keyboard_arrow_uptop

If you haven't heard the news, New York Mets wunderkind Matt Harvey has been diagnosed as having a partially torn ulnar collateral ligament. Harvey has said that he wants to avoid surgery if possible, but this sort of thing usually ends up with a visit to Dr. Tommy John for James Andrews surgery. As per usual, everyone on Twitter remembered their extensive medical training and pitching mechanics expertise and did the most productive thing that they could: lay blame for Harvey's unfortunate circumstances at the feet of his pitching coach, his manager, his conditioning, his conditioner (the hair kind), his genetic makeup and, of course, dumb luck.

Any time a team loses a young pitcher like Harvey for an extended period of time, there's an inevitable period of mourning and second-guessing as to what could have been different. Even for fans of other teams than the Mets, it's sad to see a bright star like Matt Harvey go down. Is there something that could have been done?

Warning! Gory Mathematical Details Ahead!
A few months ago, I did a study on what really predicted whether pitchers would sustain an injury, based on markers from the previous year (or two). Using a database of injuries that we maintain here at BP, I was able to locate all pitchers who had Tommy John surgery from 2002-2012 and check what risk factors would be obvious at the beginning of the year. In the original study, I found that elbow injuries (of any sort) were most closely associated with a previous elbow injury, as well as with a low home run rate and the raw number of pitches thrown the year before.

When I ran a similar regression (for the initiated, a stepwise binary logit) predicting who would end up having Tommy John in the next year, I got something similar. The best predictor was the presence of a previous elbow injury. The second was a high groundball rate in the previous year. The raw number of pitches did not enter in as a predictor. There was also no effect for walk rate, strikeout rate, or any of the other performance-based metrics. Harvey does have an above-average grounder rate, but according to our database, he had no previous significant elbow injury to speak of.

I also looked at whether there were certain "abusive" pitch counts that predicted a greater chance of undergoing Tommy John surgery (using the same Cox Regression methodology as in this article). In general, I found that there weren't really any effects of note until a pitcher crossed the 115-pitch mark in his start. When I looked at elbow injuries more generally, there was a red line at 120 pitches in an outing before effects started to appear. There were also effects for previous elbow injuries and age, with younger pitchers slightly (but significantly) more likely to get hurt. Matt Harvey did have games this year in which he threw 121 pitches (April 29th), 116 pitches (May 22nd), 116 pitches (June 18th), and 121 pitches (July 8th). There's a 114 in there from May 28th as well.

In the original article, I also pointed out that the results from the regressions that I've run are based on pitchers whom managers and pitching coaches believed could handle that level of workload. Again, we assume that managers and pitching coaches are smart enough to introduce bias into the sample (i.e., they have an idea who can and who can't do that). But even knowing that these guys are considered durable enough to handle it, we still see some pretty awful effects for injury once you get up to 115-120 pitches in an outing. The marginal risk added by extending Harvey past 110 pitches in the four highest pitch count games above (that is, the extra risk incurred from going from 110 to 116 pitches) was worth about half of the risk of a previous history of an elbow injury. For those 34 extra pitches, they pushed Matt Harvey's risk halfway to the worst risk factor there is.

These analyses are blind to issues like pitching mechanics, and the injury dataset we use does not include unreported injuries, so there could be biases in the study that we can't fully account for. I am very happy to hear other arguments that those factors may significantly moderate my findings, but as of right now, I think we need to have a small conversation. Those 34 extra pitches that Matt Harvey threw probably bought the Mets two or three extra innings on those days that he pitched, but purchased it with an increased risk of injury to Harvey.

We need to talk about whether the increased risk of losing Harvey for a year (or even for two weeks) was worth it. It's piling on to pin the fact that Harvey will miss a year on the Mets, but it's also not right to excuse them completely. They took a risk and got burned.

You could try to make the case that those were three absolutely vital innings, or that the buzz that surrounded Harvey was worth more to the team than his 2014 season (because the Mets won't win the World Series this year, and it's not likely next year either). You could make the case that even if they hadn't pushed Harvey, there's always the underlying risk that a pitcher will get hurt. You could make the case that the risk of injury and the lost productivity that might follow is actually worth it. What I think isn't fully understood is that while pushing a starter to high pitch counts is not an automatic injury sentence, it is an increased risk. Like teenagers saying "It'll never happen to me!" or investors screaming "But this market is different!", it's easy to pooh-pooh that risk, but it seems that the risk added is generally believed to be much lower than it really is.