Examining the Relief of Relieving

Which task is more difficult for pitchers to master: starting or relieving?

For decades, I’ve strongly believed that starting is clearly the more challenging role. But am I correct in that belief? And if that belief is correct, to what degree is it correct; in other words, how much easier is relieving than starting? And is it equally easier for all pitchers, or do certain types of pitchers demonstrate distinct patterns of starting versus relief effectiveness?

These are far simpler questions to ask than to answer, of course. But why don’t we take a crack at trying to answer them.

How about tackling the first one first.

Is Relieving Easier Than Starting?

If starting and relieving are equally difficult tasks, then we would expect to see roughly equal overall ERA performances from starters and relievers. But we don’t. A quick check of that magnificent resource, Retrosheet.org, clearly demonstrates that we don’t. Here are the overall major league ERAs of all pitchers, broken down into starting versus relieving appearances, over each 5-year period from which Retrosheet split data are available:

Overall MLB Starter vs. Relief ERA

   Years     Starter  Relief  Delta
   2001-05     4.50    4.10    - 9%
   1996-00     4.68    4.35    - 7%
   1991-95     4.24    3.96    - 7%
   1986-90     4.02    3.66    - 9%
   1981-85     3.93    3.50   - 11%
   1976-80     3.88    3.60    - 7%
   1971-75     3.60    3.43    - 5%
   1966-70     3.48    3.40    - 2%
   1961-65     3.76    3.56    - 5%
   1957-60     3.88    3.77    - 3%

The distinction here could hardly be more clear: over a span of almost 50 years, through many changes in playing conditions and pitching staff usage styles, reliever ERAs have been consistently better, almost always by a factor of between 5% and 10%. This is very strong evidence that starting is the more difficult thing to do.

But …

Alas, there are rather major limitations on this extremely “macro” analysis. Fundamentally, they are:

• Relief pitchers have an inherent advantage over starters in ERA, which is that they may enter a game with the starter’s runners on base, and allow them to score, which hurts the starter’s ERA without affecting the reliever’s. Starters can never return this “favor.” Thus simply looking at ERA is insufficient, and indeed the evidence from ERA alone may very well overstate the difference in difficulty between the two assignments.

And

• In any given year, a different set of pitchers is doing the bulk of the starting versus the bulk of the relieving; specialists are generally employed in both roles. Moreover, in looking at the careers of pitchers over the past several decades, at least two clear patterns emerge:

(a) The most celebrated starters, the ones who win the bulk of the Cy Young Awards and are most likely to reach the Hall of Fame, rarely make relief appearances at all in the modern era. They tend to be starting specialists for their entire careers.

(b) The vast majority of very marginal pitchers, those with the shortest and worst major league careers, make most if not all of their major league appearances in relief.

Point (a) means that the work of those pitchers generally recognized as the best in baseball is captured little if at all in the relief ERA column above. And point (b) means that the work of the worst major league pitchers is captured almost entirely in the relief column. This would strongly suggest that the evidence from league-wide ERA alone may very well understate the difference in difficulty between the two assignments.

Therefore …

We must find a reasonable method of overcoming these limitations. It would seem that instead of looking only at ERA, we should look at the component rate stats as well, such as hits, walks, and home runs per inning. And instead of looking only at league-wide aggregate totals, we should zero in on those pitchers who actually have worked a significant number of games (to minimize small-sample-size issues) in both starting and relieving roles; career-long specialists, either starters or relievers, don’t provide much help in measuring the degree of difficulty between the two assignments.

So let’s look at the full stat line, not just ERA, and let’s focus our examination only on those pitchers who have made at least 75 major league starts and 75 major league relief appearances, for whom we have full-career split data available from Retrosheet (that is, with Major League careers beginning in 1957 and later). When we isolate this population out of the huge mass of pitching careers, we discover a total of 291 such pitchers.

So looking only at these 291 pitchers, let’s ask our first question again:

Is Relieving Really Easier Than Starting?

