UZR ratings are out

[blue911]

I guess the point here is that +/-5 runs is mostly noise in UZR. I've read that, too, over there, so I'm not thinking Appelman is making something up just for this case. I think what this shows is that while maybe it is one of the best systems out there in the public domain, and an upgrade over officials scorers (e.g., "if he did not touch it, he can't get an error," "no one else would have reached it, but because he reached it and made a bad throw, he gets the error"), it should be treated as being bit fuzzier than how it is generally reported, especially over a week, month, or season.

They should not only fire any scorer that uses this line of reasoning but ban them from every sporting event in the world. Worst calls ever.

[JCA-CrystalCity]

They should not only fire any scorer that uses this line of reasoning but ban them from every sporting event in the world. Worst calls ever.

Soylent Green. 

[blue911]

Soylent Green. 

It's like they don't understand why they assess errors in baseball.

[Nick the Pig]

I guess the point here is that +/-5 runs is mostly noise in UZR. I've read that, too, over there, so I'm not thinking Appelman is making something up just for this case. I think what this shows is that while maybe it is one of the best systems out there in the public domain, and an upgrade over officials scorers (e.g., "if he did not touch it, he can't get an error," "no one else would have reached it, but because he reached it and made a bad throw, he gets the error"), it should be treated as being bit fuzzier than how it is generally reported, especially over a week, month, or season.

I wish I could have the same comfort level as you do.  But I don't.  Here's what I'm getting:

Me: What's up with this New Coke ?

MBC: Why, doesn't it taste good ?

Me: Yeah, it's pretty good.  Tastes a lot like Pepsi, actually.  And I like Pepsi.  But that's not the point.  Why change the original formula ?

MBC: Because it's better.

Me: No it's not.

MBC: Says you !

Me:  Well, yeah.  I drink a lot of cola, and each has their own charm.  Even Jolt and RC.  But I can't figure out why New Coke is supposed to be better than Original Formula.

MBC: Because we say it is.

Me: Ummm...

MBC: Is there really that much difference ?  Let me ask you this - Old Coke was good, and New Coke is good, right ?

Me: Well, yeah...

MBC: There you have it !

Me: Have what ?

MBC: There's no difference !

Me: Yes, there is.

MBC: Only in your opinion.

Me: No, it's measurable by your standards.

MBC:  So what ?  It still tastes like cola.

[PANatsFan]

I guess the point here is that +/-5 runs is mostly noise in UZR.

Well, how do they define error? I guess they have their zones, so there's some intrinsic error based on the size of the zones, but how careful is the error analysis?

[tomterp]

The idea of statistical analysis is to provide the reader with as good a picture as possible.

Understanding the limitations is all a part of understanding when a stat is useful, vs. not.  Any stat has limitations, but that doesn't by any means mean it is worthless.  Just limited in helping understand performance.  A part of a larger picture, not intended to be a whole picture.

[blue911]

Understanding the limitations is all a part of understanding when a stat is useful, vs. not.  Any stat has limitations, but that doesn't by any means mean it is worthless.  Just limited in helping understand performance.  A part of a larger picture, not intended to be a whole picture.

If a stat is misused it's meaningless. Not by the fault of the stat. But if I were to tell you that a relief pitcher was good because their ERA was 2.00, you'd say "That stat is worthless".

[JCA-CrystalCity]

If a stat is misused it's meaningless. Not by the fault of the stat. But if I were to tell you that a relief pitcher was good because their ERA was 2.00, you'd say "That stat is worthless".

I agree with this.  And PA has a point too. I think i understand how they define range, but I'm not sure I understand how they define error and arm.

[Lintyfresh85]

Couple weeks in... some updates on UZR.

Desmond is rated as the third best SS in the game

Zimm is the 4th best 3B.

Nyjer is regressing at a scary rate. He's rated the second worst CF'er in the game just ahead of Matt Kemp. Hopefully things balance out before the year is over... though, I kind of agree he's been less than great in CF this year.

Dunn is at -1.3... and his UZR/150 is still -6.4. That is a massive improvement over his -35.1 of last year!!!

Currently, the Nationlas team UZR of 3.1 is 6th overall in the NL! The Nats were 15th last year.

So, as you can see... The Nats are very much improved on defense this year!

[Evolution33]

i wonder if Desmond being so much better than Guzman hurts Zim in some of the metrics. I have noticed him multiple times pull up on a ball he can get to just to let Desmond have it and make the easier throw. I wonder if this will show up as him having decressed range in the stats even though it is the smart thing to do. 

[Lintyfresh85]

No. The range rating only accounts for plays in your zone that you attempt to make.

Thus, if Desmond makes a play in Zimm's zone, he gets a OOZ (out of zone) play added to his stats and Zimm is not even counted on the play.

