I have been writing the Team Health Reports for 14 years now. I won't bury the lede here: I'm not writing team-by-team reports this year. Instead, I want to focus on the results of the underlying Risk Ratings. Those will be published on Friday, in plenty of time for most fantasy drafts.
That said, the value of the Risk Ratings is wrapped up in one question: Can injuries be predicted?
There's two very simple answers to this, but it's a complex structure for it.
If you want to believe that a specific injury to a specific player is predictable, I will largely argue that it cannot. If you want to instead believe that risk is quantifiable, and over a large number of players will prove correct and can be used to adjust player value, every piece of work I've done will support this. I'm hardly alone in this, with both academic and anecdotal studies supporting this.
Recently, Eno Sarris wrote an article at Sports on Earth that argued the opposite. His article showed a basic misunderstanding of risk. While his review of the available literature was reasonable, he fails to understand the difference between injuries and risk. That basic issue takes everything else and skews it mightily, leaving the information of the article incorrect.
Sarris uses the example of Yu Darvish as a predictable example, using the research of Russell Carleton as the basis for this. Carleton's research shows that, as well-known, one of the most predictive variables in determining risk is past injury history.
I asked Carleton what his research showed, and he made it simple.
In an email, Carleton said, "The evidence that we have is that if a pitcher has an injury in his elbow the year before, his chances of having another elbow injury go way up. You can substitute any body part you like in there." He was hesistant to apply injury history beyond the one-to-one his research indicates, though he does say there's some likely relationship.
While what he says seems true, Sarris reaches far beyond the branch he stands on and falls. Darvish's minor injury history is hardly enough to base any sort of prediction on. Minor injuries may lead to major injuries, but they do not necessarily do so. Single variable predictions are poor.
A minor elbow injury can only predict another minor elbow injury. It could lead to a major elbow injury, a shoulder injury, an oblique injury or any number of other issues. While Carleton's research is true in a specific sense, it shouldn't be extended beyond that general sense.
When it is, it has to be in the context of something like the kinetic chain, which is a standard part of the literature from Dr. James Andrews and Dr. Glenn Fleisig. (I would link here to an illustrative article or study, but Andrews and Fleisig have done so many—and so many on this topic—that I'll just invite you to put their names in a search engine.)
Part of the issue is that what we know externally is seldom the entire story. In fact, even internally, the athletic trainers inside the game seldom get the entire story from players, who often lack trust or have incentive to hide injuries. These ATs know significantly more than us externally, even with an increase in injury reporting and specificity over the last decade.
Even with internal injury-tracking systems like the ones being used in European Football and Rugby, they are dynamic systems that use proprietary measurements and algorithms, along with very specific parameters that have to be entered. The accuracy there is much higher and can be used proactively.
My system for Risk Ratings is more like the actuarial tables that predict the class risk for any event rather than trying to do the impossible. Adding a tool like PitchFX might seem more authoritative or informative, but it tells us nothing about injury risk.
While PitchFX does tell us the release point and how it can change, which was initially researched by Eric Seidman, it's like trying to determine someone's route by the destination. Sarris may understand the graph, but he seems to misunderstand what creates the graph and a basic understanding of pitching and biomechanics.
PitchFX's release point is an isolated variable and one that is seldom used in biomechanical analysis. To be clear, using PitchFX to determine biomechanics is impossible.
Speaking of biomechanics, Sarris doesn't get back on track by introducing a new variable. Kyle Boddy's biomechanics lab is nice, but it's hardly the kind of lab that teams use. That would be my biggest criticism of using this lab as an example. It's a misdirection or at best, a red herring.
Boddy does not work with many, if any, professional pitchers. (By the way, if you'd like to see this kind of lab, I'll introduce you to the people who baseball turns to for biomechanical analysis later this spring.)
Even among MLB pitchers, very few play for one of the two teams that does extensive biomechanical analysis, and even with the teams that dabble in it, there's still nothing close to a majority use.
I should say here that I admire what Boddy has done, even if I disagree with his methods. I can remember Boddy coming to Indianapolis to work with me when he was learning about biomechanics and teaching pitching. He continued down a path I abandoned because, at the time, I learned that "eyeballing" biomechanics had low value.
But Sarris doesn't get this. He cites linear force, embeds a couple of GIFs and calls it a day. We simply don't know what the forces on Darvish's elbow, shoulder, back or anything else are. The Texas Rangers likely don't either. While the Rangers are a very progressive organization in some ways, they haven't done much with biomechanics.
So can injuries be predicted? Sarris never answers, in large part because he doesn't seem to understand the question he started with.