Two-hundred and ninety-one pitchers is a whole lot of pitchers. The group that has made at least 75 starts and 75 relief appearances includes three Hall of Famers (Gaylord Perry, Phil Niekro, and Dennis Eckersley), and a couple more who may very well become Hall of Famers (John Smoltz and Curt Schilling), as well as probably the most brilliant young pitcher in baseball today (Johan Santana). But it also includes a large number of good-but-not-great pitchers, and an even larger number of unremarkable league-average guys, as well as many journeymen who struggled to patch together a career of six or eight years. Obviously there aren’t any truly marginal cup-of-coffee types in here, but overall this group of 291 pitchers represents a broad and deep sampling of the athletes who’ve pitched the bulk of major league innings over the past half-century; in an almost exactly equal proportion of starts and relief appearances (56,000 to 55,000), their median ERA+ is 98, just slightly below league average.

Here’s how they performed in starting versus relieving assignments:

TOTAL SAMPLE OF 291 PITCHERS
  Median ERA+: 98
  96 LHP, 195 RHP (33%, 67%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (50%)           56656    n/a     6.30    8.95    0.87    2.85    5.30    3.93
As Reliever (50%)          55763    0.12    1.60    8.51    0.75    3.04    6.09    3.61
Delta                                     - 75%    - 5%   - 14%    + 7%   + 15%    - 8%

Their ERA as relievers is 8% better than when starting; very much in line with the aggregate league-wide totals we saw above. But their component rate stats are even more dramatically different between the two assignments: as relievers, working stints that are 75% shorter than when starting, they allow hits 5% less frequently, and home runs 14% less frequently. Meanwhile, they issue walks 7% more frequently while relieving (and I’ve looked only at Unintentional Bases on Balls, because of the dramatically higher rate of Intentional Bases on Balls that are issued by relievers than by starters), and they rack up strikeouts at a 15% improved rate.

The higher rates of both walks and strikeouts is strongly indicative of a different mode of pitching when relieving: without concern for pacing himself and conserving his pitch count, the reliever is free to throw harder, as well as to shoot for corners. These 291 pitchers typically have surrendered more walks, but been much tougher to hit, especially for power, when pitching just an inning or two as opposed to six or seven. And, of course, when working in relief they’ve rarely been left in the game long enough to suffer the effects of in-game fatigue, rarely had to attempt to get the same hitter out multiple times, and not had to deal with the heat-up, cool-down, heat-up-again dynamic they faced when starting.

This stands as strong evidence that the answer to our initial question, “Is relieving easier than starting?,” is unequivocally, “Yes.” And the answer to our second question, “How much easier is relieving than starting?,” would seem to be, “Somewhere between 5% and 15%, all things considered.”

So let’s delve more deeply into the population of 291, and see if we can find some answers to our larger question: “Is relieving equally easier for all pitchers, or do certain types of pitchers demonstrate distinct patterns of starting versus relief effectiveness?”

Lefty Versus Righty

Does left- or right-handedness make a difference in this regard? Just about exactly one-third of our 291 pitchers are southpaws. Is there any difference in how they adjust to the relieving role versus how right-handers do it?

LEFT-HANDED PITCHERS (96)
  Median ERA+: 98

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (52%)           19853    n/a     6.28    8.96    0.87    2.84    5.44    3.88
As Reliever (48%)          18522    0.09    1.45    8.53    0.74    3.19    6.22    3.64
Delta                                     - 77%    - 5%   - 14%   + 12%   + 14%    - 6%
Difference from Total                      - 2%     =       =      + 5%    - 1%    + 2%

RIGHT-HANDED PITCHERS (195)
  Median ERA+: 98
                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (50%)           36803    n/a     6.32    8.94    0.87    2.86    5.22    3.96
As Reliever (50%)          37241    0.14    1.67    8.51    0.75    2.97    6.03    3.60
Delta                                     - 74%    - 5%   - 14%    + 4%   + 16%    - 9%
Difference from Total                      + 1%     =       =      - 3%    + 1%    - 1%

Nothing significant emerges. The “Difference from Total” line compares the Starter-Reliever Delta figures of each subgroup from that of the total sample of 291. The lefties give up a few more walks when relieving than do the righties, but overall the patterns are quite similar. It doesn’t appear that which arm a pitcher throws with has anything to do with this phenomenon.

