Combined Use Scores for Pitchers

February 12, 2014

Hey Bill, looking back at your writing on regularity of use, I guess the request of Whitey Ford of Houk for regular work has developed into something of a sea change moment. It occurred to me that one of the underpinning complaints of Ford's request -- that under Stengel he had been held out of games against weaker opponents to toss against upcoming stronger opponents -- would leave a statistical residue after this transition period passed. It seems like to a certain degree, the back of the rotation saw weaker opponents than the aces more than today's etched-in-stone rotations. Do you feel this evened out the historic records prior to the 60s between the aces and the whozatts to a meaningful degree?

 

I certainly think that there was a sea change in the view of baseball men as to the regularity of use of starting pitchers, and that this change occurred at about the time you indicate.   I don’t know, however, whether Whitey Ford’s public request of Ralph Houk to leave him in the rotation and let him pitch actually precipitated this change or was merely in keeping with something that was happening at that time, so I got interested in that question. ..was Ford’s public request actually the trigger that led to that change?

Not precisely on point of the reader’s question, I developed the concept of a "Combined Use Score" for a pitcher.   The Combined Use Score of a pitcher is simply the number of starts that he has made, multiplied by the number of relief appearances he has made.    If you make 10 starts and 10 relief appearances, that’s a Combined Use Score of 100.   If you have 15 of one and 20 of the other, that’s a Combined Use Score of 300.   If you pitch only in relief or only as a starter, you will have a Combined Use Score of zero.

This turns out to be an extremely handy tool.  One can figure a Combined Use Score for a pitcher in a season, for a pitcher in a career, for a team, for a league, or for a year.   And by the use of these scores, a few things that previously were not entirely clear become…well, entirely clear.

First point. .. most pitcher/seasons in baseball history have non-zero Combined Use Scores—over 50%, even if you include the pitchers who only pitched in one or two or three games.  In modern baseball, of course, the opposite is true.  In 2013 there were 679 major league pitchers, of whom 537 had Combined Use Scores of zero.  That’s 79%.  Among those pitchers who appeared in ten games or more, 78% had Combined Use Scores of zero.

Between 1920 and 1939, however, 76% of pitchers had Combined Use Scores greater than zero, and among pitchers who appeared in ten games or more, 94% had Combined Use Scores greater than zero.  It was simply assumed of every pitcher that he could and would pitch whenever and however he was needed to pitch. 

The highest Combined Use Score ever (868) was by Dave Davenport in 1916, 31 starts and 28 relief appearances.  The top ten are almost all from the years 1908 to 1916:

First

Last

Team

Year

CUS

Dave

Davenport

St. Louis Browns

1916

868

Ed

Walsh

Chicago White Sox

1912

861

Ed

Walsh

Chicago White Sox

1908

833

Reb

Russell

Chicago White Sox

1916

780

Eddie

Rommel

Philadelphia Athletics

1923

775

Hugh

Mulcahy

Philadelphia Phillies

1937

775

Sheldon

Jones

New York Giants

1948

714

Ed

Walsh

Chicago White Sox

1911

703

ThreeFinger

Brown

Chicago Cubs

1911

702

Bob

Shawkey

New York Yankees

1916

702

Until the early years of the 20th century is was assumed that the starting pitcher would finish every game.  When this assumption started to crumble, the first relievers were other starters.   One starter knocked out; another one in.   There weren’t any bullpen specialists.  This resulted in pitchers with 30 starts, 25 relief appearances, or combinations like that.

