Thoughts on Pitcher Wins

I got involved in a semi-heated discussion with my friend Brent Cochran (aka @BrentCochran1) on Twitter about the “Win” statistic for pitchers. While we both agreed that it is an overrated statistic that for too long has held unreasonable sway over Cy Young voters, we disagreed over whether the statistic had any value at all. I maintain that it is not only a completely valueless statistic, it is an actively misleading one and we should discontinue referring to it at all.

With apologies for the esoteric nature of this post, I am a giant baseball nerd and besides which I apparently cannot abide someone being wrong on the Internet. Therefore, let me explain why we ought to throw the pitcher “win” statistic out the window entirely.

The “win” as a statistic, is based on an errenous belief, popular among television baseball commentators, that the starting pitcher holds an enormous amount of sway over whether his team wins a game or not. Relative to the other players on his team, this is true; relative to 100%, he does not. Let me explain.

To win in baseball, you must do two things: 1) score runs, and 2) prevent the other team from scoring runs. Roughly speaking, scoring runs is about 40% of winning and preventing runs is about 60% of winning. These figures are arrived at through approximations of baseball’s Pythagorean win probabilities, which can predict with remarkably good accuracy what a team’s record will be based on how many runs a team scores and allows.

Oversimplified, the theory states that if you start with a team that scores, say, 750 runs and allows 750 runs, then that team would expected to be a .500 baseball team. As they score more runs and allow less, their record is expected to go up in predictable increments, and vice versa. A relatively robust amount of study suggests that preventing runs is, on the whole, slightly more important to a team’s success than scoring them.

Likewise, a pretty good amount of study suggests that, on average, run prevention is about 70% attributable to the pitcher and about 30% attributable to the defense. Thus, roughly speaking, if 60% of winning is run prevention and 70% of run prevention is pitching, then pitching is roughly 42% of winning. Assuming a modern environment in which starting pitchers are responsible for about 60% of innings pitched, then starting pitchers are responsible, on average, for about 25% of a team’s chances of winning.

That’s not to minimize the role of the starting pitcher; quite to the contrary, relative to his teammates, he (on average) has a huge influence on his team’s win probability. Assuming a fairly typical modern AL game in which the pitcher leaves after the 6th inning and is replaced by 3 one inning relievers in succession, and there are no defensive substitutions, then a team’s contribution to win probability might look something like this (assuming, erroneously, that each position player contributes equally to his team’s defense. Yes, it is an erroneous assumption, but we don’t want this to get too complicated):

• Starting pitcher: 25%
• Each position player: 6.7%
• Reliever 1: 5.5%
• Reliever 2: 5.5%
• Reliever 3: 6.0% (we give this guy a bonus for leverage)
• DH 4.4%

So, relative to his teammates, the starting pitcher is by far the most important player on the field, which comports with the sense that we get when we are watching the game. It is the reason that, even in the modern age, starting pitchers get roughly as much money for what they do even though they only play in 20% of a team’s games; even in their limited playing time they exert tremendous influence on the course of their team’s fate.

However – and this is a big however – it is still important to realize that, on average, the pitcher “win” statistic is based on factors that the player we are attributing it to has only 25% control over. I say “on average” because the vicissitudes of fate and the bizarre rules of baseball sometimes decree that a relief pitcher who gets one out gets tagged with a win when he has in reality only exerted 1% of control over his team’s win, and sometimes a pitcher gets a win when he has pitched a complete game and thus is roughly 42% responsible for his team’s win (it gets more complicated than that – stay tuned).

I maintain strongly that any statistic that is less than half under the control of the player we are attributing it to is inherently misleading and should be completely discarded from use. And baseball history is replete with instances of how misleading wins can be in terms of measuring pitcher performance. By wins alone, Felix Hernandez’s 2010 Cy Young campaign looks almost exactly equal to John Lackey’s 2011 campaign, which was one of the worst full seasons turned in by a starting pitcher in Major League history. Randy Johnson and Don Sutton had roughly equivalent careers in terms of length. Don Sutton had 20 more wins, when in fact he was a vastly inferior pitcher. And so on and so on.

Now it is important to remember that no statistic is completely within the control of the player. All batter statistics – in particular home runs – are affected by a variety of factors related to the parks they play in – including variable park size, climate, etc. Pitcher ERA is dependent to a large degree on the quality of the defense behind him (not in terms of errors, which are of course excluded from ERA, but the more prosaic and far more common variables with respect to which balls his defenders reach at all).

