There's something incredibly compelling about the idea of being able to predict pitcher injuries based on their motions. I guess it's because it makes so much sense. Things look like they hurt, so people empathise with the image of a pitcher moving like a spastic scarecrow and assume that it can only lead to injury. And then they dress it up with engineering terms like stresses and force and loading, put a pretty little bow on top, and voila! Instant biomechanics expert, no education required. And people lap it up, because it's a really cool thing to think about.
I guess the problem is that these people aren't really biomechanics experts, and even most biomechanicists don't really consider failure in the materials that they're working with. There are not many people who've researched material failure (injury, in other words) in conjunction with the pure mechanics of the body, all from an engineering perspective. And if you're one of the few that have, you'll agree with me that it's actually extremely difficult.
What do you need to know to accurately model injuries?
1) Material properties.
This may be the most complicated part. Full engineering tests are yet to be conducted on most of the random connective tissue in the human body (joints are messy and generally pretty intricate), and even if we did have data for the average human male athlete or whatever there is so much random variation due to genetics, etc, that we have little to no clue how a joint as a whole will respond to loading, or the stress pathing, or any of that.
In addition to the basic problem of random deviation across population, biological materials are tricksy little bastards. To demonstrate how complicated the material models are, I'll show you one of the more recent models used to describe the behaviour of cartilage. Without the hard maths, of course.
So obviously you need a finite element solver to work that out, and most of them crash and burn given such messy material properties (have I mentioned that there's also a poroelastic component which at least doubles computational time?). The physical stuff behind all the mathematics is pretty complicated too.
2) Joint anatomy.
R.A. Dickey doesn't even have a UCL! Many people are wired up in odd ways, and that will have a big effect on all the loading/stresses on joints too.
3) The mechanics themselves.
This may be the easiest part, but it still requires a feel for mechanical engineering and a finite element solver. On the simplest level, all this is is measuring the upper body, arm, forearm, hands, etc as rigid bodies, and looking at the axial force and moments/torque imparted in each one as the result of the motion. Reality, however, does not operate on the simplest level, and when you start having to look at the big picture... well, an accurate model of the shoulder moving around in circles might be more than one Ph.D's worth of work. So that sucks too.
This isn't one of those things that I harp on to show how clever I am - I'm just trying to give an idea of how tricky the subject is to handle, and why it's so frustrating to see it all dealt with so cavalierly online. If you don't know the theory behind something, anecdotal evidence runs into the problem of small sample size, and there's nothing you can do about it. Mark Prior had perfect mechanics, remember?
The point is really that I'm not convinced we can just take all the injury predictions that analysts are throwing up online at face value. I don't think I could come anywhere near injury predictions within the next few years, and I'm supposed to be an expert on it. Maybe I'm just stupid, but I'd say it's more likely that we should all be taking commentary on pitching mechanics and their relationship to injury with a boulder of salt.