In a career, however, almost all of the highest totals are by pitchers of the last two generations.   The highest Combined Use Score ever, for a career, was by Dennis Eckersley, with 361 starts and 710 relief appearances:

First

Last

Starts

Relief Appearances

CUS

Dennis

Eckersley

361

710

256,310

Charlie

Hough

440

418

183,920

Jim

Kaat

625

273

170,625

Rick

Honeycutt

268

529

141,772

Tom

Gordon

203

687

139,461

Jack

Quinn

443

312

138,216

Kenny

Rogers

474

288

136,512

Danny

Darwin

371

345

127,995

Ron

Reed

236

515

121,540

Terry

Mulholland

332

353

117,196

John

Smoltz

481

242

116,402

Jeff

Fassero

242

478

115,676

 

And these pitchers, of course, were not pitchers who were working the two jobs interchangeably, but pitchers who switched at some point in their careers from being starters to being relievers, or vice versa.  Danny Darwin and Terry Mulholland were legitimate dual-use guys.

A couple of other things we could do here, but I will spare you, are:

1)  To give you the league leader for each season in baseball history, and

2)  To give you a list of the most effective pitchers pitching in a mixed role in each season.

The league leading totals actually have not changed very much over time.  The major league leader in 2013 was Garrett Richards of the Padres (510). . .well, here are the major league leaders over the last ten years:

 

First

Last

Team

Year

CUS

Dustin

Hermanson

Giants

2004

522

Glendon

Rusch

Cubs

2005

513

Ryan

Madson

Phillies

2006

561

Zack

Greinke

Royals

2007

532

Miguel

Batista

Mariners

2008

480

Bobby

Parnell

Mets

2009

480

Brian

Duensing

Twins

2010

520

Cory

Luebke

Padres

2011

493

Brian

Duensing

Twins

2012

484

Garrett

Richards

Angels

2013

510

These numbers actually are not very much different from the numbers of the league leaders in the 1950s.  Dave Koslo led the majors in 1952 with a score of 408.    The reason that number hasn’t changed very much over the last 60 years (while every other number we will see here has changed) is that there is a natural limit to it.   If you pitch in 50 games and they’re evenly divided between starts and relief appearances, that’s a score of 625.  You can’t go much higher than that, under normal circumstances, and there is always one pitcher in the majors who divides his season fairly evenly between the two, so that guy is always going to be up near the limit.

But now there are one or two pitchers per season who are up near the limit.  In the 1950s there were many more.   In 1954 there were 42 major league pitchers who had Combined Use Scores over 200; in 1955 there were 51, and in 1956, 45.   In 2011, with twice as many teams and a longer schedule, there 14 pitchers with Combined Use Scores over 200.   In 2012 there were 14, and in 2013 there were 15.  The number of pitchers per team being used a significant amount in both roles has decreased by about 80%.

We can chart the decrease in the acceptance of the practice of using a pitcher in both roles by charting the major league Combined Use Score total over time.   From 1876 until 1904, this total was under 10,000 every year:

Year

Total

1876

823

1877

596

1878

192

1879

1,091

1880

2,277

1881

1,082

1882

1,088

1883

3,290

1884

3,412

1885

869

1886

1,624

1887

1,397

1888

1,166

1889

6,113

1890

8,491

1891

8,459

1892

6,425

1893

6,558

1894

7,596

1895

6,815

1896

5,739

1897

5,261

1898

5,177

1899

5,615

1900

4,572

1901

7,281

1902

5,796

1903

6,828

1904

6,390

At about that time there began what we could call the Ed Walsh era, in which starting pitchers were also used to pitch relief.  From 1904 to 1914, the major league Combined Use Score total jumped almost six fold:

Year

Total

1904

6,390

1905

11,114

1906

11,997

1907

14,734

1908

19,140

1909

16,621

1910

18,850

1911

21,726

1912

21,629

1913

24,869

1914

37,619

The 1914 figure, aided by the fact that there was a third league that year, the Federal League, climbed to 37,619.   That remains to this day the highest total ever for a season.  From 1914 to 1958, the major league total was flat:

 

