Interesting on Eckstein. MGL can correct me if I'm wrong, but I seem to remember him having surprisingly decent UZR numbers when he was in Anaheim.

I'm surprised by Ichiro as well.

How does playing time affect UZR? Is it like a counting stat? In other words, would Figgins' numbers be worse if he hadn't missed a month of the season?

Perhaps Cano could play SS...

If anyone's interested, I have a worksheet for player's defensive runs using ZR (and assists for OFers) as well as their LWTS (including DP rate and ROE). I have pitcher stats too but they're not as in depth.

Anyway, here's the link to the AL spreadsheet:
AL 2007

I need to finish up pitchers in the NL, which will probably be done tomorrow. And I'll be tweaking some things as well, mostly adding park factors to positions like LF in Fenway...

Kent -23
Yep. He's awful.

Jeter the worst defensive shortstop in baseball? I don't buy it. And what's with Ichiro getting the worst AL CF? There's some bias in these numbers...

Jeter the worst defensive shortstop in baseball? I don't buy it.

I've often felt UZR or other defensive systems somewhat overstated how bad he was in the past. But this season, he really didn't seem to get to anything. I buy it.

And what's with Ichiro getting the worst AL CF?

Yes, this is odd. I thought something at midseason (I thought UZR, but I guess not) was showing him playing at a Speaker/Mays/Jones-plus level. Not that I've seen him much, but I really doubt he's that bad.

I think that four Royals made the best list (Pena [maybe], Teahen, DeJesus, and Gordon).

Yay!

How does playing time affect UZR? Is it like a counting stat? In other words, would Figgins' numbers be worse if he hadn't missed a month of the season?

The stats presented at the link are prorated to 150 games, so you're seeing a rate stat there.

How does playing time affect UZR? Is it like a counting stat?

I would think that since the "R" stands for "Rate" that it would behave like a rate stat, but not knowing the details, I'll let those who know say their piece...

Jesus, Ichiro at -14 in center? I never would have guessed. Never.

I sorta of buy the rating for Ichiro. I saw a lot games of him chugging and chugging after balls for what seemed to be relatively interminable times and either catching them by taking a quick last second stab or failing to catch them. It looked like a lot of effort and there was just something off.

There's some bias in these numbers...

Yes, I actually run the numbers and then change them for players I like or don't like. I definitely don't like Jeter. Actually, I don't mind him at all. It is the media's infatuation with him that I hate. So I change his numbers a lot to make him look really bad. I think Dewan does the same thing, but I am not sure.

Scareduck, R is for "rating" which does not necessarily mean a "rate" but in this case it is based on the number of defensive games they play and then everyone is prorated to 150 defensive games (no matter how many games they actually play, even though it will be close to the same).

I may sound like a broken record here, but sample numbers are just that, sample numbers, a sample of a player's defensive performance in this case. That means that for several reasons, some percentage will be A LOT different from a player's true talent performance, or that performance he would have if he played an infinite number of games. Ichiro may be one of those and he may not be. His past UZR numbers in RF are not all that great, so there is a great statistical suggestion that he is not as great at running down and catcing fly balls and line drives as most people and scouts/teams think. But again, that may or may not be true. It is only a suggestion, albeit a strong one, if you include his many years of UZR.

If you pinned me down and put a gun to my face, I would have to say that Ichiro is probably not nearly as bad as his UZR numbers over the years suggest and that it is a statistical aberration (or perhaps some bias or methodological weaknesses with UZR, or some combination), especially if the one holding the gun is a Mariners fan. Seriously though, I am certainly not 100% confident in that notion (that he is much better than his UZR numbers).

And for those of you not familiar with PBP fielding metrics like UZR (there are a number of them out there), they are very good assessments of player's actual fielding talent and value in runs. And like any sample metric (almost all metrics are sample metrics, which simply means derived from a sample of a player's performance, even if that sample is an entire career), they are subject to sample error. That sample error can be mitigated, especially for small samples, if you know other things about that player (observations, scouting, speed, etc.)...

Interesting on Eckstein. MGL can correct me if I'm wrong, but I seem to remember him having surprisingly decent UZR numbers when he was in Anaheim.

Eckstein has been a very good shortstop in the past. This could just be an aberration, but it could also be age, and injuries affecting his quickness. Eckstein obviously isn't making plays with his arm, if he loses a step it could be disastrous for him.

http://spreadsheets.google.com/pub?key=pfk_WuYpfduzefA8-s0_iYw

There is a lot more data in this spreadsheet, calculated by a reds blogger using what reads to me like the dial method. Their is a link to the actual blog in yesterdays dug out for anyone who wants to read the methodology.

Their are some severe disagreements between the two methods, which makes it very tough to distinguish how good many players defensivly may be.

The lack of a strong directional correlation between statistics attempting to measure the same thing, makes it very difficult to put much faith in the accuracy of the numbers.

I kind of like how Rowand + Victorino - Burrell = 0. It takes a lot of good fielding to balance out Burrell.

