If you haven't read Russell A. Carleton's article from Monday on the factors that really predict pitcher injuries, go do that now. Then listen to his subsequent tour of the baseball podcast circuit, from Buster Olney's Baseball Tonight to Ian Miller's and Riley Breckenridge's Prodcast. I'll wait.
Finished? Okay, you're all caught up. There was one part of Russell's piece that mystified me: his finding that home run rate was the best predictor of elbow injuries, and that pitchers with lower home run rates have elbow injuries more often. The rest of the links he found were immediately intuitive. Previous elbow injury history? Sure. More batters faced, and thus more pitches thrown? Makes sense. But I wondered why a low home run rate would be a bad omen for elbows, so I decided to look into the relationship a little more. Whatever the reason, it would have important implications: as commenter jfranco77 noted, the low-homer-rate link might mean that Brett Anderson is an even bigger injury risk than his already frighteningly long list of DL stints would suggest.
First, I asked Russell a couple questions. Was it possible that the home run rate relationship was statistically significant only because something is bound to be significant if you have enough variables? Russell's response: It could be a freak case of statistical significance, but using his stepwise method makes random, throwaway "significance" less likely—home run rate beat out the other variables as the most significant. Okay, well, was he looking at only one year? If so, maybe a bunch of low-homer guys happened to get hurt at once. The answer: No, this was from 2006-2012.
We reasoned that low-homer guys are generally high-grounder guys, and that high-grounder guys throw sinkers. So is it something about the sinker? Neither of us knew, so I emailed BP injury authority Corey Dawkins and mechanics expert Doug Thorburn. As it turns out, the relationship probably can't be blamed on one problem pitch. I've reprinted their email responses below. You can reach your own conclusions in the comments. —Ben Lindbergh
Corey Dawkins: I don't know of any research that shows that sinkers are more dangerous than any other type of pitch, but it could be velocity that is the main thing here. The higher the pitch velocity, the greater the risk for Tommy John Surgery—that’s pretty well accepted. Of course, high velocity is a big reason why many pitchers are effective, so it's a double-edged sword.
I think the groundball/sinker link most probably comes down to this. Groundball pitchers (especially the extreme ones) don't use the changeup or curveball as much as other pitchers, so they’ll end up with a higher percentage of high-velocity pitches. If it's a difference of 15 pitches per start, that's easily over 450 extra high-velocity pitches per season, not counting warm-up pitches, bullpen sessions, etc. Surely that plays a role.
Doug Thorburn: I have to say that I am surprised by the finding. However, the multitude of variables that contribute to HR rates seriously clouds the issue. Even assuming the GB relationship, and then further assuming the sinker connection, we are: A) creating a pattern of reducing the reliability by grouping multiple tendencies that are not indicative of the whole class, and are also not 100 percent connected within said class of pitchers, thus diluting the sample to the point that it would be impossible to conclude that sinkers are more predictive than having a previous injury; and B) we are ignoring the critical element of pitch command, which is not only impossible to measure directly (walk rate does not suffice), but is generally accepted as having a relationship with elbow injuries (ie command is the last thing to come back for TJS pitchers). Walk rate does not work for two reasons: 1) based on anecdotal evidence, most home runs that I witness come as a result of a pitch that misses IN the zone; and 2) walks fail to describe the degree to which the pitcher is missing targets.
Regarding A), it is theoretically possible that the pre-set pronation that is used to throw sinkers and two-seamers is an added risk, though I have my doubts, given that the pre-set pronation actually lessens the kinetic toll during follow-through/deceleration phase because the arm goes through less pronation after release. This is in contrast to curves and sliders, whose pre-set supination results in greater degrees of pronation after pitch release (pronation is an inevitable aspect of every pitch for every player after release point). This is critical, as the arm decelerates in roughly half the time that it takes to accelerate, though that kinetic toll is lessened a bit due to the subtraction of the 5-oz baseball after release.
Regarding B), we will be hosed by confounding statistical variables until we have the ability to accurately measure pitch command. A tracking system that measures the position of the catcher's glove at pitch release AND his glove at the point that the ball crosses the plate would be the best approximation of pitch command, but even in that case, we would have variation due to the discrepancies between the Jose Molinas and the Jesus Monteros of the world. There is also the issue of pitchers with poor "control" getting into more hitter's counts, resulting in more fastballs aimed closer to the middle of the zone. Basically, you have both the RIcky Nolasco model (throws strikes but misses within the zone) and the Ubaldo Jimenez version (can't hit targets at all and is forced to throw meatballs).
In sum, I do believe that it is a bit of a statistical blip, not because it is untrue or artificial, but rather because it may not lead to any specific connection that would allow coaches and teams to change the protocol in order to prevent injuries. If anything, it could be that the results are slanted by pitchers who tend to miss targets so badly that their wayward pitches are hard to hit out of the yard (it's a whole lot easier to mash a strike), and such pitchers tend to have particularly poor mechanics. To be a little more specific, such pitchers who miss targets by feet rather than inches tend to have serious issues with timing, and the greatest precursor to elbow injuries that we have found is strongly related to timing. Namely, I'm talking about elbow drag, where the elbow trails behind the shoulder axis as the arm triggers into internal rotation, and which is often caused by a heavy scapular load paired with late initiation of trunk rotation.
*Update* Harry Pavlidis tested Corey's hypothesis that groundball pitchers throw a higher percentage of "hard stuff" than other pitchers, which would mean more high-velocity offerings and potentially more stress on their arms. He found the following "small but probably meaningless" relationship between groundball rate and "hard stuff" (four-seamer, two-seamer/sinker, and cutter) rate from 2010-12 (click to expand).
Special thanks to Doug's wife, who allowed Doug to answer emails about mechanics on his honeymoon. Or maybe she didn't know he was doing that, in which case: sorry for blowing your cover, Doug.