Year

Total

1915

36,960

1916

28,695

1917

25,100

1918

13,248

1919

16,592

1920

20,342

1921

23,241

1922

25,145

1923

24,902

1924

23,930

1925

23,338

1926

25,372

1927

23,946

1928

23,816

1929

24,869

1930

26,033

1931

24,186

1932

26,023

1933

22,686

1934

28,185

1935

27,974

1936

26,907

1937

24,933

1938

22,035

1939

23,587

1940

20,872

1941

21,085

1942

18,796

1943

19,549

1944

20,901

1945

18,571

1946

22,141

1947

23,982

1948

24,551

1949

23,215

1950

25,482

1951

24,402

1952

20,139

1953

24,756

1954

22,150

1955

23,550

1956

23,074

1957

24,991

1958

24,513

 

What was established as the accepted practice, in the Ed Walsh era, remained the accepted practice until the late 1950s, except that managers backed off the very high games and innings loads of the 1910s.

And then—as was stated at the top of the article—there was a relatively sudden change.    Because of the 1961/1962 expansion I am going to list the per-team figures as well as the major league totals.   The major league average per team dropped from 1,532 in 1958 to 914 in 1968:

Year

Total

Teams

Per Team

1958

24,513

16

1,532

1959

22,591

16

1,412

1960

22,357

16

1,397

1961

24,601

18

1,367

1962

28,068

20

1,403

1963

24,924

20

1,246

1964

25,376

20

1,269

1965

23,132

20

1,157

1966

23,338

20

1,167

1967

22,767

20

1,138

1968

18,277

20

914

After 40+ years of almost no change in the major league norm, the norm dropped by 40% in ten years—and continued to drop.    By 1971 the major league norm (per team) was down to 804, by 1978 down to 730, by 1985 down to 662, by 1988 down to 586, by 1997 down to 570, by 2004 down to 450.   In the last three years the major league norms (team total CUS) have been 326, 353, and 355 (1911-12-13).   The five-year trend line has continued to decrease from 1959  to the present time.

Overall, then, Combined Use Scores have dropped almost 80% since 1958.   Getting back to the question with which we began this exercise:  Was it Whitey Ford’s public request to be left alone in the starting rotation that triggered this change?

No, it was not.  The change started in 1958; Ford’s request came two years later.   There may have been a triggering event; there may not have been.  I don’t know.

Let’s look a minute at team totals.    Casey Stengel’s team totals were high.  In 1956 the Yankees’ team total was 1,522, fourth highest in baseball.   In 1959 the Yankees were at 1,591, again fourth highest in baseball; in 1960 they had a team total of 1,408, sixth highest in baseball. 

One could say, then, that Stengel’s totals were high, but not remarkably high.   There were other managers who were more willing than Stengel to shift pitchers back and forth between starting and relieving.

You know the description of a statistician, that a statistician is a person who, if you have one foot in the fire and the other in a block of ice, will tell you that on average, you’re comfortable?  There is something about Casey Stengel’s use of starter/relievers that is too subtle for the average.

The teams with very high numbers were generally the bad teams.  This is still true today.   A bad team doesn’t necessarily know who its best starting pitchers are.   A bad team is searching for answers, and, in searching for answers, one of the things they will typically do is give relievers a shot in the starting rotation.   In 1960 the major league team with the highest Combined Use Score was the Washington Senators.   In 1957 the American League leaders were the Kansas City A’s.

Yes, there were other managers who were even more inclined to switch pitchers between roles than was Stengel, and this is something that I didn’t see clearly until I had this method.    I think that I may have written that Casey Stengel was the last manager to use his ace to close out games for other pitchers.   Not strictly true.   The major league 1-2 leaders in Combined Use Scores, by season beginning in 1954. .1954 Chicago White Sox (Paul Richards, manager) and Pittsburgh Pirates (Fred Haney.)   1955 Washington Senators (Chuck Dressen) and Pittsburgh Pirates (Fred Haney).    1956 Cincinnati Reds (Birdie Tebbetts) and Washington Senators (Chuck Dressen).   1957 Cincinnati Reds (Birdie Tebbetts) and Kansas City A’s (Lou Boudreau).   1958 San Francisco Giants (Bill Rigney) and Cleveland Indians (Joe Gordon).   1959 San Francisco Giants (Bill Rigney) and Cleveland Indians (Joe Gordon).    1960 Washington Senators (Cookie Lavagetto) and Baltimore Orioles (Paul Richards).

