Catcher fielding heuristics
In the wake of the silliness known as the Gold Glove Awards voting, in which Yadier Molina did not win one as catcher, we again were faced with the inadequacy of fielding metrics, specifically for catchers. In perusing the reactions to the Gold Glove announcement, we’ve read about Caught-Stealing %, stolen bases allowed, passed balls allowed and errors. How very 20th century.
Yet it seems that two areas of catcher skill can more readily be quantified using data that is available, if not yet in easily-manipulated form to the public, certainly to the major baseball web sites. We’re thinking of two metrics that can be developed to address major areas of a catcher’s defense: Blocked-Pitch Percentage and Quality of Runners Faced.
Let’s start with the first: how well a catcher blocks pitches. This is certainly an important defensive skill, since not only does it prevent opposing runners from taking extra bases (and therefore prevents runs), it allows a pitching staff greater leeway in the types and locations of pitches thrown. Commonly, catchers are assessed by how many wild pitches and passed balls they yield. The Hardball Times at least converts these into a rate stat, WP+PB/G (Wild Pitches and Passed Balls per Nine Innings played). But that is nearly as misleading as errors per game, since one is left wondering just how many chances the catcher had in the first place. We propose simply taking the total WP+PB and dividing by the total number of pitches in the dirt, something that MLB GameDay tracks, to give Blocked-Pitch Percentage. It would be hard to find a more accurate metric for a catcher’s plate coverage (and we’d like to hear about it).
The second major area of catcher skill, preventing runners from stealing base, is also quantifiable. Most observers are happy to point out a catcher’s Caught-Stealing Percentage as a way to validate his success against the running game. But that’s inadequate unless all catchers face the same baserunners. For example, catcher A throws out six of 10 (60%) would-be basestealers, and catcher B only throws out two of five (40%). Catcher A is the better receiver, right? Well, if the six runners that catcher A throws out are among the slowest, least adept basestealers in the league, that’s not saying much. And if the two that catcher B has nailed happen to be the two best basestealers in the league, that’s a big deal. Moreover, if both catchers caught the same number of innings, isn’t it important that catcher B had 50% fewer attempts against him?
Enter Quality of Runners Faced (QRF). Take all of the runners who challenged a catcher, subtract their stolen-base attempts and successes vs. that catcher, and you’ve got a QRF factor for that catcher. You could refine it further: Of the runners that a catcher allowed to steal a base, determine their QRF — call it the catcher’s SB QRF. And get the QRF for all the runners that the catcher nabbed, and call it his CS QRF.
With the pitch-by-pitch data now available through the work of MLBAM, these kinds of helpful metrics are possible. Of course, Blocked-Pitch Percentage and Quality of Runners Faced won’t solve the problem of coaching staffs selecting the Gold Glove winners, not yet, anyway. But they can help those observers who are interested in basing their views on something other than the conventional and insufficient stats we have today.
November 9th, 2007 at 8:53 pm
I asked Dan Fox about the application of the pitch f/x data to catchers and he basically said that the way they label the pitches is bogus. He implied that there isn’t consistency about when a ball is actually blocked in the dirt or if it skirts away, etc. I absolutely agree that catcher defense very much needs to be quantified (and hopefully disproved in large part — but that’s just my own slant).
November 10th, 2007 at 11:44 am
I respect Dan’s opinion, so I’m not sure if I’m misunderstanding what he’s saying or simply disagree. Having done about half the games at Busch this past summer for MLBAM, I can tell you first-hand that we tracked blocked pitches very accurately. The f/x display onscreen or the x-and-y coordinates for balls in the dirt may be off, perhaps. But as for whether a catcher actually blocked a pitch in the dirt, I’m confident that that data is highly accurate (in addition to trying to get it right live at the ballpark, we also had support people in New York who review each game to make sure).
November 11th, 2007 at 5:51 am
I’m not following you or Mr. Fox, Pip. In the main entry you said you were just after “chances” and then we could go from there. I agree that this should certainly improve things.
However, in both of the comments, Dan and you seem to imply that you’d need to know which pitches were blocked and which weren’t. If you knew the amount of balls in the dirt, and then the amount of past balls and wild pitches for your metric.
Perhaps I’m misreading something.
As for QRF, I wonder how well a runner’s CS% correlates from yty. It’s always seemed a little fluky to me on a player level, but that’s just a guy reaction. I wonder if you did this for enough runners (let’s say 60 different players try and steal on a catcher in a given year) that you aren’t going to see something pretty similar for most catchers.
On the other hand, it stands to reason that only the best thieves would run on someone like Molina. Maybe you’d find that QRF is inversely proportional (to some degree) with a catcher’s career CS%, and maybe offensive environment as well (or they only run in high leverage situations). Perhaps there’s a constant to be found and Molina’s CS are actually substantially more valuable than Joe Averagecatcher.
Just thinking out loud.
November 11th, 2007 at 5:54 am
Couple of typos:
“guy” should be “gut,” and there’s probably a joke to be made there.
The last sentence of my second paragraph should read “If you knew the amount of balls in the dirt, and then the amount of past balls and wild pitches, that should be sufficient for your metric.”