Deployed Evenly In Both Roles

While every one of these 291 pitchers made at least 75 of both types of appearances, most made quite a few more of one than the other, and a few are rather extreme: they range from Gaylord Perry, at 7.93 starts per relief appearance, to Jose Mesa, at 0.12. So perhaps the inclusion of so many near-specialists in the sample is distorting our sense of the dynamic for those who have performed both roles in nearly equal proportion.

So let’s focus only on those who compiled a ratio of between 1.10 and 0.90 starts per relief appearance. There are 30 such pitchers, ranging from Lew Krausse (1.09) to Mike Kekich and Tom Griffin (both 0.91). The best in this group are Charlie Hough, Danny Darwin, and Juan Pizarro; the worst, guys like Kekich, Jim Bullinger, and Billy Champion.

CLOSEST TO EQUAL NUMBER OF STARTING/RELIEF APPEARANCES (30 Pitchers)
  Median ERA+: 96
  13 LHP, 17 RHP (43%, 57%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (50%)            5157    n/a     6.13    9.10    0.99    3.06    5.28    4.17
As Reliever (50%)           5152    0.10    1.67    8.55    0.79    3.39    5.94    3.74
Delta                                     - 73%    - 6%   - 20%   + 11%   + 13%   - 10%
Difference from Total                      + 2%    - 1%    - 6%    + 4%    - 2%    - 2%

Again, no real distinction in their pattern from that of the total group. They give up a few less homers and a few more walks when relieving than the rest, but their overall improvement in effectiveness when relieving is just about the same as everyone else’s.

Deployed Evenly In Both Roles Through Heart of Career

Okay, but many of these guys who started and relieved in roughly equal proportion over their full careers were specialists at one or the other in different phases of their careers. Maybe the real test of the comparative difficulty of starting versus relieving is to look only at those pitchers who did a significant amount of both through the peak years of their careers.

It’s rare for an established veteran relief specialist to be converted to starter (Hough was one of them), but established veteran starters are converted to relief all the time. Moreover, most starting pitchers who do make a fair number of relief appearances in their careers do so either in their first few seasons in the majors, before they’ve reached their peak, and/or in their final few seasons in the majors, after they’ve passed their peak; indeed, among these 291 careers, this is by far the most prevalent pattern. So let’s remove all these types of guys from our sample, and isolate only the careers of pitchers who worked a lot of games in both roles while they were at their best.

So I’ve selected only those pitchers, of whom there turn out to be a total of 71. The selection process is obviously rather subjective, and one can quibble with a few of my inclusions/exclusions, but unquestionably these 71, as a group, meet the criteria far better than the other 220. Not all of them started and relieved in equal proportion: they range from Dean Chance, who started 2.63 games for every relief appearance, but even in his peak years was being used out of the bullpen by Angels manager Bill Rigney on a regular, planned basis; to Greg A. Harris, who was mostly a reliever (0.16 starts per relief appearance), but who had a few years with lots of starts in mid-career, when he was at his best. Most of these guys are classic swingmen, doing both roles on a regular basis, while a few went back and forth between full-time starter and full-time reliever in their peak phase (such as Smoltz, Gary Bell, and Don Robinson).

WORKED AS BOTH STARTER AND RELIEVER DURING PEAK YEARS (71 Pitchers)
  Median ERA+: 95
  18 LHP, 53 RHP (25%, 75%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (46%)           11060    n/a     6.12    9.00    0.92    2.97    5.34    4.06
As Reliever (54%)          12892    0.11    1.65    8.34    0.77    3.21    6.37    3.62
Delta                                     - 73%    - 7%   - 16%    + 8%   + 19%   - 11%
Difference from Total                      + 2%    - 2%    - 2%    + 1%    + 4%    - 3%

Nope. No different pattern for these guys. They follow the pattern of the larger group almost exactly.

General Overall Quality

So let’s try this. Maybe it works differently for really good pitchers than it does for really bad pitchers. Let’s sort the total group of 291 by total career ERA+, and compare the best 10% against the worst 10%.

The five best are Santana, Schilling, Smoltz, Jimmy Key, and Derek Lowe. The worst of the best is Sonny Siebert, with a career ERA+ of 110. The five worst are Kekich, Wade Blasingame, Champion, Todd Van Poppel, and Bill Greif. The best of the worst is Balor Moore, with a career ERA+ of 87. These are two very, very different groups of pitchers.

