Two Quick UZR Points

by Jeff Sullivan on Oct 5, 2009 10:54 PM PDT in Miscellaneous

By the way, there's a scheduled network outage around 11pm tonight. Should last ~1hr or so.

I can't tell you how happy I am that this is a community that both uses UZR like it's nothing and understands the significance of playing defense. The conversation around here is miles ahead of what you'll find most anywhere else, and that fills me with no small amount of pride. It's with that in mind that I'd like to issue the following two brief reminders:

(1) Everybody talks about how UZR is unstable and unreliable, but you might want to check out this post over at Fangraphs. Read the comments, too, keeping an eye out for the tangotiger cameo. UZR isn't the most stable stat at our disposal, but in that regard it's not really that much worse than wOBA - a number nobody doubts - and I think that should put many fears to ease. So Raul Ibanez's UZR jumped from -10.4 to +7.3. Adam Lind's wOBA jumped from .325 to .394. Misinterpretation isn't the fault of the statistic.

(2) Never ever ever use one season (or less) of UZR to back up an argument if you can help it. MGL has made a ton of comments about this over at The Book Blog and I'd link to them if I could find them, but right now I'm scrambling to get this out before the scheduled outage for some reason. One season or half a season of UZR can tell you something, but just as is the case with pretty much any other statistic, sample size is king, and you have to look at a broader window of information. You wouldn't just use Adrian Beltre's performance at the plate in 2009 to project him going forward, right? No, you look at his track record and put everything together. It's the same with defense. Raul Ibanez isn't a good defender now. We just have more information that allows us to call him less bad.

If you're dealing with a rookie and you don't have much of a sample size, it's probably best to just ignore UZR altogether, or at the very least to try and combine it with scouting evaluations. If Carlos Triunfel comes up and UZR's -10 at some position through his first three months, he might be bad, but claiming he's a bad defender because he posted a bad figure through his first three months would be like saying Jack Zduriencik's an alcoholic because you saw him in a strip mall with a liquor store. Don't rush to judgment. The payoff of maybe being on to something isn't worth the risk of getting burned.

Tweet Comment 13 comments | 0 recs |

Do you like this story?

Comments

Whenever I hear the words "strip mall" I always imagine a mall with strippers

Thanks for the heads up about UZR.

by Sam Regens on Oct 6, 2009 3:51 AM PDT reply actions

Informative and a welcome addition to the community!

Thanks!

by Aaron Campeau on Oct 6, 2009 4:06 AM PDT up reply actions 3 recs

Something I've always wondered:

I was looking at the FanGraphs leaderboards, and I noticed this:

In 1353.33 innings, Franklin Gutierrez has a UZR of 27.1 and a UZR/150 of 19.2
In 1337.66 innings, Ryan Zimmerman has a UZR of 18.9 and a UZR/150 of 16.0

Even though Gutierrez has only about 1% more innings than Zimmerman, he gets alot more UZR for each UZR 150:
Gutz: 27.1/19.2 = 1.411
Zimmerman: 18.9/16.0 = 1.181

What’s the reason for discrepencies like this? Is it positional? Or is it something else?

by Decatur on Oct 6, 2009 4:07 AM PDT reply actions

UZR/150 is based on a different stat known as DG (defensive games) which is not available on fangraphs.

...and now I'm here

by CapSea on Oct 6, 2009 5:26 AM PDT up reply actions

Yes it is

It’s on the left side of the UZR box.

Basically, Gutierrez had a bunch of balls hit to him this year, so he played the equivalent of 1.5 seasons of center field this year.

by davidcameron on Oct 6, 2009 6:26 AM PDT up reply actions

So

In addition to our OF defense making our pitching numbers appear much prettier, the nature of a flyballing staff can inflate the UZR numbers of our OF.

by CKel on Oct 6, 2009 7:29 AM PDT up reply actions

That sounds like common sense after typing it out.

Just found it interesting.

by CKel on Oct 6, 2009 7:30 AM PDT up reply actions

I liked that post a lot

and its great that UZR is now at a point where it is as predictive as hitting stats. I think people still treat it as soft data because while it may be equally predictive as wOBA it is not based, like wOBA is, on easy to understand, indisputable counting stats. The concept of ‘average fielder’ is both subjective and mutable (eg, if there were a dozen Gutis in the league his UZR would suffer for it), and there is the nagging question of observer bias which Tango was hinting at in a post about Ichiro that was linked to in the recent past.

Of course, all these questions are addressed by increasing sample size, which is precisely what everyone using these stats repeats like a Hare Krishna. I think the implementation of Hit ƒx will go a long way toward calming the detractors.

One thing I wonder about is how much of a factor the home park is in a player’s UZR performance. Ibanez cited Safeco’s LF as one of the reasons he scored so poorly, and Mets fans often claim that switching stadiums hurt Beltran’s numbers. I guess once we have sufficient data to have a good sample of players switching teams we’ll know the answer to that.

De Gutibus non disputandum est

by Bearskin Rugburn on Oct 6, 2009 6:50 AM PDT reply actions

Screw you

Nothing brings me more pleasure then rushing to judgment. I will not let you try and ruin my happiness Jeff

Don't believe the lies Bill!!!! look at the sparkly ERA!!! Sparkly, Sparkly!!! - McCovey Chronicles

by Trenchtown on Oct 6, 2009 9:42 AM PDT reply actions

Thanks for this.

It’s always bugged me that people always attribute variation in offensive performance to injury or a different batting stance or luck, but variation in UZR means that there’s a problem with the stat.

by Manzanillos Cup on Oct 6, 2009 11:45 AM PDT reply actions