The Anatomy of a Pitcher Win

At the end of the 1972 season, Steve Carlton was awarded the NL Cy Young award for his efforts with the Philadelphia Phillies. He certainly earned it. He posted a 12.1 WAR, a 22% strikeout rate, and an ERA- of 56. What many voters and fans were looking at, however, was his wins — 27, the most in the National League.

That number is impressive for a different reason to my father. Steve Carlton’s 27 wins in 1972 are important to my dad because it is included in one of his favorite baseball tales, a not-so-secret weapon for dads — Steve Carlton won 27 games in 1972, the same year the Phillies won all of 59. For those keeping score at home, that means that Steve Carlton was awarded a pitcher win in 46% of the games that the Phillies won all together.

This is more in the “fun fact” category than the “useful information” one, but it’s an eyebrow raiser nonetheless. Still, I needed to set the record straight and, during the 14th or 15th time my dad relayed his trivia nugget I, politely, mentioned that pitcher wins perhaps aren’t as important of a stat as we once thought. This led to a discussion involving all the usual stuff statistically-minded people mention when discussing wins. We sort of agreed to disagree, I guess, but something still stuck with me.

I began pondering pitcher wins some more. This is something that probably should very rarely be done, but hey, it’s January. After a while, I found a nasty thought materializing in my head — a thought that went against so many things I’d been hearing and saying for so long. Good pitchers might not always get wins, but do winning pitchers generally do well?

Famed stats person Brian Kenny has been pushing his “kill the win” campaign for some time now. I get the idea behind it, and I generally agree with it. Throughout last season he famously tweeted stat lines in which a good performance did not get a win for a certain pitcher, or vice versa. And while the logic is sound — using this fairly outdated statistic is not the best measure for a pitcher’s talent — it tended to cherry pick a bit. Whether its pitching or the catalog of a band or the quality of a grocery store’s tomatoes, there will always be outliers — warts and blemishes that look bad close up, but get washed away in the aggregate. So, out of sheer curiosity, some digging was done.

A note on the process: I found all the pitching performances from 1974 (the farthest back we have reliable data for such things) that resulted in wins or losses, then averaged some basic stats from those performances. I also took the league average for each corresponding year as a comparison. I used a 5IP threshold to negate vulture wins by relievers. I realize that this may help the losers group a little, since we won’t see the truly awful losses that led to pitchers getting the hook early, but we’ll have to make due.

The chart is somewhat interactive, as you can select different tabs, single out certain years, and show and hide certain filters. Play around with it. I’ve added no-decisions to the mix as well, but hid them by default. Click on the measure name to activate it. The large version is available here.

I won’t insult anyone by relaying everything the chart tells us, but the general point is there. A pitcher who logs a win historically, on average, did better than league average on that particular day. Though, he didn’t really do that much better. The biggest gap occurs in ERA, which sort of makes sense. A high number of runs means the pitcher is getting pulled and probably won’t be in line for a win. A winning pitcher does have a higher K%, but only by about 2% at the most extreme. That’s the difference between Matt Harvey and Clayton Kershaw. The same goes for walk rates — 1% to 2% difference. Winning pitchers pitch longer into games and are generally more efficient in that game, as well.

Look how all those lines mirror each other so well. There really isn’t a year in which the losers were way more terrible than average, or the winners that much better. They just kind of coast along, varying just a bit on either side of the orange line. Winners, losers, it doesn’t matter. They all just teeter slightly on one side of the league-average fulcrum.

Every loss that a bullpen blew for a starter, every cheap win a pitcher got thanks to a high-powered offense, it all gets blended away. The Steve Carltons or Max Scherzers of the world — dominating pitchers with the win total to accompany it — get dragged back down to Earth by the bottom of the average. No one is safe from the anchor of mediocrity.

No analysis on pitcher wins is going to win me any awards, and this might rank at the very bottom of even that in terms of usefulness. This is simply for fun, at least if you call database querying fun. I do. I’m sure none of you needed any more fuel in your arguments against the merit of pitcher decisions. But it’s a fun thing to look at and ponder. A stat that many fans and writers — and even pitchers themselves, for that matter — care about hang on the thinnest of threads above the boringness of the norm. A percentage point here, a half-percentage point there. Think of all the big contracts handed out on win numbers, the awards, too. Think of all the fantasy leagues won and lost. We always knew that a pitcher win (or loss) has a lot to do with luck. The numbers show it has a lot to do with doing just slightly better than league average, too.

Print This Post

David G. Temple is the Managing Editor of TechGraphs and a contributor to FanGraphs, NotGraphs and The Hardball Times. He hosts the award-eligible podcast Stealing Home. Dayn Perry once called him a "Bible Made of Lasers." Follow him on Twitter @davidgtemple.

