from Amazon
Monday, April 21, 2008
Now this is good stuff: Fastball speeds
With all due respect to the previous work, by all of the pitch f/x researchers. This is exactly the kind of stuff that needs to be done to understand what the heck goes on from the mound to home plate, and exactly the kind of stuff that teams need to know in order to train and evaluate pitchers. Any team that isn’t reading these articles, or doing their own similar research, is, let’s say gently, not yet on the technological bandwagon, and is in jeopardy of being left in the dust.
One thing you can clearly see from John’s data and analysis is that you don’t need to pitch down in the zone to be effective, at least with the fastball. In fact, the best place to throw the fastball, in general, is UP in the zone. And that goes for soft tossers as well as guys that can bring the heat.
Every pitcher has been taught since Little League, to keep the ball down, no? That may be bad advice, at least as a general rule. Of course, what makes the high fastball seemingly a little better than the low one, may be the fact that it is not thrown high all the time. There is a lot of game theory in pitch location as well as selection per se. It might just be that pitchers (not all, of course) throw the right amount of pitches up and the right amount down. But, no matter what, I think it is pretty clear that it is NOT correct to consistently throw your fastball down in the zone. Now that obviously does not apply to ALL fastballs. Maybe it is better to throw more 2-seamers down and more 4-seamers up and it is probably likely that most sinkers (even the 93-94 mph sinkers that Webb and Wang throw) should be thrown low, even if the batter knows that.
What do you think many pitching coaches would say if you told them that the most effective fastball, in general, is the high one and not the low one, given the percentage of high and low fastballs that are currently thrown (I’m not sure that many of them would understand that last clause)? I think a lot of them would look at you as if you were crazy! Isn’t Mazzone lionized because he proposes throwing everything (not literally) down and away?
Anyway, when I first saw John’s graph of the low and away fastball and the fact that fast and slow fastballs were equally effective, I was a little surprised, but not that surprised. I right away thought that there was a huge selective sampling effect at work when you look at the effectiveness among the various speed fastballs. In order to survive, a slow fastball MUST have more movement, have more deception by the pitcher, and the pitcher probably mixes up his pitches more than a guy that throws 95 mph. Also, even if you are looking at a specific location in the zone, like John was looking at the low and away segment of the zone, if a pitcher is more wild in general (like you expect the 95 mph guy to be), you might expect that the batter is going to be more selective on those pitches as opposed to against a control guy, like you expect the soft-tosser to be, such that, again, the soft tossers fastball is going to be more effective even given its slower speed.
I did not explain all of that very well, but what I am trying to say is that you CANNOT compare the quality of fastballs across different pitchers any more than you can do an aging study by comparing how all 25 year olds do as compared to all 26 year olds, etc. The two problems you have with the pitch data is, one, a huge selective sampling issue, as I said, which is that there is going to be a gigantic difference between an 87 mph fastball thrown by a successful major league pitchers and an 87 mph fastball all of a sudden thrown by CC Sabathia or any other pitcher who normally throws in the 90’s, and two, as John mentions, that the effectiveness of a pitch depends on so many other factors other than speed, yet these other factors are hard to control for while still working with large samples of data. (I also want to congratulate Johns for NOT doing a multiple regression. They are WAY too opaque for anyone who is not a pHD level statistician to understand and really see what is going on. When someone does a multiple regression, you pretty much have to take the results on faith. Not that there is necessarily anything wrong with that.)
Anyway, to be honest, I’ve only read half of the article. When I finish the rest, I’ll come back with more comments. Despite some of the problems, some of which I have articulated, this is a great start, and long awaited.
Anyway…