clock menu more-arrow no yes mobile

Filed under:

Why Pitchers Can Be Difficult To Project

God bless the ST articles that still manage to feed us something worthwhile. On Derek Lowe:

"I literally changed everything from my feet to my head," he said. "I looked at video from before, watched video of successful guys and what they do. I changed my arm angle, arm path, stride, knee, shoulder, head, elbow, height of my arm, we can go on all day long. There isn’t one thing I did last year I’m doing right now – which a lot of people will be happy to hear."

Now, there are two options, here. One is that pitching mechanics don't make a difference. That's silly. You're silly.

The other is that pitching mechanics do make a difference, and given that condition, if Lowe's telling the truth and he really did overhaul the way he throws, then we should expect him to be different in 2010 than he was in 2009. Granted, even with a new delivery, Lowe will still throw the same pitches. I imagine he will forever be an extreme groundballer. But when you change the way you throw a baseball, there exists the possibility - or likelihood - that you see a change in the results you generate. And that's the sort of thing that doesn't get picked up by projection systems.

Hitters are easier - not only because they're not as fragile as pitchers, but also because they seldom change the way they hit. You'll hear about the occasional re-tooling, such as we saw with Michael Saunders, but by and large it's just about learning what to swing at and what to avoid. That's a lot more projectable than a pitcher who makes a change to his delivery, and pitchers mess around with their deliveries all the time. Oftentimes they don't make a difference, but sometimes they do, and that can make predicting their performances quite the challenge.

It would be neat to create a database separating pitchers who changed their mechanics from pitchers who didn't. Of course, you'd have to break them down by type of change, and degree of change, and then you're dealing with total sample size issues, but it'd be great to see some work done on the issue, because as of right now this is something we're not really sure how to handle. (Note: I realize this is impossible. I also think it would be neat if my laptop made bagels.)