It has long been common wisdom that breaking balls damage pitchers’ arms. More so, breaking balls are supposed to be especially bad for young pitchers, whose bodies are still developing, and not yet ready to handle the stress of a curveball or a slider.
It was also once well-known that the earth was flat. My point is that all ideas, no matter how intuitive, need to be examined. Unfortunately, up until a few years ago, there was no way to question this tenet, because there was no resource that gave us detailed information about the repertoires thousands of major league pitchers throughout history.
Then came the Neyer/James Guide to Pitchers. This is a fabulous resource, a monumental achievement which includes a plethora of extremely useful information. Unfortunately, that information has not been utilized much in the sabermetric community.
Let’s take a step towards changing that, and find out if breaking balls actually do harm to young pitchers. First, we need to determine what we mean by “young.”
Luckily, Nate Silver and Will Carroll have already done that work for us. Silver and Carroll found that pitcher attrition rates are extremely high prior the age of 23, and then they drop like a rock. What that means is that pitchers are highly susceptible to injury before turning 23, and much more stable after. Thus, Silver and Carroll call 23 the “injury nexus” (also the number of ingredients in Dr. Pepper). If a pitcher can reach the injury nexus healthy, he is unlikely to suffer a future injury, but if he does not, his career may well be ruined.
So what I did was looked at how many pitches every pitcher between 1955 and 2005 threw before his age-23 season, using Tom Tango’s expected pitch count estimator. For the sake of examining pitchers whose major league workloads at that age would have been considered pretty heavy, I drew a cutoff point at 5,000 pitches, or about 320 innings. That left us with 96 pitchers.
I then entered pitch selection information for each pitcher, and removed any pitcher not included in the Neyer/James Guide or who was still active. After that paring down, we were left with just 75 pitchers.
Sixty-one of those 75, or more than 80%, threw a breaking ball as one of their top two pitches when they came into the major leagues. 14, therefore, did not.
That was the basis of our comparison groups. What next? First, I calculated each pitcher’s ratio of pitches thrown after turning 23, and before. I then ran a regression trying to predict that ratio based on whether a pitcher relied heavily on his breaking ball. The results were mixed.
On the one hand, breaking balls had a strong negative effect on the number of pitches thrown after the “injury nexus,” meaning that breaking ball pitchers were expected to miss more playing time after the injury nexus than guys that did not rely on their curveball or slider. On the other, statistically, this result was not significant. Specifically, the p-value, which is simply the probability that we could have obtained such a result due to random chance alone, was .153, whereas the general statistical threshold is .05.
However, I am not too concerned about the p-value (if you don’t care about the math, by the way, skip this paragraph). First off, p-values are generally measured using a two-tailed distribution, whereas a one-tailed distribution is clearly called for here, given that our alternative hypothesis would be that breaking balls have a negative effect on a pitcher’s longevity, not just some effect, period. So that cuts down our p-value to .077. Okay, so technically that is not below .05, but it’s close enough, I think. Given how arbitrary the .05 threshold is, and given everything we do know here, I feel comfortable calling this result significant.
Of course, regression analysis is not the only statistical test in our toolbox. We can also do something called a “one-sample T-test,” which measures whether our sample differs meaningfully from the expected mean. In plain English, we can examine if the difference between the ratio of pitches after the “injury nexus” to pitches before age 23 is meaningfully different for breaking ball pitchers and non-breaking ball pitchers.
We find that non-breaking ball pitchers throw, on average, 382% more pitches after turning 23 than before, while breaking ball pitchers throw 311% more. That’s a pretty big difference, but is it significant? The T-test says yes; the odds of such a big disparity occurring by random chance alone are just 3%.
Better still, if we look at the results of the regression and the more direct T-test, we find that they both spit out a very similar answer in terms of how much playing time pitchers lose by throwing a lot of breaking balls at such a young age. Breaking ball pitchers throw 20% fewer innings than non-breaking ball pitchers after the “injury nexus,” according to both tests. That’s equivalent to the difference between a healthy season, and missing five weeks on the disabled list!
It seems pretty clear that breaking balls hurt pitchers’ arms, but there are some caveats:
- Seventy-five pitchers is not a huge sample. I had to balance my desire for a sample of pitchers who threw a meaningful number of innings before crossing the “injury nexus,” and my desire for a larger sample size. If we could somehow include minor league statistics in this study, it would expand greatly, and perhaps offer more accurate answers.
- By choosing pitchers who threw a lot of pitches in the major leagues before turning 23, I may have introduced a bias into the study, since we’re looking at a very good group of pitchers overall. What effect that bias might have had, I don’t know, but it’s just something to keep in mind.
- Given the size of the sample, and given that you could slice it and dice it a number of ways, these results should not be taken as gospel. They are at best a theory, a starting off point that can and should be examined more thoroughly.
But caveats aside, I think we may well have found the statistical effect of breaking balls on pitchers’ arms, and the effect is great. It seems that believing in the destructive effect of breaking balls is not quite so crazy as believing that the world is flat.