Did you remember to factor in pickoffs at first or second by the catcher? Yadier does that a ton, I’m also sure that just based on watching Yadier play he is undoubtedly the best defensive catcher in the league. Plus, did you see that throw he made against the Mets from his knees and it was still right where it needed to be? He gunned down Cora (I think), plus, the pitch was a breaking ball and Cora got a good jump.

It’s good to pro-rate “expected” SB attempts, but you might want to index it to number of runners in “steal” situations faced (i.e., runner on first with second open) rather than innings. Factor in how often a runner goes with nobody out, one out, two out. Some catchers have pitchers who are good at keeping runners off base to begin with, so they never get “attempted” on.

On WP and PB, consider that some pitchers are harder to catch (cough*WAKEFIELD*cough) and some throw more sinking stuff that has a greater chance of biting the dust. There was some pitch f/x stuff done on the subject some time ago. In the absence of that, you might try a WOWY approach to how often one might expect a PB/WP given this pitcher.

Finally, I’ve done some research that says that a pitcher who throws to first to “check on” the runner doesn’t make the runner more or less likely to attempt a steal, but it does have an effect on whether or not he makes it. (Faster runners are also more likely to draw throws.) You may want to correct for that. Maybe start by having different expected “kill” rates for a runner who’s been checked on vs. one who has not.

And if you want to get really crazy, you can consider the speed score of the runner.

This is good. Keep going with it.

at catcher. Seeing him up there validates most of this list, imo.

This is really interesting, and for the most part seems pretty accurate.

I’d love to know what Brian Schneider’s defense looked like with this analysis, because he’s always been regarded as an excellent defender and I personally just don’t see it.

I agree with the comment above on pick-offs. Also, it might be more accurate to use blocked ball % instead of # of wild pitches and passed balls.

There is no reason to use an “adjustment” for the SB and CS totals. A catcher who has a great reputation (whether it is deserved or not) against whom no one attempts a steal, is not being penalized. He gets zero runs either way just as he is supposed to get. If a catcher has a great arm and no one attempts a steal, then his arm is wasted and so be it. That is the nature of catcher arm defense – you run against the bad catchers and pitchers and you don’t against the good ones. A catcher’s arm is like a tree falling in the forest. If there is no one to run against it, it has no value. The run value of a SB and CS is what it is. That is what you should use for catchers. If a catcher has lots of steal attempts against him with a high CS percentage, either he got lucky or teams are not aware of his talent and are running too much against him. Either way, his runs are his runs (not that I like it).

That being said, there is simply too much noise in ALL of these stats for it to have much meaning as far as catcher talent is concerned. I really hate catcher defensive stats. The only way they are going to have much meaning is in the long run and if you adjust for the pitchers, for example by using a WOWY method. About the only things that a catcher has most of the control over is PB and errors and even with the PB, the pitcher obviously bears some responsibility.

I wouldn’t even include WP in your data. That is mostly about the pitcher. Sure, if we knew how many potential WP (balls in the dirt) were blocked, that would be valuable information about a catcher’s defensive talent, but if we just know how many WP a catcher “allows” that does not tell us much. Maybe 25% at most of a WP is a catcher’s responsibility. Even for SB/CS totals, maybe 40% is in the catcher’s control.

So, bottom line, IMO, you will get these really nice looking numbers by compiling runs saved and allowed based on SB/CS, WP, PB, and the like, but I am afraid it is not really telling you much about catcher defense. Don’t like it. Don’t like it at all. Thing about UZR and other defensive metrics for non-catchers, as you increase your sample size, your numbers converge on a player’s true talent – the number of runs per whatever that you can expect him to save or cost his team given a typical distribution of batted balls. Not so for catchers using this kind of data. Not only do you have the usual sample size issues, but even in the long run, these numbers are not going to converge on a catcher’s true defensive talent. The will converge on maybe 30% of a catcher’s defensive talent and the other 70% is a reflection of all of their pitchers’ talent at holding runners and throwing or not throwing pitches that are hard to catch. Once you have sample size issues (which you always have) AND you are not really capturing that much talent in the first place, you are in trouble to start with…

MGL

I was trying a similar thing recently and found it extremely difficult (my method and results were heading in a similar direction when I stopped). I strongly disagree with the comment that questioned your inclusion of baseunner holds. I thinnk that is essentail to understanding the catcher’s impact. I used the linear weight difference btw first and second times SB% and the results were similar, but I thnk your method is better.

The biggest reason for my failure to finish the study was an uncertainy in the pitcher’s impact on both CS and WP/PB as often expressed here in the comments.

I think of it as the Doug Mirabelli syndrome after Tim Wakefield’s personal catcher for many years. Wake does not hold runners very effectively, his pitches are very difficult to catch and he throws slow. By any statistical measure, you would conclude that Mirabelli is a lousy catcher. However, if he was, he wouldn’t be given the task of catching baseball’s most uncatchable pitcher.

I consider a few ideas that might be relevant:

1: Using the League’s PB rate to prorate the staff of a team for PB, since the scorer is supposed to determine that a PB is the pitcher’s fault this might help indicate wildness in a staff and allow for some correction in that respect. It could be done for each pitcher, but that gets into a small sample size issue and is very time consuming.
2. Looking at pitcher/Catcher data for outlier SB attempt numbers. For example if runners attempt 4X times as many stolen bases against Omir Santos when Mike Pelfry is pitching prehaps Pelfrey is the one they are stealing against and not Santos. This has both the small sample size problem and the labor intensity problem, but it might be something that can be done through graphing once the data is collected. If I knew how to do it I would have tried it.

Even with such problems I doubt that we will see many differences in rankings. I tend to think Varitek and Posada are particularly victimized by their pitching staffs, but the top guys like Yadier and Mauer are the best inspite of the difficulties in such a study.

Nice work none the less.