Last week I wrote about how losing velocity at different ages impacts a pitcher’s chance to throw 40 or fewer innings the next season (what I labeled “attrition”).
The overall finding was that losing velocity at any age increases the likelihood of attrition for pitchers, and that likelihood only increases with age. Overall, pitchers in the data set had a 29% chance of attrition between years one and two. If they lost at least 1 mph on their fastball, however, that rate jumped to 39%. Pitchers that didn’t lose at least 1 mph only had an 18% attrition rate–so half the odds. Starting at age 34, the attrition rate jumped to 50% and climbed for each age cohort until roughly age 39. (Thirty-eight-year-olds who lost velocity magically bucked the trend, attriting at about the same rate as all other pitchers.)
Eno Sarris asked me whether, as some have suggested, pitchers who relied on a change-up as their primary secondary pitch (such as a James Shields or Mark Buehrle) gained some kind of advantage, in terms of attrition. Do these pitchers have a lower chance of injury or ineffectiveness than someone who relies heavily on either a curveball (e.g. A.J. Burnett) or a slider (e.g. Ervin Santana)?
This is, of course, part of a broader question around pitch usage and whether some pitches are “better” for pitchers to rely on than others. Jeff Zimmerman’s research into injuries has shown that extreme use of a curveball or a slider does increase a pitcher’s chances of landing on the disabled list the following year. And, of course, injuries are a big reason that a pitcher might fail to throw more than 40 innings in a subsequent year.
To tease this out I used a similar methodology as the earlier velocity analysis. I collected all the seasons where a pitcher threw 40 or more innings and then looked at whether they managed to hit that same threshold the following year. I used FanGraphs’ pitch-type data to code a pitcher as either throwing a slider, curveball, change-up or split-finger fastball as their primary secondary pitch*. I then calculated attrition rates for each type of pitcher. Finally, I calculated odds ratios to get a better sense of whether throwing different types of pitches led to an increase or decrease in the likelihood of attrition. This was done by simply taking the attrition rate for a category and dividing it by the attrition rate for all other pitchers in the sample.
If we look at the difference in attrition rates for pitchers who throw a certain secondary pitch, versus all other pitchers, we find that the rates are basically indistinguishable.
|N||Attrition Rate||Comparison N||Attrition Rate||Odds Ratio|
For example, throwing a slider as your secondary pitch doesn’t negatively impact your chances of attrition. The odds ratio for slider pitchers is slightly higher, but we are talking about .03 times higher. Similarly, change-up heavy pitchers don’t seem to gain an advantage as some suggest.
Now we might be tempted to walk away thinking it really doesn’t matter what kind of secondary pitch someone throws. But that would be inaccurate. It may not just be what kind of secondary pitch your throw, but how often you throw it. Here is where it gets interesting:
First, let’s look at how often any secondary pitch is used. I created “buckets” of pitchers based on the percentages of their secondary pitches. We see that as the reliance on a single secondary pitch increases, so too does the pitcher’s likelihood of attrition:
|N||Attrition Rate||Comparison N||Attrition Rate||Odds Ratio|
|Any 10% – 15%||467||28.1%||2834||29.6%||0.95|
|Any 15% – 20%||910||27.5%||2391||30.2%||0.91|
|Any 20% – 25%||812||29.3%||2489||29.4%||1.00|
|Any 25% – 30%||514||37.5%||2787||27.9%||1.35|
|Any 30% – 35%||245||32.7%||3056||29.2%||1.12|
|Any 35% – 40%||138||33.3%||3163||29.2%||1.14|
|Any >= 40%||104||40.4%||3197||29.1%||1.39|
Even without knowing what kind of pitch someone is relying on, we can see that the odds that they will fail to throw 40 innings in a subsequent year increase the more they become a two-pitch pitcher.
But what about pitch type? Are sliders and curveballs more dangerous than change-ups?
The short answer is yes. As reliance on sliders and curveballs as the primary secondary pitch increases, pitchers subsequently increase their odds of attrition the following year. The graph below plots the odds ratios for the various categories of pitchers, including by pitch type:
Overall, pitchers rely more heavily on sliders than any other secondary pitch. As a result, we can see how the change in odds ratios for slider usage map well onto our earlier look at overall secondary pitch usage. By breaking up the sample by pitch type we can see some interesting differences. The most striking pitch is change-ups. Pitchers who rely more heavily on change-ups drive down their odds of attrition, while slider-heavy and (especially) curveball-heavy pitchers see their odds increase.
Of course, at the extremes of usage we see sample sizes fall pretty dramatically. For example, we only have 38 seasons where a pitcher threw their curveball between 30% and 35% of the time. That number drops to 15 for 35%-40%, and to five for greater than 40%. The samples are even smaller for split-finger fastballs. However, if we look at pitchers who threw each type of pitch greater than 30% of the time, the results aren’t all that different:
|Total N||Attrition Rate||Total N||Attrition Rate||Odds Ratio|
|SL > 30%||389||44.2%||3268||35.3%||1.25|
|CH > 30%||87||28.7%||3570||36.5%||0.79|
|CB > 30%||64||46.9%||3593||36.1%||1.30|
|SF > 30%||11||45.5%||3646||36.3%||1.25|
Once again, we see that sliders and curveballs (and splitters, although we are still only looking at an N-size of 11) increase attrition odds while change-ups decrease them.
So how can we account for this pattern?
There is still a lot of debate about the impact of pitch type on arm health. The data seem to point to a negative impact for sliders and curves, but a 2006 study by the American Sports Medicine Institute showed that there was little kinetic difference between throwing a fastball and a curveball (results for sliders were inconclusive). And while the change-up did seem to put less stress on an arm, the controlled tests suggested that other secondary pitches were not that dangerous. Will Carroll, however, pointed out that that assumes the pitches are being thrown the correct way. As Carroll said, “If you throw a bad anything, you’re going to get hurt.” While I don’t have any data, my intuition is that as pitchers fatigue, it is easier to throw an incorrect slider or curveball than it is to throw an incorrect change-up.
If this is correct, it still leaves open the question of how to explain pitchers that don’t attrit due to injury. Why is it that being a two-pitch change-up pitcher is any better than being a slider or curveball pitcher? I don’t have an answer for this yet; maybe it has something to do with being able to better control a change-up’s location? Maybe pitchers can more easily throw multiple change-ups (i.e. speed, location, movement), creating the same effect as throwing multiple secondary pitches?
Either way, the data does provide us with a signal that we can use to identify pitchers who are pitching themselves into an increased likelihood of trouble the following year.
UPDATE: Originally, I did not split out starters from relievers as I was worried about sample size issues. However, Andrew’s comment below prompted me to run the numbers and here’s the results:
For starters, we see the trend continue for curveballs, but reverse for sliders and become more erratic for change-ups based on usage. Relievers seem to fit the overall pattern better (e.g. higher slider and curveball usage brings higher odds of attrition).
I don’t have much more to add at this point, but will try to dig into it more in a future post.
Here is a link to all the results in table form, complete with N-sizes–all pitches, starters only, and relievers only.
*Secondary pitch was determined by simply looking at which pitch a pitcher threw the most often after their fastball. The gap between usage of the secondary and tertiary pitch was not taken into consideration.
Bill leads Predictive Modeling and Data Science consulting at Gallup. In his free time, he writes for The Hardball Times, speaks about baseball research and analytics, has consulted for a Major League Baseball team, and has appeared on MLB Network's Clubhouse Confidential as well as several MLB-produced documentaries. He is also the creator of the baseballr package for the R programming language. Along with Jeff Zimmerman, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Twitter @BillPetti.