BEST CAREER ERA+ (29 Pitchers)
  Median ERA+: 115
  11 LHP, 18 RHP (38%, 62%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (52%)            8034    n/a     6.75    8.34    0.78    2.64    6.29    3.41
As Reliever (48%)           7358    0.28    1.47    8.08    0.67    2.65    6.95    3.11
Delta                                     - 78%    - 3%   - 14%    + 0%   + 10%    - 9%
Difference from Total                      - 3%    + 2%     =      - 7%    - 5%    - 1%

WORST CAREER ERA+ (29 Pitchers)
  Median ERA+: 84
  9 LHP, 20 RHP (31%, 69%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (49%)            3415    n/a     5.70    9.50    1.00    3.40    4.98    4.69
As Reliever (51%)           3583    0.05    1.66    9.13    0.84    3.61    5.76    4.29
Delta                                     - 71%    - 4%   - 16%    + 6%   + 16%    - 9%
Difference from Total                      + 4%    + 1%    - 2%    - 1%    + 1%    - 1%

Both the best and the worst mimic the pattern of relief improvement shown by the overall group. The best guys are able to keep their walk totals constant when relieving, but then their strikeout rates don’t rise as much as those of the rest. Overall the effect is a wash.

Old-School Versus Newbie

Hmm … has the effect changed over time? Let’s look at the 10% who came along earliest, compared with the 10% who’ve come along most recently. The old fogey group includes the likes of Bob Shaw, Don Lee, and Dick Drott, while the whippersnappers are headed up by kids such as Victor Zambrano, Santana, and Vicente Padilla.

EARLIEST DEBUT (29 Pitchers)
  Median ERA+: 100
  7 LHP, 22 RHP (24%, 76%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (55%)            6115    n/a     6.39    8.80    0.89    2.74    5.15    3.74
As Reliever (45%)           5087    0.30    1.76    8.35    0.75    2.99    5.53    3.46
Delta                                     - 72%    - 5%   - 16%    + 9%    + 7%    - 7%
Difference from Total                      + 3%     =      - 2%    + 2%    - 8%    + 1%

MOST RECENT DEBUT (29 Pitchers)
  Median ERA+: 95
  8 LHP, 21 RHP (28%, 72%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (46%)            3682    n/a     5.91    9.60    1.17    3.10    5.84    4.74
As Reliever (54%)           4278    0.01    1.33    9.03    0.95    3.18    6.92    4.28
Delta                                     - 77%    - 6%   - 19%    + 3%   + 18%   - 10%
Difference from Total                      - 2%    - 1%    - 5%    - 4%    + 3%    - 2%

The newer guys see more of a strikeout boost in relief, but overall the pattern hasn’t meaningfully changed.

Walk Prevention

All right then, maybe it’s all about the control. The top 10% control artists are led by Jim Merritt, David Wells, and Jim Barr. The wild things include John D’Acquisto, David West, and Ron Villone.

BEST CONTROL PITCHERS (29 Pitchers) 
  Median ERA+: 103
  8 LHP, 21 RHP (28%, 72%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (54%)            7311    n/a     6.59    9.14    0.88    1.90    5.13    3.72
As Reliever (46%)           6147    0.15    1.61    8.77    0.74    1.94    5.73    3.41
Delta                                     - 76%    - 4%   - 16%    + 2%   + 12%    - 8%
Difference from Total                      - 1%    + 1%    - 2%    - 5%    - 3%     =

WORST CONTROL PITCHERS (29 Pitchers)
  Median ERA+: 90
  12 LHP, 17 RHP (41%, 59%)

                             G     SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (46%)            3812    n/a     5.92    8.59    0.85    4.28    6.36    4.23
As Reliever (54%)           4462    0.08    1.59    8.13    0.76    4.29    7.10    3.96
Delta                                     - 73%    - 5%   - 11%    + 0%   + 12%    - 6%
Difference from Total                      + 2%     =      + 3%    - 7%    - 3%    + 2%

The pattern between the two groups is remarkably similar. Neither shows the tendency to give up more walks while relieving that everyone else does, and everything else is just about spot on with the full group.

