The Career Won-Lost Record

              The purpose of this research is to try to get a clear answer to this question:   Over the course of a substantial career, does a pitcher’s won-lost record become a reasonably reliable measurement of how good he really was?   In other words, we know that in a single season, a pitcher can pitch extremely well and be stuck with a won-lost record of 12-15, just because his team doesn’t score runs for him, while another pitcher—a teammate even—might pitch not as well, but might wind up with a record of 18-7 because he wins a lot of games 7-5 and 11 to 8.   But over the course of a career, do those things "even out", or do they not?

              "Even out". . . it’s an expression.   Of course the discrepancies of random chance do not disappear with more trials; they become larger with more trials.   But since the sum of the random discrepancies doesn’t grow at the same rate as the number of trials, what can happen is that, given enough trials, the effects of luck become insignificant relative to the size of the whole.   The question is, does that happen, to a pitcher’s won-lost record?   Is a career long enough, and are the random deviations small enough, that they become relatively insignificant over the course of a career?

              It’s a very live issue, in the public debate.   My friend Brian Kenny wants to Kill the Win.   Traditionalists may have moved a few feet in defending the Win in the context of a Cy Young debate—that is, in the context of evaluating seasons—but they haven’t moved an inch in the Hall of Fame debate.   And I have some sympathy for them there.   I tend to use won-lost records and rely on them, in a career context.    But the question is, should I be doing that?   Or should I move toward Brian. . . I’m not going to join Brian in trying to Kill the Win, certainly, but should I adjust my thinking on this issue in his direction?  

              I have done studies like this before, of course, but this time I did the study better than I had ever done it before, I think better than anyone has done this particular study.   I’m sure I studied this in the 1970s or eighties, but I wouldn’t have had the ability to do then the things I did now, to make the adjustments that I made now.  I spent weeks doing the research; I don’t necessarily know that it was worth the time and effort, and I acknowledge that a skilled programmer could have done the same work in a third of the time, but I put a lot of time into it, so I’m going to spend a few days explaining the research. 

              In pursuing the ultimate question of the study, you have to begin by asking whether the pitcher’s won-lost record IN EACH SEASON is a fair representation of how he has pitched.    In order to answer that question, you have to figure what his won-lost record SHOULD have been in that season—in other words, the won-lost record that he deserved, based on how he pitched. 

              To answer THAT question as well as you can answer it, you have to consider dozens of pieces of information about each pitcher’s season—his innings pitched and ERA, his un-earned runs as well, the league runs allowed rate, the park factors, but also you have to consider whether his Fielding Independent numbers (strikeouts, walks, hit batsmen and home runs allowed) are consistent with his runs allowed, or whether he may have been fortunate or unfortunate in his defensive support, or in the way that his run elements fell together into runs.  

              So—re-iterating that the question I am trying to get to here is whether or not the career won-lost record becomes a reasonably reliable measure of how a pitcher has pitched—in the process of answering THAT question, we have to answer hundreds of other questions as well, which are not our main focus but which are interesting questions as well.   Among those are:

              The luckiest and unluckiest pitchers ever (career),

              The luckiest and unluckiest pitchers ever (season),

              Pitchers who made the Hall of Fame basically because they were lucky,

              Pitchers who missed the Hall of Fame basically because they were unlucky,

              Pitchers who won Cy Young Awards they didn’t deserve because they were lucky,

              Pitchers who deserved the Cy Young Award, but were exceptionally unlucky,

              Pitchers who won 20 games due to luck,

              Pitchers who deserved to win 20 games, but pitched in tough luck,

              Historically significant pitchers whose won-lost records are accurate reflections of how they pitched, and

              Pitchers who pitched at a .600 level for their careers.  

              I’ll write about all of those things, I think, and also I will have to spend one day in this series of articles thoroughly explaining how the calculations were made, what the pathway toward a pitcher’s "deserved" won-lost record was.  