Probably the most player independent statistic is the strikeout – for both the hitter and pitcher. In every park, the mound is exactly the same and the same distance from the plate, and weather factors (other than in Colorado) are generally considered to not affect pitching enough to measurably alter K rates. But even here, a pretty good amount of data shows that the quality of a catcher as a receiver can influence a pitcher’s success in achieving strikeouts – I recall seeing an analysis that I can’t find now that David Ross earned Jon Lester something like 15 extra strikeouts last year due to his skill as a receiver.

Further, with unbalanced schedules, both hitters and pitchers can be expected to have some marginal impact on their K rates just based on who they face. For instance, in the mid- and late-2000s, pitchers for the Baltimore Orioles had to face the combined orders of the Red Sox, Yankees, and Rays a disparate number of times a year – teams that were notorious for cultivating batters who were skilled at drawing walks and avoiding Ks.

But these differences are marginal and exist around the edges. If Jon Lester were forced to throw to Ryan Doumit (a terrible receiver) the difference in his K rate would be less than 10% (probably less than 5%). Similarly, quality of competition, over the course of a season, levels out to account for a very small percentage of a pitcher (and a hitter)’s K rate. So these statistics, while not perfect measures of a pitcher’s performance, are a good measure of his performance, or at least some important aspect of it (and let no foolish TV personality convince you that strikeouts are not important for a pitcher – they are hugely important, vital to a pitcher’s long term success, and a vast amount of study confirms this).

So, no one is saying, if you look only at Ks, you can figure out who the best pitchers necessarily are. Nolan Ryan was not the best pitcher in Major League history, or even close. However, by looking at Ks, we learn something important about a pitcher’s performance that gives us useful data about how good of a pitcher he was.

Looking at wins, on the other hand, we learn useful data mostly about the team around the pitcher. So much of the stat is dependent on other players that its inherent tendency is to mislead rather than to inform.

Which brings this long and convoluted post around to the point of the discussion that Brent and I were having on twitter: who should win the Cy Young award for the NL this year? Traditional analysis would say that either Jake Arrieta (who had a sparkling 22 wins) or Zack Greinke (who had a very stellar 19 wins) should win. I maintain that Clayton Kershaw, who sported a relatively pedestrian record of 16-7, was far and away the best pitcher in the NL this year, with all due respect to Mr. Arrieta, who had a fine season and will probably win the award for sentimental reasons without too much protest from me.

The reason for this is fairly simple: a pitcher has essentially three jobs, in descending order of importance:

1. Do not walk batters – this is the most harmful thing to a team’s run prevention efforts and each walk must be offset with more than one strikeout to give a pitcher’s team a chance.
2. Don’t allow home runs – In theory, this is more important than number 1, but a pitcher who allowed so many home runs that it would outweigh the damage of walks would also probably be so terrible at achieving strikeouts that it would not matter. Thus, the actual difference in allowing home runs among the caliber of pitcher who is likely to actually pitch in the majors makes this less important than not allowing walks.
3. Strike out hitters. A lot of study shows that a pitcher simply cannot succeed without striking out at least5 batters/9 innings, and then only if they never walk anyone and induce a ton of ground balls (and probably they also need to be left handed).

So here’s why Kershaw (whose K and BB numbers are both better than Arrieta and Greinke, by a measurable margin) was a more important pitcher to his team’s success this year than either Arrieta or Greinke.

Recall that the average starting pitcher figure of 25% influence on a team’s success is based on two assumptions – 1) pitching about 60% of the innings in the games you start and 2) exerting 70% influence on your team’s run prevention efforts. The two easiest ways a pitcher can increase his influence on his team’s success is to 1) pitch more innings and 2) strike out more batters (without walking any more). Pitching more innings means you get credit for more of the 42% of the game attributable to pitchers overall, whereas striking out more batters means that you are exerting more control over run prevention personally – the 70% figure being an average that is based on an assumption of league average strikeout rate. So, when Jacob deGrom struck out a billion Dodgers the other night, his influence on his team’s run prevention was worth far, far more than the 70% average and thus his overall influence on his team’s win probability was much higher.

That is a major part of the reason why Fangraphs sees Kershaw as far and away the best pitcher in baseball this year (better than Arrieta by about a game and better than Greinke by a huge margin) – it is worth noting that Baseball Reference sees Greinke as the most valuable of the three because they don’t (I assume) dock pitchers wins for factors attributable mostly to luck like LOB%. But in terms of the actual merits of what the pitchers contributed to their teams, Kershaw’s numbers stand out.