July 1, 2014
Do Some Pitches Do More Damage Than Others?
In their recent “position paper” on preventing elbow injuries in Major League (and Minor League and College and High School and Little League) Baseball, Drs. James Andrews and Glen Fleisig had an interesting recommendation for young pitchers: Don’t throw with 100 percent effort on every pitch. The arm, particularly the elbow, isn’t made to take that much stress all the time.
In a recent article at the site RotoScouting, Ben Flajole picked up the idea from a more pragmatic point of view. He looked at the case of Jose Fernandez of the Marlins, who was recently shelved by Tommy John surgery. Last year, Fernandez played on a Marlins team that, outside of Giancarlo Stanton, was, shall we say, offensively challenged. Despite his heroics, he was pitching either behind in the score or with only a small lead more often than the average bear. On another team, he might have had the luxury of throwing more innings in which he was ahead 7-2, and where he wouldn’t have had to worry as much. (Even if he gives up a leadoff home run in the inning, it’s only 7-3.) Flajole suggests that because Fernandez was often pitching with the game “on the line” he might have over-extended himself a bit and pitched at max effort more of the time than most, and that that could have contributed to his eventual demise.
I know that there are already some people snickering at the idea of pitchers “pitching to the score,” but the idea at least passes the silly test. Pitchers might view some hitters or situations as more important and might alter their arm action accordingly. Whether or not that’s a logically sound strategy to follow is irrelevant. Humans aren’t logical creatures and pitchers are human. Whadayasay we take a look?
Warning! Gory Mathematical Details Ahead!
I counted the number of pitches that pitchers threw in these types of situations over the course of a year. Then, as I did in a few other articles on pitcher injuries that I’ve previously written, I looked at starting pitcher injuries from 2002-2012. In the past, I’ve found that two very powerful (and very non-surprising) predictors of a pitcher’s injury risk for the upcoming year are his prior injury history, particularly an injury history to a specific body part (i.e., elbow injuries predict elbow injuries, but not necessarily shoulder injuries) and his overall pitch count from the year before.
To keep the model nice and neat, I ran a logistic regression predicting whether or not a starter would end up with a shoulder injury based on four factors: a shoulder injury last year, a shoulder injury two years ago, his total pitch count from last year, and the number of pitches thrown in these “solo homerun situations.” I did the same for elbow injuries.
The results: As expected, previous injury history entered the regression first (for the initiated, I ran it stepwise) indicating that it was the most powerful predictor. However, the pitch count from situations where a homerun could tie the game or put a team into the lead was next, before the overall pitch count entered, at least for shoulder injuries. The number of high-stress pitches last year is a better indicator of shoulder injury than the total number of pitches. It was not a predictor (nor was overall pitch count) for elbow injuries. When I defined high-stress pitch counts as those coming during times when the game was within a run or tied, that didn’t predict anything.
Reaching Back for an Injury?
This could be a case where the aggregate results don’t do the individual effects justice. For example, in the article on Jose Fernandez, Mr. Flajole suggests that when Jose Fernandez reached a two-strike count, he increased his use of breaking pitches more than a pitcher usually does with two strikes. Maybe some pitchers, when faced with a lot of close situations, default to riskier pitch selections or throw at 100 percent effort more than they should. The fact that Fernandez pitched in a lot of close situations combined with this own habits may really have been what hurt his elbow. Mathematically, we would say that the two factors interact with one another to create a moderator effect. Once again, the simple model probably isn’t sensitive enough to get at what we want to see.
It’s not that the hypothesis that a player’s circumstances might affect his injury chances is necessarily false. It’s that the variables that drive those injuries are related to each other in ways that are complex, both conceptually and mathematically. The tough part is that the combinations of factors that we’re looking at might not occur in big enough numbers to provide the sample size to power the sorts of research designs that would be helpful in figuring this all out. But I’d argue that Mr. Flajole’s argument points to a reasonable question. If Jose Fernandez had been pitching for a team where he didn’t always feel that he was “pitching with the game on the line,” would he still be pitching today? Right now, we don’t know the answer to that question, but it’s not a silly one to ask.