              But first, I am going to address the issue that I was trying to study, and I am going to address that issue today because I know that, when you publish a series of related articles, you can only lose readers as the series runs on, not add them.  I’m going to start by answering the question I was really trying to get to, because I don’t want to hemorrhage readers before I get to that issue.

              Let’s start with the assumption that there are three positions:

              1)  The traditional position is that a pitcher’s Won-Lost record, over the course of his career, is a reliable indication of how good a pitcher he was,

              2)  The Brian Kenny position is that Won-Lost records are useless and should be ignored, and

              3)  The compromise position is that Won-Lost records are generally accurate, but are not reliable in all cases.  

              Having studied this as thoroughly as I can study it, my conclusion is that the best answer is either (2) or (3), but that it isn’t (1).   It seems to me clear that, even over the course of a career, won-lost records are not so reliable that one can site them without concern for the possibility that they may be seriously misleading.

#

              I included in my study

              (1)  Every pitcher who had 150 games started in his career AND 150 decisions, plus

              (2)  Any pitcher who had twenty starts in that season.

              In other words, if a pitcher had 20 starts in a season, he was included in the study for that season, but not necessarily for his career.    If a pitcher had 150 decisions in his career but not 150 starts (like Hoyt Wilhelm or Goose Gossage or Rollie Fingers), he would be included in the study only in seasons when he made 20 starts.    If a pitcher had 150 starts in his career but not 150 decisions (like Al Fitzmorris or Ed Halicki or Brandon McCarthy) then he would be included in the study for those seasons in which he had 20 starts, but not other seasons.   But if a pitcher had 150 starts in his career AND 150 decisions, then all of his career was included in the study, even if he had no starts in a season.  

              There are 872 pitchers, I think, who had both 150 starts and 150 career decisions.   Of those 872 pitchers:

              52 were extremely lucky over the course of their careers, and have won-lost records which are clearly better than they deserved,

              177 were somewhat lucky, had records a little better than they deserved and notably so, but not enough that you could definitively say that the won-lost record was not a fair representation of their ability,

              429 had career won-lost records which were accurate representations of their ability as pitchers,

              193 were clearly unlucky, but not (in my judgment) so unlucky that you could definitely say that the won-lost record did not represent them, and

              21 pitchers were just snake bit, and did not have won-lost records that fairly represented how they had pitched.  

              Of the 872 pitchers, then, less than one-half had career won-lost records which fairly represented how they had pitched.   Only 73 had career won-lost records which were extremely different from how they had actually pitched, a little less than 10%.   That’s not enough to go all the way to the Brian Kenny position, but if a majority of pitchers have career won-lost records which are as much misleading as accurate, it is difficult to argue that the won-lost record should be relied on in describing a pitcher’s career.  

              If you are willing to accept that what I just said is true, if you are willing to take my word for it, then you can stop reading today’s article now; tomorrow’s will be more interesting.   But if you are curious as to what the dimensions are of my five classes, I’ll try to explain it.   My five classes are:

              1)  Extremely lucky,

              2)  Lucky (or somewhat lucky),

              3)  Fair.   Pitchers whose won-lost record is a fair representation of how they have pitched. 

              4)  Unlucky (or somewhat unlucky),

              5)  Snake bit (or extremely unlucky.)  

              How do we decide which of those five groups to put a pitcher into?   There is a formula, of course, which has arbitrary elements.   The formula puts the pitchers into a "luck line" based on how lucky they have been, and then there are cutoff points along that line by which we sort the players.   The cutoff points are also arbitrary, based on what seems to me reasonable.   The key point, then, is whether you accept that these designations seem reasonable.   

              We are trying to fit language and mathematics together.   I am trying to construct math that fits with the words we are using.   The first line is between "Extremely lucky" pitchers and "somewhat lucky".     What I should show you, then, is a group of pitchers who could be described either as "extremely lucky" or "somewhat lucky", the point being that if it seems reasonable to you to describe a pitcher as either one of those things, then I’ve hit our mark. 

