August 13, 2002

Doctoring The Numbers

The Five-Man Rotation

by Rany Jazayerli

The five-man rotation is a failure.

I don't mean to be overly dramatic here. I'm not trying to frame 'failure' in a pejorative sense, the way we might describe Tony Muser, or airport security pre-9/11, or Bud Selig's ceaseless efforts to acquire a human soul. I use the term "failure" in a purely literal sense. How else to describe a concept which has not succeeded in accomplishing the precise objective for which it was created?

The five-man rotation was a product of the early 1970s, when an era of free experimentation in our society leeched its way into baseball and a few teams dabbled in the concept, on the premise that it would keep their starting pitchers healthy. Most notable among these were the Dodgers, who made the switch midway through the 1971 season. In 1969, three Dodger pitchers (Claude Osteen, Don Sutton, and Bill Singer) each made 40 or 41 starts, providing just 37 leftover starts for the rest of the staff. By 1972, Sutton, Osteen, and Singer were joined by Al Downing and Tommy John, as all five pitchers started between 25 and 33 games, combining to start 150 of the Dodgers' 155 games. In retrospect, it's easy to understand why the Dodgers went with five starters - because unlike almost any other organization, they actually had five quality starters. How many teams can boast five starting pitchers whose names are still recognizable a quarter-century later?

The Dodgers' transition to a five-man rotation was not without its hiccups--they returned to using just four starters in 1974 and 1975--but it was ultimately irresistible. Since 1975, the Dodgers allowed a pitcher to make more than 35 starts in only one season, 1982. And during the 1970s, wherever the Dodgers tread, the rest of baseball was sure to soon follow. As ex-Dodgers and minor league managers trained in the Dodger Way promulgated their philosophy throughout baseball, the gospel of the five-man rotation was successfully spread to nearly every corner of the baseball map. By the early 1980s, only the Baltimore Orioles had not been successfully converted to the cause.

"It is easier to find four starting pitchers than five."
- Earl Weaver's Seventh Law. (From Weaver on Strategy, 1984)

In 1973, no less than twelve major league pitchers made 40 starts or more, the highest total of the 20th century. Following this trend year by year:

Year    # of Pitchers >= 40 GS

1973    12
1974    9
1975    3
1976    2
1977    2
1978    3
1979    1
1982    1
1987    1

(The 1987 outlier is Charlie Hough, a knuckleballer.)

Over the years, the concept of spreading starts out to as many pitchers as possible has been taken more and more literally. Once upon a time, the term "five-man rotation" meant that your best starters would throw every fifth day, and starters in the back of the rotation would be bumped when necessary to keep better pitchers on rotation. Today, teams are loathe to skip even their fifth starter, preferring to give their starters five or even six days of rest on occasion. The typical season runs around 180 days, meaning that an Opening Day starter working every fifth day, as opposed to every fifth game, should make at least 36 starts over a full season. A look at the number of pitchers making 36 or more starts in a season since 1973:

Year    # of Pitchers >= 36 GS          Year    # of Pitchers >= 36 GS

1973    31                              1985    14
1974    33                              1986    10
1975    22                              1987    11
1976    20                              1988    4
1977    17                              1989    7
1978    16                              1990    2
1979    14                              1991    4
1980    14                              1992    2
1982    12                              1993    5
1983    9                               1996    2
1984    5                               1998    1

Where have all those starts gone? To less-qualified pitchers, naturally. Whereas a generation ago, a team could expect to place 80% of its starts in the hands of its four best starting pitchers, today that percentage is far less.

Compare the status of major league rotations in 1973 compared to 1999. The following chart lists the average number of starts made by each slot in the rotation:

                                Cumulative
Slot    1973    1999    Diff    1973    1999    Diff

1       37.3    33.2    -4.1    37.3    33.2    -4.1
2       34.1    30.7    -3.4    71.4    63.9    -7.5
3       29.6    28.3    -1.3    101.0   92.2    -8.8
4       23.2    23.6    +0.4    124.2   115.8   -8.4
5       14.9    17.9    +3.0    139.1   133.7   -5.4
6+      22.8    28.2    +5.4    161.9   161.9   0

(The cumulative numbers add up to only 161.9 because of a few rainouts.)

What this chart tells us is that the average "ace" starter made 37.3 starts in 1973, but only 33.2 starts in 1999 - a dropoff of 4.1 starts. Number two starters lost almost as many starts (3.4), and #3 starters lost 1.3 starts apiece. Those 8.8 starts have been redistributed to far less qualified pitchers. A miniscule number went to #4 starters, but the vast majority have gone to pitchers ranked 5th or worse on a team's depth chart.

That's nearly 9 additional starts - 7.5 of which came from the two best starters on a team - that have been redistributed to fifth starters, spot starters, long relievers, and emergency call-ups. That's a pretty steep price to pay for the luxury of a five-man rotation.