Strikeout Ability

Could strikeouts be the secret? The top five Kings of the K are Santana, Sam McDowell, Schilling, Scott Sanders, and Darren Dreifort. The leaders of the Let-‘Em-Hit-It League are Jeff Ballard, Lary Sorensen, Paul Hartzell, Steve Comer, and Al Fitzmorris.

BEST STRIKEOUT PITCHERS (29 Pitchers)
  Median ERA+: 103
  13 LHP, 16 RHP (45%, 55%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (41%)            5128    n/a     6.33    8.32    0.89    3.32    7.51    3.87
As Reliever (59%)           7499    0.18    1.36    7.90    0.79    3.58    8.15    3.62
Delta                                     - 79%    - 5%   - 11%    + 8%    + 9%    - 6%
Difference from Total                      - 4%     =      + 3%    + 1%    - 6%    + 2%

WORST STRIKEOUT PITCHERS (29 Pitchers)
  Median ERA+: 97
  10 LHP, 19 RHP (35%, 65%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (56%)            5704    n/a     6.40    9.41    0.81    2.49    3.54    3.96
As Reliever (44%)           4422    0.10    1.90    9.21    0.69    2.57    4.30    3.58
Delta                        56%          - 70%    - 2%   - 15%    + 3%   + 21%   - 10%
Difference from Total                      + 5%    + 3%    - 1%    - 4%    + 6%    - 2%

The soft-tossers gain more strikeouts in relief (not that they had many to begin with), but overall the patterns of both of these disparate types are pretty much the same.

The Long Ball

Perhaps home run prevention is the key skill that differentiates pitchers here. The Keep-It-in-the-Park Club includes Bob Veale, Bruce Dal Canton, and Luke Walker, while the Gopher Ball Group is headed up by Bruce Chen, Jose Lima, and Van Poppel.

BEST HOME RUN PREVENTERS (29 Pitchers)
  Median ERA+: 101
  10 LHP, 19 RHP (35%, 65%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (46%)            5428    n/a     6.33    8.77    0.60    3.07    5.31    3.65
As Reliever (54%)           6370    0.15    1.70    8.61    0.54    3.01    5.49    3.37
Delta                                     - 73%    - 2%   - 10%    - 2%    + 3%    - 8%
Difference from Total                      + 2%    + 3%    + 4%    - 9%   - 12%     =

WORST HOME RUN PREVENTERS (29 Pitchers)
  Median ERA+: 93
  8 LHP, 21 RHP (28%, 72%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (49%)            4162    n/a     5.96    9.52    1.30    3.03    5.62    4.72
As Reliever (51%)           4372    0.06    1.48    9.03    1.05    3.13    6.64    4.23
Delta                                     - 75%    - 5%   - 19%    + 3%   + 18%   - 10%
Difference from Total                       =       =      - 5%    - 4%    + 3%    - 2%

This is interesting. Both groups demonstrate a similar result in terms of ERA, but they get there taking different routes: the low-HR guys show significantly less change in their walk and strikeout rates when relieving than the total group shows, while the high-HR guys are pretty much the same as everyone else. I’m not sure what this means, but something might be going on here. This is one slice of the pie that could merit some further scrutiny.

Closer Deployment

How about simply looking at the group most likely called upon to serve in the closer mode, compared against those least asked to do so. The most active and successful closers here are Smoltz, Eckersley, Rick Aguilera, Ryan Dempster, and Jose Mesa, while those who never got a Save at all include John Halama, Chen, Woody Williams, Curt Young, and Paul Byrd.

MOST SAVES PER RELIEF APPEARANCE (29 Pitchers)
  Median ERA+: 104
  4 LHP, 25 RHP (14%, 86%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (36%)            5793    n/a     6.51    8.77    0.82    2.86    5.81    3.82
As Reliever (64%)          10331    0.30    1.46    8.25    0.67    2.64    6.56    3.22
Delta                                     - 78%    - 6%   - 18%    - 8%   + 13%   - 16%
Difference from Total                      - 3%    - 1%    - 4%   - 15%    - 2%    - 8%

FEWEST SAVES PER RELIEF APPEARANCE (29 Pitchers)
  Median ERA+: 94
  12 LHP, 17 RHP (41%, 59%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (48%)            4027    n/a     5.93    9.46    1.13    2.99    5.43    4.56
As Reliever (52%)           4435    0.01    1.48    9.05    0.95    3.18    6.00    4.23
Delta                                     - 75%    - 4%   - 16%    + 6%   + 10%    - 7%
Difference from Total                       =      + 1%    - 2%    - 1%    - 5%    + 1%

Not surprisingly, the Did-Lots-of-Closing group shows a more pronounced effectiveness boost when relieving, caused almost entirely by their ability to avoid the base on balls in relief. The other bunch is almost exactly in line with the larger group. I don’t see anything very meaningful here.