              Tom Browning, who I believe was, all things considered, a slightly less than average pitcher, deserving of a won-lost record of 109-111, posted an actual won-lost record of 123-90, 33 games over .500.    I have marked him as extremely lucky in terms of his career won-lost record.    Jamie Moyer, who I believe had a "deserved" career won-lost record of 247-226, had an actual record of 269-209, 60 games over .500.    He was 60 games over .500 when he should have been 21 games over .500.    I think that was extremely lucky.  

              On the other hand, Deacon Phillippe was 80 games over .500, but deserved to be probably 44 games over .500.   I have marked him as "fortunate", but not "extremely fortunate".    Bob Forsch should have been 155-164; he was 168-136.   He was lucky, yes, but was he EXTREMELY lucky?

              Your answer to that question is as good as mine, of course; I am just trying to give you a sense of what these terms mean within the study, so that you can decide whether you want to accept the conclusions of the study, or want to debate them.  I marked Bob Forsch as "lucky", but not as "extremely lucky".   Feel free to do your own study and mark him however you want if you have the time.  There is a line there, separating Jamie Moyer from Deacon Phillippe, and there are 52 pitchers in baseball history who are above that line.   There are 52 pitchers in history who were as lucky as Jamie Moyer in terms of their won-lost record, or more so.  

              I’ll explain part of the system.    Deacon Phillippe, because he pitched for a great team (the Pirates of Honus Wagner’s era), had better won-lost records than he would have had if he were pitching for an average team, and with average luck.    He was a good pitcher, and I believe that he deserved to have a career won-lost record of 169-125, which is a very good record.    Because he pitched for outstanding teams, he was actually 189-109.   189-109 is 36 half-games, or 18 whole games, better than 169-125, so we credit him with 18 points on that scale.   He is +18; actually he is +17.79.    But since we all pay much more attention to career wins than we do to career losses and that process counts wins and losses evenly, we then count his wins AGAIN, so that wins count two and losses count one.  Phillippe won 20 games more than he probably deserved to win.   That’s +20, which makes him +38.  Carrying more decimals, he’s actually +37.60.

              Also, Phillippe’s career winning percentage, .634, is 58 or 59 points higher than the winning percentage that I believe he deserved, which would be .576.   If you save more decimal points, he winds up .0586 better than he should have been.   Multiply that by 200, and that makes 11.72.    We credit him with another 11.72 points for having a better winning percentage than he deserved.

              To say that Pitcher A has been luckier than Pitcher B could refer either to the Gross Impact of his good luck, or it could refer to the Proportional Impact of his luck.    Jim Palmer, also pitching for great teams, had a career won-lost record about 22 games better than the record he deserved, 268-152 as opposed to 261-189.   Spud Chandler, also pitching for great teams (the Yankees of the late 1930s and early 1940s) was about 15 games better than he should have been (109-43 as opposed to 104-68).    Chandler was the American League MVP in 1943.  

              However, because Palmer’s career was much longer than Chandler’s, Palmer’s "luck advantage" in terms of winning percentage is only 58 points (.638 vs. .580), whereas Chandler’s is 111 points (.717 vs. .606).    So who is "luckier"—Palmer, who has a 22-game advantage vs. Chandler’s 15, or Chandler, who has a 111-point advantage vs. Palmer’s 58?   Either answer is reasonable.    You really have to look at both of those things and combine them somehow, though, because if you don’t you’ll get rankings that just aren’t reasonable.          