21 Responses to “The Anatomy of a Pitcher Win”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. DD says:

    Could you just look at the R squared for Wins vs. ERA- or RA/9 or something?

    Vote -1 Vote +1

  2. daniel says:

    If you use a 3 IP threshold, which would still negate almost all vulture wins, you would get a much greater difference in the statistics of pitchers who earn a W or an L next to their names. Almost every starter who last 5 innings has pitched somewhat decently, so you are not going to get that much variation between winners and losers.

    Vote -1 Vote +1

    • Iron says:

      Less difference than you’d think. It would have no effect on the winning pitchers still only being a hair above league average. It would have a very small impact on losing pitchers, because games in which the losing pitcher pitchers 3-4 innings are only a relatively small percentage of games, too small to have a huge impact.

      Vote -1 Vote +1

  3. KCDaveInLA says:

    A great read and terrific research. I had always liked to think that pitcher wins are not totally meaningless (even though I’ve been reading Fangraphs for a while). The problem as with any stat, is that if you isolate it and judge a player by it, you lose a lot of context. Although I’m sure Jack McDowell and Bob Welsh are grateful for that lack of context.

    Vote -1 Vote +1

  4. Ian R. says:

    Good analysis. I’ll repeat an old Bill James (I think?) thought on this one – there’s no such thing as a truly meaningless stat. Pitcher wins and losses mean something. If you ask me to pick between Pitcher A and Pitcher B based purely on their W-L record, I’ll pick the pitcher with the better record every time, and more often than not, I’ll come away with the better pitcher.

    The issue with the win is just that it’s called, well, a “win.” Even enlightened fans put an inordinate amount of weight on it because, hey, baseball is all about winning. Even calling it a “winning effort” or something might be preferable.

    +6 Vote -1 Vote +1

    • Peter B. says:

      I agree with you on the “not a useless stat” point. I think that a lot of people negate the Win’s value because there are good performances that don’t get a pitcher the win, and bad performances (“bad”) that do get a pitcher the win. What I have thought about the Win as a stat since I really started thinking about it is that a Win is just like any other stat- in small sample sizes, it is basically useless, but over the course of a long time they do become a reliable indicator of some talent. For example, Justin Verlander won 24 games in 2011. Is he a true-talent 24-game winner? Probably not, and because of all the noise surrounding wins, we don’t really know what that looks like. However, Greg Maddux won 355 games over the course of a 23-year career. Is he a true-talent 355-game winner? Probably. The 8 IP/ 0ER/ 10 K no-decisions and the 5.2 IP/ 6ER/ 3K wins all even out over the course of a career.

      Vote -1 Vote +1

      • Simon says:

        But the offensive support that a pitcher gets will not. Elite players tend to spend most of their careers with two or three clubs, and if those clubs are good, it will give a significant boost to their wins total which has nothing to do with the pitcher’s skill.

        Vote -1 Vote +1

  5. snydeq says:

    Interesting analysis.

    Reverse-engineering the data from 2013, here’s the “average win” vs. “average loss” lines:

    Win (26.4 BF): 6.68 IP, 4.67 H, 1.4 ER, 5.5 K, 1.69 BB, 0.41 HR
    Loss (26.0 BF): 5.89 IP, 6.51 H, 3.45 ER, 4.5 K, 1.82 BB, 0.94 HR

    So that’s roughly the difference of 2 outs (one by strikeout) vs. 2 hits (with a double likelihood for those hits being HRs).

    Or, looking at K%, BB%, HR/9, and, just for fun, filtering over the course of the season, you could say your average winning SP pitched like Clayton Kershaw, Matt Harvey, or Adam Wainwright that game, while your average losing SP pitched like Bronson Arroyo, Jeremy Guthrie, or Kyle Lohse:

    Win: 21.0 K%, 6.4 BB%, 0.55 HR/9,4,5,13,6,45,62,120,121,40&season=2013&month=0&season1=2013&ind=0&team=0&rost=0&age=0&filter=11452&players=0&sort=11,a

    Loss: 17.6 K%, 7.07 BB%, 1.43 HR/9,4,5,13,6,45,62,120,121,40&season=2013&month=0&season1=2013&ind=0&team=0&rost=0&age=0&filter=11453&players=0&sort=11,d

    Vote -1 Vote +1

  6. ksclacktc says:

    Something I never hear anyone speak of is the fact that a Pitcher’s Won-loss record becomes more reliable for a career. Therefore, at the career level it can have some real meaning. Given other available stats, it may not be the best. Never the less, would you rather know a pitcher’s career won/loss record or his K/BB ratio? The won-loss record is more context-neutral, and thus more useful if other stats are not available. Bill James talked about this some year’s back. The significance of Games Played at the career level, and a pitchers won-loss record.

