This site is getting better faster than any other website I’ve ever followed. One year ago, I only came to fan graphs to grab ZiPS projections. Now, I read it almost every day, and I catch up when I miss a day.
For the sake of argument, I’m not entirely convinced that an outfield are makes only a minor difference to an outfielder’s value. Assuming that an outfield assist is worth the same amount as a put out and assuming that outfield assists are a repeatable skill, then the best arms appear to create about 5-10 outs more than an average outfield arm and the worst about 5-10 less than average. (I’m eyeballing all of this from baseball reference.) If that’s right, then a good outfield arm is worth about .75 wins above average and a bad one is worth the opposite of that.
Comment by philosofool — December 29, 2008 @ 9:08 pm
They do make a difference, as you said. I think some people believe the difference is more than that, however.
I have more of a general question about defense. I have been following the defensive value that FanGraphs has been using and have noticed that it seems to be rather random for a player from year to year. Maybe someone can help me answer this: Do you think this variance has more to do with the imprecision of the stat or with the nature of defense?
Of course, this stat is demonstrating how well a player defends in a certain season, not the players defensive abilities. I am just trying to make sense of projecting these new dollar values over 2009 contracts, and it is difficult when it seems defense is a stat with a very large confidence interval.
You know what would be really really cool? If there could be a plot consisting of each putout made by the player on a mock baseball diamond. I remember seeing a graphic like this for Derek Jeter and Troy Tulowitzki, and it was clear that Tulowitzki was a better defender than Jeter.
philosofool: THT has a feature up by John Walsh which quantifies OF arms (not yet updated for 2008, but you can check previous years). I’m not sure how MGL is going to deal with arms, but from Walsh’s rankings you can see that arms, however important, do not transform a poor fielder into a good one, or viceversa.
qqqq: I think you’re referring to David Pinto’s PMR graphs. They’re certainly nice in showing areas of strength and weakness.
Dave: as far a catcher defense, I believe there are several ways to work around this, also using linear weights. I’ve made my own rankings using the linear weights for WP, PB, CS, SB and also including a quantified “intimidation bonus” for C’s that have less SB attempted against with respect to the average catcher (runners will go less against Yadier Molina for example, and there’s value there), and the results are fairly similar to what Chone got with his own studies. Moreover the rankings could be improved with a hand from Pitch F/X (Dan Turkenkopf checked C block % on balls in the dirt, which could skew the evaluation of certain C’s). Really, the only troublesome evaluation in my opinion would be gamecalling, because we’re never quite sure what the pitcher shakes off and “calls for himself”, but aside from that, quantifying catcher defense doesn’t seem to be as hard as a few years ago with the data we have right now.
Having said this, I had everybody in the +15 to -15 range, which is a 3 win spread. In fact, aside from Ramon Hernandez, nobody was lower than -9, and nobody aside from Kurt Suzuki and Jason Kendall was higher than +11. Saltalamacchia, Barrett and Max Ramirez would have been a lot worse than Hernandez with more playing time, while Quintero, Ausmus and Towles would have been better than Suzuki.
But that’s what I got… honestly I think that by evaluating all events versus league average and using the run value of each of them we could do a pretty good job of evaluating the defense of a catcher. To be honest, I’m already quite satisfied with my own results, even though I know that they could be improved.
Very interesting stuff on the catcher defense issue. But how much of those numbers are solely dependent on the catcher? What if a catcher has a handful of pitchers that are bad at holding runners, but is actually a superb thrower. Or if he has a lot of wild pitchers that throw a lot of passed balls and wild pitches? In general it would probably show you who is good and who is bad, but I wouldn’t take accuracy for granted. If you haven’t already, you should read Tango’s With or Without You catchers article in the 2008 annual (last year’s). It’s a similar, but more comprehensive (and I believe, accurate) approach to what you’ve described here.
I would guess that variance in year to year UZR has more to do with the data we have than the nature of defense. I think that as we get better data (specifically stuff like how long a ball was in the air, how hard it was hit, etc…), we’ll get more precise measures of defensive value and see less variation.
I generally just use a +/- five runs when looking at UZR. So, if a player is +10, then I’ll agree with anyone who wants to claim that he was between +5 and +15.
my work isn’t all that accurate in actually quantifying ability, and I really just use it as a proxy to see who’s good and who’s not. Obviously there are points that I don’t address, ranging from gamecalling (which I cannot address) to other stuff, some of which you’ve mentioned, such as pitcher wildness or pitcher control of the running game.
My point is that I was just looking for a proxy to define skill, so I’m not too worried if Mauer’s true ability is better than Suzuki’s (I had Mauer at +11 and Suzuki at +14), but the point is that they’re pretty similar and both good.
Those points I mentioned could be addressed more formally by someone who has more time on his hands. How?
Pitcher wildness could be checked like Dan Turkenkopf did last year: what percentage of balls in the dirt does the catcher block? Obviously we won’t fault catchers who don’t block balls which are 5 feet away from the strike zone, laterally. To identify where pitches ended, we could use Pitch F/X, and we would be set.
For the control of the running game, we could implement a WOWY approach to see which pitchers are the best at controlling the game. I recall Tom Tippett doing something along these lines many years ago.
I’m sure something would change by adding these features, even though not dramatically. I hope someone does decide to do this. It could be time consuming, but certainly worth it.
wouldn’t it be better to use something like a rolling 3 year average for defense? also, isn’t it pretty clear that average defense != replacement level defense at this point? that was a pretty common meme for a while .
could you run this for, say, the past three seasons? seeing how well they correlate from year to year would be a big deal, plus i’d just like to have the data. in general, it looks like this lines up with reputation (i’d been using -10 for AJ Pierzynski, for example…). the raw numbers themselves look pretty worthwhile to me and adjusting them by rough inference shouldn’t be too hard. i’d like to see fangraphs publish this stuff, frankly.
using AJ as an example, Contreras is terrible at holding runners and is prone to WPs thanks to the forkball, but Buehrle probably cancels that out. Danks, Floyd and Vazquez were probably average-ish, so I’d guess that is AJ’s approximate talent level at this point.
Do you have a post and all that I could link to with a thorough methodology and some conclusions and all that? Looking at the data provided in the macro is pretty interesting. For example, I plugged the 2008 season into excel and looked at RAA/120 v. playing time and there’s actually a significant negative correlation between the two, which surprised me. In light of Tango’s 2009 THT annual article, though, I guess that makes sense. The whole idea is that it’s incredibly taxing just to play back there. Breaking the 1000 Inn barrier is a huge accomplishment in itself and only about 15 are able to do that per year.
No, sorry, I haven’t listed the formal methodology and so on, and that’s because as I said it’s kind of a proxy for myself, really.
But I do take into account WP, PB, SB, CS, E and SB attempts against, measuring their linear weights and accounting for playing time. That’s about it.
Publishing this data is great. That said, it looks at defense using only one metric. There is a real possibility that even adding +/-5 defensive runs might not capture what a survey of metrics would suggest. For instance fangraph’s UZR might actaully grade Dunn 10 runs worse than a survey of Dewan’s, CHONES, PMR, and MGL’s UZR would suggest he was in ’08.