              So we look at three things:  the "game difference", the "win difference" and the "winning percentage difference".   Combining those three things, Deacon Phillippe has a "Luck Score" of +49.3, whereas Jamie Moyer has a Luck Score of +50.04.   We draw the line at 50.   That’s why Phillippe is just "lucky"’, whereas Moyer is "extremely lucky".    You have to draw the line somewhere; that’s where I drew it.   The question I am asking is, does that seem to you like a reasonable place to draw the line?    If you accept the premise that Carl Mays, who was 207-126, should have been just 193-152, then does it seem reasonable to say that Mays was "extremely lucky"?   If you accept the premise that Juan Marichal, who was 243-142, has a deserved record of 228-171, then does it seem reasonable to say that Juan Marichal was "lucky, but not extremely lucky" in his won-lost record?  

              I’ll leave that up to you, but there’s another thing I need to explain.   I said there were three factors in the system, but actually there are seven, although the other four are minor, and effect a limited number of pitchers.  There are also bonus points:

              1)  A pitcher who has a winning career record but a less-than-.500 "deserved" record, like Browning, Blass or Forsch in the chart above, is charged an additional 3 "luck points" for crossing the line.

              2)  A pitcher who wins 100 games in his career but has less than 100 deserved wins, like Mark Mulder or Steve Blass, is charged an additional 3 luck points for crossing that line.

              3)  A pitcher who wins 200 games in his career but has less than 200 deserved wins, like Carl Mays in the chart above, is charged an additional 6 luck points for crossing that line.

              4)  A pitcher who wins 300 games in his career but has less than 300 deserved wins is charged an additional 9 luck points for crossing that line.     That’s a big deal, if you get to 300 career wins based on good luck.  

              OK, that’s all I have to tell you about the dividing line between "lucky" and "extremely lucky".   The next line I need to illustrate is the line between "his record is a fair representation of how he pitched" and "lucky".    In the chart below, Vida Blue, Pete Donohue, Mort Cooper, CC Sabathia and Nelson Briles are grouped as "lucky" or "Group 2", while Andy Hawkins, Fritz Peterson, Bill Sherdel, Mike Flanagan and Storm Davis are grouped as having fair career won-lost records, or "Group 3".  

              Storm Davis in 1989 was 19-7 although he probably should have been about 8-11 based on how he had pitched, a famously lucky season; we might see that season again on the chart of the luckiest seasons of all time.   But for the rest of his career, his luck was pretty poor, not terrible in any one season, but just generally poor.   He went 3-9 in 1991 when he probably should have been 6-8.    He went 9-12 in 1986 when he probably should have been 10-8.   By the end of his career, he wound up with pretty much the won-lost record that he deserved. 

              If you’re thinking that you don’t really see much difference between the luck of the top group in the chart and the luck of the bottom group. . .well, yes.   That’s the point.   The point is that you have to draw the line somewhere, but there aren’t actually "gaps" in the chart.  The dividing line is 20.00.   Briles is at 20.15.   Hawkins is at 19.88.   That’s where I drew the line, because that seemed to me to be a reasonable place to draw the line.   

              Group 3, which contains 429 of the 872 pitchers in the study, is pitchers whose career won-lost records are a fair representation of how they pitched.   There are 234 pitchers in the group whose won-lost records are a little bit better than their deserved records, and 195 whose records are a little bit worse, but one way or the other, their records are about what they ought to be.    Chris Carpenter had a deserved won-lost record of 150-110, but was actually 145-94; not a big difference.   Bruce Chen was 82-81 with a deserved record of 85-95.   John Tudor should have been 120-86, but was 117-72.     At the end of the series I’ll give you a full chart of all of the deserved and actual won-lost records.  

              Doc White, who pitched from 1901 to 1913, held the record for consecutive scoreless innings for more than 60 years, and became briefly famous as an old man when Don Drysdale broke his scoreless inning streak record in 1968.    He died a few months later.   I think he was actually a doctor; I think he was a dentist, like Jim Lonborg.   Not sure about that.   Not sure about Lonborg, either.   Anyway, White had a career record of 187-156—and a deserved record of 187-156, the longest career for anyone who exactly matched the won-lost record that he deserved.  