We can see, then, that

a)  there were some managers who were comfortable doing this,

b)  it tended to be done most on bad teams, and

c)  Casey Stengel was not the manager MOST likely to use a pitcher in dual roles.

When Ralph Houk replaced Stengel in 1961, the Yankees’ position on the list really did not change.    The 1961 Yankees had Bill Stafford (25 starts, 11 relief appearances, Combined Use Score 275), Rollie Sheldon (294), and Jim Coates (352).    In 1962 Yankees had Jim Bouton (320), Jim Coates (264), Rollie Sheldon (288), and Bud Daley (222).    These are high numbers.   The Yankees didn’t stop using pitchers in mixed roles; they just stopped using Whitey Ford in both roles.

One more thing in closing.   By the use of this method, we can designate any pitcher with a Combined Use Score of 300 or more as a Mixed Role Pitcher, and we can then give the Ed Walsh Award for each season to the best Mixed Role Pitcher in the major leagues.

We could do that, but I promised you I wouldn’t.    But I can do one per decade or so.   The greatest Mixed Role season in major league history was by Ed Walsh in 1908.    Walsh made 49 starts and also 17 relief appearances, won 40 games, lost 15, pitched 464 innings and posted a 1.42 ERA.

1910s, the greatest was Walter Johnson, 1913; Craig Wright just did a nice piece on that season for Pages From Baseball’s Past, so I’ll leave that one alone.  1920s, Jim Bagby had 9 relief appearances for a Combined Use Score of 351 in 1920, when he won 31 games.  1930s, Lefty Grove had a Combined Use Score of 330 in 1931, when he went 31-4.   1940s, Hal Newhouser had 34 starts, 13 relief appearances in 1944, when he was the American League MVP with a record of 29-9, 2.22 ERA.   Again, being used in a mixed role was the norm in that era, not the exception.

1950s, the best was Mike Garcia, 1952; 22 wins, 11 losses, 36 starts, 10 relief appearances, 2.37 ERA.

1960s, the best season in a mixed-use role was by Dean Chance in 1964:  35 starts, 11 relief appearances, 20 wins, 1.65 ERA, Cy Young Award Winner (that was Bill Rigney again. Rigney was a manager who would use a pitcher both ways.)   970s, Wayne Garland in 1976 made 25 starts, 13 relief appearances, was 20-7 with a 2.67 ERA.  Mike Cuellar bombed out of the Orioles’ famous starting rotation in mid-season, and Garland slipped into the job.  Rudy May with the 1980 Yankees was 15-5, led the league in ERA at 2.46, made 17 starts and 24 relief outings.   1990s, Curt Schilling in 1992 made 26 starts, 16 relief appearances, finished 14-11 with a 2.35 ERA.    Since 2000. .. Johan Santana in 2003 made 18 starts, 27 relief appearances, went 12-3 with a 3.07 ERA, 169 strikeouts in 158 innings.

Since then we can go year by year.  2004, Rodrigo Lopez with the Orioles, 23 starts, 14 relief, 14-9 with a 3.59 ERA.

2005, Jorge Sosa of the Braves, 20 starts, 24 relief, 13-3, 2.55 ERA.

2006, Nobody really.   Brett Tomko of the Dodgers if somebody has to win—8 and 7 with a 4.73 ERA.

2007, Chad Billingsley of the Dodgers, 20 starts, 23 relief, 12-5 with a 3.31 ERA.

2008, Nobody really.  Jorge Campillo of the Braves was 8-7 with a 3.91 ERA. 