[NatsDad14]

What's interesting is that after last weeks update, Willingham was like 20 UZR/150, now he is negative. Morgan was at -20 so that means he had another below average week.

[blue911]

No. The range rating only accounts for plays in your zone that you attempt to make.

Thus, if Desmond makes a play in Zimm's zone, he gets a OOZ (out of zone) play added to his stats and Zimm is not even counted on the play.

Zimmerman is no longer getting the out of zone plays to his left. It has to hurt his rating.

[Lintyfresh85]

Zimmerman is no longer getting the out of zone plays to his left. It has to hurt his rating.

As far as range goes... from Fangraphs:

RngR (range runs): The number of runs above or below average a fielder is, determined by how the fielder is able to get to balls hit in his vicinity.

So, if Desmond is getting to balls in Zimm's zone, it's not hurting Zimm, it's helping Desmond.

I don't even know if OOZ actually factors into UZR at all... kind of like scoops by 1st basemen are counted and listed, but have no stat value when adding up UZR.

[Lintyfresh85]

Yep. As I was saying, OOZ doesn't count.

Revised Zone Rating is the proportion of balls hit into a fielder’s zone that he successfully converted into an out. Zone Rating was invented by John Dewan when he was CEO of Stats Inc. John is now the owner of Baseball Info Solutions, where he has revised the original Zone Rating calculation so that it now lists balls handled out of the zone (OOZ) separately (and doesn’t include them in the ZR calculation) and doesn’t give players extra credit for double plays (Stats had already made that change). We believe both changes improve Zone Ratings substantially. To get a full picture of a player’s range, you should evaluate both his Revised Zone Rating and his plays made out of zone (OOZ). You can read more about the Revised Zone Ratings in this article.

[NatsDad14]

Yep. As I was saying, OOZ doesn't count.

It does count. You are looking at the wrong stat.

[Lintyfresh85]

Well, post the info... cause I'm confused as heck then!

[tomterp]

Zimmerman is no longer getting the out of zone plays to his left. It has to hurt his rating.

He's certainly holding back a bit due to the hamstring issues.  Still, he cuts off quite a few medium hard ground balls to the SS.

[blue911]

He's certainly holding back a bit due to the hamstring issues.  Still, he cuts off quite a few medium hard ground balls to the SS.

Desmond is taking control of the left side of the infield. That's the mark of a good shortstop. The left side is far better than at any time since the Nats moved here and if Zim is still a little sore, so much the worse for the rest of the league.

[JCA-CrystalCity]

Yep. As I was saying, OOZ doesn't count.

I read that as saying OOZ is broken out in RZR but not UZR. It speficially mentions that, for a full pictue, "evaluate both his Revised Zone Rating and his ... OOZ.

The other thing to note in the UZR update is how the RF platoon has fared. Combined, our RFs are 3.3 Runs in UZR, with Willie and J-Max providing all of that in terms of range (Willie T's arm is the other big positive, which offsets the negatives from the other RFs).  The rest of the team is 0.4 - IOW, LF, CF, 2d, and 1b match (negate) RZ and ID.

[Nick the Pig]

On April 21st, Zimm's UZR/150 was -8.4 and he had played 10 games in the field.

Since then, he has played 9 more games in the field.  And his UZR/150 now stands at 25.4

So can we say that he was awful early on, and has been exceptional since ?

Or maybe, just maybe, as the inventors of UZR are the first ones to admit, small sample sizes mean nothing ? 

[PANatsFan]

On April 21st, Zimm's UZR/150 was -8.4 and he had played 10 games in the field.

Since then, he has played 9 more games in the field.  And his UZR/150 now stands at 25.4

So can we say that he was awful early on, and has been exceptional since ?

Or maybe, just maybe, as the inventors of UZR are the first ones to admit, small sample sizes mean nothing ? 

It's both, man. It's both.

I saw Zim unable to get to a routine pop up early in the season, and that was just sad. They rushed him back.

[Nick the Pig]

It's both, man. It's both.

I saw Zim unable to get to a routine pop up early in the season, and that was just sad. They rushed him back.

Perhaps...

I really need to see the whole UZR formula.  I have to think that some of it has to be based on opportunity.

With just 3 weeks (or 6 weeks now) of data, there's still lots of weirdness in the UZR #'s.

[PANatsFan]

Perhaps...

I really need to see the whole UZR formula.  I have to think that some of it has to be based on opportunity.

With just 3 weeks (or 6 weeks now) of data, there's still lots of weirdness in the UZR #'s.

I know a bit about small number statistics  - the thing is that UZR is all relative to the average player. There's 750 players, 30 starters at each position, so multiply 30 x 30ish games x 3 chances per guy, and that's actually a lot of data.