              Our next distinction is between pitchers whose records are a fair representation of how they pitched, and those who were somewhat unlucky.    As we drew that line on the other end at +20, we draw it here at -20, and as pitchers were charged "luck points" for crossing certain lines, they are given negative luck points for being on the wrong side of those lines.   A pitcher who deserves to have a .500 or better career record but does not, that’s -3; a pitcher who should have won 200 games but did not, that’s -6.    This is the chart that separates the "fair record" group from the "unlucky" group.  

              Passeau, Farrell, Redman, Harmon and Benz are considered to have "fair and accurate" career won-lost records; Randy Jones, Vern Kennedy, Joe Horlen, Kevin Brown and Dutch the Knuckler are considered to have been meaningfully better pitchers than their career won-lost records reflect.  Turk Farrell is the Storm Davis of this chart, although you have to be older to remember it.   Turk Farrell lost 20 games for an expansion team in 1962 although he actually pitched well.   He was the subject of an article that fall in the Saturday Evening Post, entitled "It Takes a Hell of a Pitcher to Lose 20 Games."   (That’s from memory, and I haven’t been able to confirm that such an article was actually published.  At the time, the Saturday Evening Post was pushing the envelope by using the word "hell" in the title of an article.)   Anyway, it was a famously unlucky season; Farrell went 10-20 despite an outstanding ERA and the third-best strikeout/walk ratio in the league.    He probably should have been about 17-11 that season, but then, he went 10-2 in 1957, and he didn’t deserve that won-lost record, either. 

              There are two Dutch Leonards in baseball history, both pitchers.   One of them was a Ty Cobb-era pitcher most famous for making baseless allegations against Ty Cobb and Tris Speaker, which many people still believe today because people like to believe the worst about others even with a complete absence of supporting evidence.  The other was a knuckleball pitcher in the 1940s, 1950s; in my data files I list them as Dutch the Accuser and Dutch the Knuckler.

              This is Dutch the Knuckler.   When I started this study, I sincerely expected Dutch the Knuckler to be perhaps the unluckiest pitcher of all time.   In 1948 Leonard went 12-17, leading the National League in Losses although he was second in the league in ERA, 2.51.    The next year he went 7-16 despite an ERA within a few points of the league norm, and in 1940 he had gone 14-19, leading the American League in Losses despite an ERA almost a run better than the league norm (3.49 vs. 4.38).   He led both leagues in losses in seasons when he actually pitched very well, and he won 191 games in his career; it’s not like he was only around for a few years.   Maybe he deserved to win 250? 

              I am always looking for hidden stars, always looking for guys who were really good, but whose records disguise their true ability.    I thought Dutch Leonard MIGHT be that guy; I thought he might have been the unluckiest pitcher ever to walk the face of the earth, over the course of his career.    But what I didn’t quite understand, until I did the work, is that almost all of Leonard’s big years were in pitcher’s parks, so his ERAs are a little misleading.   He was unlucky, yes, but not historically unlucky.  

              Our final dividing line is between pitchers who were unlucky, and pitchers who were REALLY unlucky.   As that line is drawn at +50 on the other end of the chart, it is drawn at -50 on this end.  

              Eckersley, Matlack, Newsom, Lockwood and Matt Young are considered "Unlucky" by this method, whereas Weiland, Nolan Ryan, Gaylord, Mike Morgan and Nap Rucker were "Snake Bit".    (There were also other pitchers who should be on this chart, but who I omitted because they pitched a long time ago and weren’t very good, and modern readers wouldn’t recognize their names.)   Bob Weiland was a 1930s guy who started his career by doing a tour of the worst teams in the American League.   He was wild at that time, and at one point his career won-lost record dropped to 21-57, which, you know he had a good arm, because you don’t hang around long enough to go 21-57 unless you have a great arm.   He finally got to pitch for the Cardinals in the late thirties, where he won 15 and 16 games.