2009, Matt Palmer of the Angels, 13 starts, 27 relief, 11-2 with a 3.93 ERA.

2010, Brian Duensing of the Twins, 13 starts, 40 relief, 10-3 with a 2.62 ERA.

2011, Kyle McClellan of the Cardinals, 17 starts, 26 relief, 12-7 with a 4.19 ERA.

2012, Kris Medlen of the Braves, 12 starts, 38 relief, 10-1 with a 1.57 ERA.  

2013, Joe Kelly of the Cardinals, 15 starts, 22 relief, 10-5 with a 2.69 ERA.  

 
 

COMMENTS (13 Comments, most recent shown first)

bjames
Responding to JDW, it is actually extremely common historically (or was, from 1940 to 2000) for pitchers to start the season in the bullpen, move into the starting rotation in mid-season, and wind up leading the league in ERA. Among the pitchers who did this: Elmer Riddle, 1941, Gene Bearden, 1948, Dave Koslo, 1949, Chet Nichols, 1951, Bob Friend, 1955, Stu Miller, 1958, Frank Baumann, 1960, Hank Aguirre, 1962, Gary Peters, 1963, Phil Niekro, 1967, Diego Segui, 1970, Luis Tiant, 1972, Buzz Capra, 1974, Rudy May, 1980, Danny Darwin, 1990, Bill Swift, 1992, and Steve Ontiveros, 1994 (might also count Ed Heusser, 1944, Bobby Shantz, 1957, Hoyt Wilhelm, 1959, Sam McDowell, 1965, and Rick Sutcliffe, 1982.)

In addition to those, there is an even longer list of pitchers who started the season late or had injuries, made only about 20-25 starts, just skimmed over the qualification standard, and led the league in ERA. The reason this happens is that a pitcher’s ERA, while tremendously important, is subject to huge random fluctuations, and consequently subject to “bad reads” . ..anomalous data. ..in smaller data samples. It is A LOT easier to lead the league in ERA if you are pitching 170 innings than if you are pitching 250 innings, because you are much more likely to get a bad read in 170 innings. Since the “league leader” is, by definition, a search for an outlier, the outlier tends to occur just over the qualification standard. If you set the standard at 130 innings, a pitcher working 135 innings will very often be the league leader.

10:35 AM Feb 15th
 
David Kowalski
Allie Reynolds pitched from 1942 to 1954 and had combined use score of 38,625 with a season high of 390 in 1954, his final season. I thought that Reynolds would have a higher score. What got me thinking about him was that he, too, was managed by Casey Stengel. Unlike Lefty Grove, Reynolds' best season was not his best for combined use score.
12:51 AM Feb 13th
 
bjames
The major league combined use score is the total score for all the individual pitchers. Somebody asked about the median, rather than average, for a team. Well, that's hard to figure, because of two-team pitchers. Suppose a pitcher makes 15 relief appearances and 2 starts for one team, is traded, and makes 12 starts for another team. His CUS is 30 for the first team and zero for the second, but it is 210 for the season as a whole. We can manage this irregularity in the average by taking the league individual total and dividing by the number of teams in the league, but it's more of a problem with the median.
12:18 AM Feb 13th
 
doncoffin
I'm a little confused about how the "major league Combined Use Score total" is calculated. It's obviously not GS*(ReliefAppearances) for MLB. But what is it?
8:08 PM Feb 12th
 
jdw
A side note on Darwin - his 1990 season is an interesting "challenger" to Schilling's 1992 as the decade's best CU. Both pitchers started the season in the pen, then moved to the rotation and did not make another relief appearance in the season. Schilling made his quicker, hitting the rotation on 5/19/92 while Darwin didn't make it until July. It does nicely divide their seasons:

Darwin (1990)
Relief: 2-1, 2.40 ERA, 31 Appearances, 45 IP
Start: 9-3, 2.14 ERA, 17 GS, 117.2 IP
Total: 11-4, 2.21 ERA, 48 G, 17/31 GS/GR, 162.2 IP

Schilling (1992)
Relief: 2-2, 2.86 ERA, 16 Appearances, 28.1 IP
Start: 12-9, 2.27 ERA, 26 GS, 198 IP
Total: 14-11, 2.35 ERA, 48 G, 26/16 GS/GR, 226.1 IP

Darwin won the ERA Title exactly a decade after May pulled off the same trick in the AL.
7:30 PM Feb 12th
 
jdw
I wonder if the simplicity of the formula invites in some outliers. The reference of Mulholland is the reason I mention it.

If you pull his top seasons, you get this:

2000 - 680 (20 GS / 34 GR)
1999 - 432 (24/18)
1998 - 384 (6/64)
2004 - 360 (15/24)
1997 - 351 (27/13)

All of them look like what we would think of as a CU Pitcher... except 1998, which ends up his third highest CUS. He had so many games in relief that it takes very few starts to move the dial. 4 of them were in September in a GS + 5 GR + 3 GS stretch. If he had another GS in the place of the five GR, it would have been 7/59 and moved it up to 413. That still doesn't quite feel like a quite like the rest, let along "higher" than 2004 & 1997.

Did you run into many season where the ration of GS to Games in Relief wasn't quite a CU Pitcher?
7:06 PM Feb 12th
 
ajmilner
Thanks for the clarification, Bill.

BTW, Shawkey in 1916 and Three-Finger Brown in 1911 are the only pitchers with 20+ games finished and 20+ complete games in the same season.
12:40 PM Feb 12th
 
ventboys
With 16 teams early, then 20 in the 1960s (and so on), the sample size is small enough that maybe a couple of teams could ruin the curve for everyone? Its not my study or my effort, but it makes me wonder what the median numbers are?
11:46 AM Feb 12th
 
bjames
Just reminding everybody that there was no "Saves" data at the time that Lefty Grove retired--even unofficial. The "Save" was invented in the late 1950s by the late Jerome Holtzman, and became an official stat about ten years later. The data you see now for Saves is unofficial data, and was figured using a Saves rule which was only in effect for a few years. If a pitcher a) finished the game for the winning team, and b) was not the winning pitcher, then he was always credited with a save, no questions asked. Ernie Shore in his major league debut (1912) entered the game with a 21-2 lead, gave up 10 runs in an inning, and is credited in modern sources with a Save.
11:28 AM Feb 12th
 
ajmilner
"I read somewhere that Robin Roberts was among the Phils' leaders in career saves- amazed me."

When Robin left the Phillies after the 1961 season, his 24 career saves were third-most in team history.

What really blows my mind is that Lefty Grove, when he retired in 1941, was tied for eighth in career wins... and FIFTH in career saves.
11:20 AM Feb 12th
 
bjames
Obviously I meant 2011-2012-2013, not 1911-12-13.
10:58 AM Feb 12th
 
AJD600
I read somewhere that Robin Roberts was among the Phils' leaders in career saves- amazed me.
10:55 AM Feb 12th
 
jwilt
That's very interesting. But what it doesn't quite capture is the difference between pitchers who genuinely pitch relief in between starts, and pitchers who pitch relief for three months then start for three months. My hunch is that the former practice would see an even more precipitous decline over time than the raw CUS indicates.

For example, Rodrigo Lopez' league leading total in '04 was him relieving for all of April/May, then from June 26th-on he never relieved again. Constrast with Sammy Stewart in 1982, who had 26 relief appearances, 12 starts, almost the same CUS as Lopez (322 vs 312), but his starts were spread intermittantly over four different months.
10:11 AM Feb 12th
 
 
©2019 Be Jolly, Inc. All Rights Reserved.|Web site design and development by Americaneagle.com|Terms & Conditions|Privacy Policy