Relief Boost Versus Relief Bust

All right, so how about we just do this: let’s look at the top 10% of pitchers who show the greatest improvement in ERA when relieving, and compare them against the 10% who show the least. Indeed, not all 291 pitchers actually show an ERA improvement when relieving; most do, but 79 of the 291 (27%) show a decline, and in 37 of those, it’s a decline of more than 10%. So the 29 least-improvement pitchers we’re looking at here in fact all declined in ERA performance when pitching in relief.

The top five Big Improvers are LaTroy Hawkins, Sammy Ellis, Billy Swift, Chris Hammond, and Jim Bibby. The top five Big Decliners are John Smiley, Veale, Ken Johnson, Fred Norman, and Tracy Stallard.

MOST IMPROVED ERA WHEN RELIEVING (29 Pitchers)
  Median ERA+: 98
  5 LHP, 24 RHP (17%, 83%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (45%)            5257    n/a     6.10    9.15    0.90    3.11    5.56    4.25
As Reliever (55%)           6297    0.17    1.49    7.91    0.62    2.91    6.72    2.99
Delta                                     - 76%   - 14%   - 31%    - 6%  + 21 %   - 30%
Difference from Total                      - 1%    - 9%   - 17%   - 13%    + 6%   - 22%

MOST DECLINED ERA WHEN RELIEVING (29 Pitchers)
  Median ERA+: 98
  14 LHP, 15 RHP (48%, 52%)

                              G    SV/G    IP/G     H/9    HR/9   UBB/9    SO/9     ERA
As Starter (64%)            6647    n/a     6.32    8.62    0.82    2.79    5.55    3.67
As Reliever (36%)           3674    0.07    1.51    9.08    0.87    3.67    6.34    4.47
Delta                                     - 76%    + 5%    + 6%   + 32%   + 14%   + 22%
Difference from Total                      - 1%   + 10%   + 20%   + 39%    - 1%   + 30%

Clearly, both of these subgroups deviate hugely from the norm, and in quite opposite directions. Let’s take a closer look at the members of each group and see if we can make some sense of this.

When we examine the careers of each of the 29 pitchers in each group, we will specifically look for dynamics that might help explain these results, including the phase in their development/decline in which they were predominantly used in each role, what kind of run environments they pitched in at each phase, and so on, what we find is (these categorizations are obviously less than perfectly precise):

GROUP: MOST IMPROVED ERA WHEN RELIEVING

• Swingman* (11); Used pretty much interchangably in both roles throughout their careers, with no particular factors that would explain their unusually strong relief advantage: Bibby, Shawn Boskie, Bullinger, Pete Falcone, Joe Gibbon, Halama, Bart Johnson, Barry Latman, Bob Miller, Charlie Puleo, and Bob Walk.
• Struggling Starter, Reborn as Reliever* (7); Young starters whose careers were going nowhere, who then resurfaced as relievers and found sudden success: Mark Davis, Dempster, Jim Gott, Hammond, Hawkins, Brian Meadows, and Jeff Russell.
• Reliever Converted to Starter** (4); Oustanding young relievers converted to the starting rotation, who experienced some success at starting but soon declined: Ellis, Greg W. Harris, Lowe, and Swift.
• Swingman Converted to Starter* (3); Broke into the majors in the swingman role, then promoted to full-time starter, where they spent the rest of their career, including the decline phase: Doyle Alexander, Bill Bonham, and Bill Hands.
• Starter Converted to Swingman* (2); Starters who moved into the swingman role in their decline phase: Kevin Gross and Steve Renko.
• Swingman Converted to Reliever** (1); A good swingman who does even better as a relief specialist, the impact of which is amplified by doing most of the relieving in a lower-scoring environment (post-2000)—a quite singular story: Tom Gordon.
• Starter Converted to Reliever then Back to Starter* (1); A great starter who becomes a great reliever and back again, a quite singular story: John Smoltz.