              Nolan Ryan is, with Pete Rose and Barry Bonds, among the most controversial, spectacular, memorable and unusual players of my lifetime.    Although he was only 32 games over .500 in his career, he is regarded by some people as the greatest pitcher of all time.  

              It is beyond the scope of this article to sort all of this out, but regarding Ryan, there is an argument which goes as follows.   Ryan had spectacular positives, yes; he struck out 2,000 more hitters than anyone had before him, and still holds the career strikeout record by a huge margin.  He individually pitched more no-hitters than some entire franchises which have employed many hundreds of pitchers.   Literally.    Some years he would strike out 10 batters in a game more times than the rest of the league combined; I don’t know whether that one is literally true, but it seemed like it.   He led the league in strikeouts by more than 100 several times, and he was still throwing no-hitters and leading the league in strikeouts when he was in his 40s.    He would throw 200, 220 pitches in a game, come back three days later and throw 180.  

              But—the argument goes—his negatives are as impressive as his positives.   He has almost a thousand more strikeouts than any other pitcher—and almost a thousand more walks.   The percentage margin by which he holds the walks record is much higher than the percentage margin by which he holds the strikeout record.    He threw more wild pitches than any other pitcher ever, 50 more wild pitches than any other pitcher since 1900.     He committed more errors than any other pitcher of the last 100 years, regularly leading the league in errors.   In 1977 he walked more than twice as many batters as any other pitcher in the league; in 1976, almost twice as many.  He allowed more stolen bases than any other pitcher ever, often leading the league in stolen bases allowed by ridiculous margins.   People think Maury Wills should be in the Hall of Fame as a base stealer, but Ryan allowed almost 200 more stolen bases than Wills’ took.  

              It is not difficult to understand how he won 300 games, the argument goes, and not difficult to understand how he lost almost as many as he won.

              Sabermetrics searches for the balance of the positives and the negatives.   A colossal figure like Nolan Ryan presents a colossal challenge for us.    While Ryan was active, I had more faith in won-lost records than I do now and less ability to see the balance, and I generally endorsed the anti-Ryan argument.  

              The anti-Ryan argument is not an empty vessel, but with the passage of time I have come to see that Ryan was also unlucky, and that he was meaningfully better than his won-lost records.   In 1973 he won 21 games and lost 16, but his team started four regulars with slugging percentages under .300, eight regulars with slugging percentages less than .370.   It’s difficult to "win" consistently with a team like that, and here’s the kicker.   That team scored 173 more runs than they had the year before, when Ryan was 19-16.  In 1972 they scored 454 runs.   All year.  

              Studies like this one (and others) have helped me to see that the balance of Ryan’s positives and negatives was better than I thought it was at the time. 

              Gaylord Perry went 24-16 in 1972, but by rights, he should have been 27-12.   That season would be more famous than it is, had not Steve Carlton trumped it.   Gaylord went 15-17 in 1967, when he should have been 20-14.   He had one season in his career when he won 20 games by good luck, but four seasons when he should have won 20, but didn’t.  

              Ryan and Perry were unlucky pitchers, but not the unluckiest of all time.   Of course, the traditionalist is entirely free to reject my conclusions.    "Bill James says that Jamie Moyer was extremely lucky," the traditionalist might say, "and that his won-lost record does not represent his actual performance.  But Moyer won only 7% more games than James says that ought to have won, and how do we know that that discrepancy was all luck?   James says that Nolan Ryan was unlucky, but Ryan won only 7% fewer games than James says he should have won.  Seven percent doesn’t seem to me to be a huge distortion.   James says that there are 73 pitchers in his study who have won-lost records in no way reflecting the actual skill of the pitcher, but it seems to me that the number might be a third of that.  His study shows, if anything, that pitcher’s won-lost records ARE reliable."   

              The traditionalist can say that, and there is no way I know of to prove that it is untrue.