    Vote -1 Vote +1

    • Ian R. says:

      In general, yeah, career W-L records are more useful than single-season W-L records. That’s because a lot of the noise in single-season records (varying run support, team defense, etc.) tends to even itself out over the course of a career. There are still outliers, though – some pitchers had the misfortune of playing for bad teams throughout their careers, and some had the good fortune of playing for good teams throughout their careers.

      I read somewhere that career W-L record is roughly as useful as career non-adjusted ERA. In other words, it’s meaningful, but there are tools (like ERA+ and RA9-WAR) that are better.

      Vote -1 Vote +1

    • Los says:

      Except for Bert Blyleven. I love his 1972 season where he was worth 11 wins but finished 20-17 on an 81-81 team. His overall W/L record was only .537 for his career as well.

      Vote -1 Vote +1

  7. Brian Kenny says:


    Vote -1 Vote +1

  8. baruchk says:

    as has already mentioned, there’s alot of bias here. shorter outings by starters are almost always negative appearances, and disregarding them. For an example of what impact these could have, just think about that time, late in the season, that your starting pitcher on your fantasy team had a 2 inning, 8 run outing, and how much your era got affected from 2 innings even after having over 1000 innings to your credit (yes, this has happened to me.)

    Vote -1 Vote +1

  9. Juan says:

    I believe Mr. Temple has over-analyzed this “Wins” Article.
    That happens at times to persons overzealous about stats.
    They cannot understand why teams that had all the stats
    (i.e. the 1960 Yamkess,1969 Orioles, and 1970 Big Red
    Machine)could not win the World Series. But, that was
    because the opposing teams had much better pitching.

    To people who are not familiar with Carlton (a Cy Young
    Award Winning pitcher), he had just been traded to the
    Phillies in 1972. What is significant about his 27 Wins
    (AND 30 COMPLETE GAMES THAT YEAR)is that he pitched for
    a last place team that season, a team with a .236 team
    batting average (10th in a 12 team league), 503 Runs
    Scored (11th), 98 Home Runs (10th), and 469 RBIs (10th).
    With this type of offensive stats, most ace pitchers
    would feel lucky if they won 10 games.

    However, the question that should be asked is how many
    games Carlton would have won if he’d pitch for the
    National League Champion Reds, or even a lesser team
    with an average offense.

    Carlton was every bit as good as the great Sandy Koufax.
    But, because Carlton did not pitch for a team based in
    Los Angeles or New York, he never got the fame. Of
    course, it did not help that he quit giving interviews.
    But, Steve Carlton had the same blazing fastball and
    devastating overhand curve (screwball) that Sandy had.
    And, their mentality was that if their teams could not
    score, neither would the opposition.

    Vote -1 Vote +1

    • Ian R. says:

      “With this type of offensive stats, most ace pitchers would feel lucky if they won 10 games.”

      This is a common fallacy in examining Carlton’s 1972 season. It doesn’t especially matter how bad the Phillies were in general. What matters is how well they hit when Carlton was pitching.

      Overall, the Phillies scored just 3.22 runs per game, but during Carlton’s starts, they averaged 3.88. The National League averaged 3.91 runs per game in 1972, so Carlton received almost precisely league-average run support.

      Steve Carlton had a legitimately fantastic season in 1972, but don’t act as though he did it in spite of a horrible offense. On the days he was pitching, the Phillies hit like an average offense.

      Vote -1 Vote +1

      • Ian R. says:

        Typo in the above comment – the Phillies actually averaged 3.83 R/G in Carlton’s starts, which is still close enough to league average to be basically the same.

        Vote -1 Vote +1

  10. bob says:

    It would be nice if the offense produced by the pitcher’s team could be taken into account to estimate the team’s chance of winning with an average pitcher, then determine if the actual outcome shows if the pitcher in question increased or decreased the team’s number of wins with any statistical significance. Or maybe there is a stat for that already, I haven’t studied this enough.

    Vote -1 Vote +1

  11. Ben says:

    Isn’t this expected? That in the long run, pitchers who get a win pitch better than pitchers who do not? I think even if you took all pitchers from one season, this would also be the case. Wins don’t tell us *nothing* just very very little that is of any use when we have several other metrics that do a far better job.

    Vote -1 Vote +1

  12. Trey says:

    Great article, although i believe in killing the win also

    Vote -1 Vote +1

  13. Hurtlockertwo says:

    I did a quick count of Pitchers in the top five in wins versus the top five in WAR since 1901. 230 of the pitchers that had the most wins in any given year were also in the top five in WAR. (230 out of 560, or 41.07%)
    Just in this simple statistic it seems that wins don’t equal the value of the pitcher about 60% of the time.

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>