GROUP: MOST DECLINED ERA WHEN RELIEVING

• Starter Converted to Reliever*** (6); Starters who went to the bullpen in their decline phase: Ballard, John Candelaria, Jose DeLeon, Bob Forsch, Kirk McCaskill, and Lynn McGlothen. (Ballard’s, DeLeon’s, and McCaskill’s stats are amplified by moving to the bullpen just as the run environment increased in the mid-1990s.)
• Reliever Converted to Starter*** (5); Used in the bullpen as struggling young pitchers, then as starters in the peak years: Omar Daal, Dustin Hermanson, Key, Smiley, and Ed Whitson. (Hermanson has been moved back to the bullpen following arm trouble, a move met with intermittent success.)
• Swingman Converted to Starter then Reliever*** (4); Broke into the majors in the swingman role, then promoted to full-time starters, then moved to the bullpen during their decline phases: Ken Johnson, Norman, Pizarro, and Veale.
• Star Young Starter Who Encountered Arm Trouble*** (4); Made early splashes as a starters, then got hurt, then struggled to hang on as relievers (4): Steve Barber, Danny Cox, Steve Hargan, and Craig McMurtry. (Hargan’s stats are amplified by doing much of his starting in the low-scoring 1960s, and much of his relieving in the higher-scoring 1970s.)
• Swingman* (4); Used pretty much interchangably in both roles throughout their careers, with no particular factors that would explain their strong starting advantage: Bob Hendley, Larry McWilliams, Pete Schourek, and Carlton Willey.
• Swingman Converted to Starter*** (2); Broke into the majors in the swingman role, then promoted to full-time starters, where they reached their peak: Santana and Stallard. (Santana has not yet approached a decline phase. Stallard got hurt and disappeared before much of a decline phase could take place.)
• Only Decent Year as a Starter* (2); Not really good pitchers, who had one decent early year as a starter: Willie Banks and Jesus Sanchez.
• Starter Converted to Swingman*** (1); A starter who moved into the swingman role in his decline phase: Burt Hooton.
• Starter Converted to Swingman then Back to Starter*** (1); A young star starter who struggled with alcoholism in mid-career, and then recovered to be a veteran star starter, a quite singular story: Dennis Martinez.

– – – – – – – – – – – – –

* The career dynamics of these 30 pitchers don’t seem to hold much explanation for why they performed as extremely as they did in starting vs. relieving roles, in either direction.

** The career dynamics of these 5 pitchers provide a good indication of why they performed so much better as a reliever than as a starter.

*** The career dynamics of these 23 pitchers provide a good indication of why they performed so much better as a starter than as a reliever.

Thus we find far more prevalent mitigating circumstances, if you will, explaining the better-as-a-starter performances.

What It All Means

This data suggests several conclusions:

• Over the period of the past half-century, the performances of major league pitchers who’ve done a significant amount of both starting and relieving clearly demonstrates that relief pitching is a substantially less challenging assignment than starting pitching.
• There have been some pitchers whose performance have trended in the opposite direction, but they are distinctly in the minority, and there often appear to be particular career-dynamic circumstances that explain at least a major proportion of their unusual pattern. There might well be some pitchers who are inherently better suited to starting than relieving, but they are a very small minority, and impossible to easily predict by any apparent means.
• Given this, we should routinely apply some degree of effectiveness “equalizer” when considering the performance of Starting Pitcher A versus Relief Pitcher B. Something on the order of 5% to 15% on all rate stats, including ERA, would seem reasonable.

References & Resources
This sample of pitchers was pulled from Sean Lahman’s database, on the Baseball Archive website. Thanks to my buddy Vinay Kumar for actually pulling it for me, since Vinay knows how to do it and I don’t.

Then, all the stats were found at Retrosheet.org.

Anyone who wishes to improve upon this less-than-sophisticated analysis is more than welcome to make use of my spreadsheet. Email me at
.