in pitches thrown lowers his expected ERA by approximately 0.10.

While I don’t expect Sabathia or Hughes necessarily to benefit from having a lighter load in Game 1, I believe each pitcher should pitch about like he has all season if they were to go on short rest. Now maybe the strain of the post-season and the quality of opponent puts a little extra strain on each pitch, but the loads these pitchers are bearing hasn’t been exceptionally high.

Of course, there may be factors affecting the Yankees decision that aren’t public, but if things don’t go well for the Yankees tonight, I won’t be surprised to see Girardi hand the ball to Sabathia in Game 4. Even if the Yankees do win, I think Girardi would be wise to move Sabathia up. The Yankee rotation has struggled as a whole lately, but I don’t think more rest offers much help. I think a tired trio of Sabathia, Hughes, and Pettitte is preferable to allowing Burnett to make a start.

5 Responses “In Defense of the Three-Man Rotation”

  1. Ken Houghton says:

    It is necessary for Burnett to pitch Game Four, and to be the long relief option for Game Seven.

    Only then will there be equivalence between Girardi’s and Ron Washington’s managerial skills, letting the game be decided on the field, by the players.

    (My heartfelt desire that the Yankees and their $207MM payroll that is subsidized by my cable bill lose 100+ games every year has absolutely nothing to do with the above. It is only out of Concern for the Game Itself that I suggest Burnett needs to be a key pitcher in this series.)

  2. Marc Schneider says:

    Steve Avery pitched and won Game 4 of the 1995 World Series as the fourth starter. I always thought that having a solid (although Avery had had a bad year)fourth starter pitching on normal rest might have given the Braves an advantage that year as opposed to having Maddux pitch on short rest. But maybe it wouldn’t have made any difference and Bobby Cox got lucky.

  3. Dan says:

    The effect of days of rest is unlikely to be linear. What if you just compare pitchers with 4 days of rest to those with 3 days?

  4. JC says:

    We used a estimation technique designed to capture subtle nonlinearities in the data. It wasn’t just a study of 3 vs. 4 days rest, but to capture the impact of longer rest as well, such as skipping a rotation turn.

Trackbacks/Pingbacks

  1. [...] This post was mentioned on Twitter by J.C. Bradbury, Dan. Dan said: "Every rest day lowered a pitcher’s ERA by ~0.015 but estimate not statistically different from there being no effect." http://bit.ly/aN33ER [...]