"I expect the real reason baseball will eventually return to the four-man rotation will be the simplest of all: It helps win games. The five-man rotation is not on that evolutionary path; it's a digression, a dead-end alley. Just as baseball once believed that walking a lot of batters was better than throwing a home-run pitch, we are now chasing an illusion that our pitchers work better on four days' rest and that the five-man rotation significantly improves their future."
- Craig Wright, in The Diamond Appraised, 1989.

As Wright points out, the crux of the entire argument for the five-man rotation is that it keeps pitchers healthier. Which makes the crux of my argument this: the five-man rotation does NOT keep starters any healthier than the four-man rotation. And if it doesn't provide any health benefits, what benefit is there?

Between 1973 and 1975, 68 pitchers made between 37 and 43 starts, roughly conforming to expectations in a four-man rotation. Coincidentally, between 1991 and 1993 exactly 68 pitchers made between 34 and 35 starts, the high end of expectations for starters in a five-man rotation. Let's compare the two groups:

Group        G    GS    CG   ShO     W     L   Pct     IP      H     HR    BB      K    ERA
Four-Man  2676  2642  1088   226  1230   973  .558  19341  18005   1522  6123  11494   3.31
Five-Man  2343  2335   368   101  1040   771  .574  15960  14874   1313  4784  10527   3.51

In terms of quality, the two groups are very similar. The guys in the five-man rotation had a slightly higher winning percentage (but fewer decisions per start). They had an ERA 20 points higher than the earlier group; since the average major league ERA from 1973 to 1975 was 3.69, and the average major league ERA from 1991 to 1993 was 3.94, the relative ERA of the five-man group was about 5 points lower than that of the four-man group.

The average age of the guys in the five-man rotation was 29.2 years old; the guys in the four-man rotation were 29.4 years old.

While the two groups are of comparable quality, their value--owing to the increased use of the pitchers in the four-man rotation--is not comparable. The earlier group averaged more starts (38.9 vs. 34.3) and more innings (284.4 vs. 234.7) than the latter group. Fifty additional innings are a tremendous commodity - remember, we're talking about the best starters in the league here.

Here's the kicker: all those extra starts, and all those extra innings, did NOT cause any long-term damage to those pitchers. If you take the same group of pitchers and examine their performance 5 years later, here's what you find:

Group        G    GS    CG   ShO     W     L   Pct     IP      H     HR    BB      K    ERA
Four-Man  1562  1236   342    75   545   431  .558   8968   8666    756  2940   5200   3.58
Five-Man  1476  1351   121    36   538   473  .532   8722   9010    916  2857   6677   4.31

Five years later, 50 of the 68 pitchers in the original four-man group were still active, compared to 54 of the 68 pitchers in the original five-man group. The most likely explanation for the discrepancy is simply age; while the average age of the two original groups was similar, 9 of the 68 pitchers in the earlier group were 35 and older; only 5 of the 68 pitchers in the later group were.

But even with four fewer members to count, the earlier group was still going as strong, if not stronger, five years later. The surviving members of the four-man group made fewer starts, but pitched in more games, and threw more innings than the five-man group. (They did suffer a greater percentage drop in their innings, but keep in mind a big reason for that is that in the ensuing five years, many of their teams switched from a four-man to a five-man rotation.)

Their winning percentage, which was 16 points lower than their counterparts in the original study, is now 26 points higher. Their combined ERA is 24 points better than league average (3.82), while the pitchers in the five-man group had an ERA only 17 points better than league average (4.48). This is also a reversal from five years prior; while the original four-man group had a relative ERA 5 points worse than the five-man group, the same group now has an ERA 7 points better.

The last paragraph is a little confusing, so let's put that in chart form:

          Year 0   Year 5          Year 0   Year 5
Group    Win Pct  Win Pct  Diff   Rel ERA  Rel ERA   Diff

Four-Man   .558    .558    .000    -0.38    -0.24   +0.14
Five-Man   .574    .532   -.042    -0.43    -0.17   +0.26

Rel ERA = Relative ERA

Five years after our study, the pitchers in a four-man rotation had the exact same winning percentage, and a relative ERA only 14 points higher than before. By comparison, the pitchers in a five-man rotation saw their winning percentage drop 42 points, and their ERA rise by nearly twice the margin of the four-man group.

Bottom line: if these numbers suggest anything, it's that pitching in a four-man rotation is less damaging than pitching in a five-man rotation. Now, the difference between the two groups isn't enormous, and neither is the sample size, so I'll concede the point that these differences are not statistically significant. I'm not trying to argue that working on three days' rest is more healthy than working on four days' rest, only that it isn't less healthy. Given the obvious tactical benefits that come from taking innings away from the worst pitchers on your staff and giving them to your best, shouldn't that be enough?

Next week, I'll address the apparent paradox: if limiting pitch counts is such a good idea because it reduces the risk of pitcher injury, how can it also be a good idea to increase pitch totals by increasing the number of starts a pitcher makes?

Rany Jazayerli is an author of Baseball Prospectus. You can contact him by clicking here.

Rany Jazayerli is an author of Baseball Prospectus. 
Click here to see Rany's other articles. You can contact Rany by clicking here