EDIT: make that + Burrell

A semantic nitpick, but something that makes me scratch my head:

Isn't a Zone Rating a decimal? Isn't this something more like Fielding Runs?

Yeah Manny the worst LF at -33. One thing which nobody has ever come out and explicity explained to me: are uncatchable balls high off the wall counted against him, or not? And if they are counted what is the rationale?

Part of the difference could be that Justin's using THT's revised zone ratings. Remember, Chris developed his method to go with vanilla ZR, so he may have already been factoring in some of the system's weaknesses when he worked out the problems.

MGL,

How did Buck and Swisher fare for Oakland? Swisher's ZR is pretty good in the corners, but awful in a small sample in CF. Do you think either of them could handle CF?

Yeah Manny the worst LF at -33. One thing which nobody has ever come out and explicity explained to me: are uncatchable balls high off the wall counted against him, or not? And if they are counted what is the rationale?

Stats is counting balls off the wall against Manny. Since his RZR rating converted to runs is almost identical, they must be doing that as well. I don't know the rationale, or why they continue to do so when it makes no sense.

My adjusted rating for Manny is -17 (actual playing time, not per 162) which puts him in the same category as a generic bad left fielder like Dunn, Lee, or Willingham.

Where was Polanco in the 2nd base rankings? Does he just not have as much range as the leaders?

I use a park factor for Fenway, which essentially (not explicitly) does NOT count balls hit off the wall against Manny or any other left fielder when they play in Fenway. In any case, over the many years, his road UZR happens to be almost exactly the same as his overall and home UZR (in the high 20's and low 30's the last few years), at least the last time I looked at his home/road UZR splits (last year, I think).

Polanco was +12 in 124 games. He is one of the very best every year. One of the most underrated and under-paid (I think) players in baseball for a while now.

Swisher was +6 runs overall in RF and -3 in CF (these are total runs and not per 150 games), both in limited time of course. I don't know if he can "handle" CF or not, but when moving from a corner to CF, you are going to lose several runs (I forgot how many, off the top of my head), maybe 6-8. Buck was +3 runs overall in RF and LF and a few games in CF. Last year, Swisher was +5 runs overall in around 100 games in RF and LF, and in 05, he was a stellar +12 runs in 90 games in RF. He appears from the numbers to be a well-above average corner OF, which should make him at least an average CF'er, again, at least according to the numbers.

Would intentional deep positioning of the outfielders affect how they look to UZR? What I mean is, would positioning the outfielders a lot deeper than normal make their UZR look bad, because they allow a lot more singles to drop in but are more focused on preventing the extra base hit?

Shoewizard wrote about this in regards to AZ's outfield this year. The Dbacks are intentionally playing all of their outfielders really deep; that's especially the case with Chris Young. As a result, the Dbacks have given up A LOT fewer doubles and triples this year, compared to last year.

I don't really know how UZR corrects for positioning, or if it even cares for it, but I do know that the Dbacks front office doesn't care much for what the defensive metrics out there think about their players; I believe the term that was commonly used is GIGO. I'm certainly curious to know what MGL does to correct the first "G".

Here's a question about UZR I've been meaning to ask for years...

If you develop an statistic that measures overall offensive ability using some kind of +/- system, and you want to measure how accurate/effective it is, just add up the totals for each player on a team to see the number of runs the team is predicted to score, and compare that with the actual number of runs a team scores. The more accurate the statistic, the more closely it should predict runs scored. Of course, people have made many attempts to determine the best offensive statistic using this very method.

Obviously, defensive statistics are more difficult to validate, because of the interaction with pitching. Some team statistics are better than others (i.e. DER, or BABIP), but none are as perfect as runs are for offense. If anyone has any ideas for better team defensive stats, I'd be interested.

Still, a team that is full of players that rate very well using UZR or another system should also tend to have a high DER, and so on. Is this actually the case? Has anyone ever tried to compare the sum of individual defensive statistics to some team stat? Do the teams that look especially good or bad using individual UZRs in 2007 also look bad using other measures?

http://spreadsheets.google.com/pub?key=pfk_WuYpfduzefA8-s0_iYw

There is a lot more data in this spreadsheet, calculated by a reds blogger using what reads to me like the dial method.

Actually it looks like the raw data was obtained from THT (it's what I did).
THT obtained their raw data from Baseball Info solutions- so it may be similar to some of the stuff used in The Fielding Bible.

What I did to figure out fielding runs was to calculate the average RZR (ZR stands for revised zone rate) for each position, as well as the average # of plays out of zone (OOZ) for each position. Each player then got a +/- # of plays made, which I then converted to runs by plugging them into an "average" team's statline and redoing its estimated runs produced.

My end results (which are on my home PC- I'm at work now) look a lot like this guy's results, so I suspect what he did was not too disimilar.

I was just playing around any way I haven't really tried to work with defensive stats before.

But FWIW if you look at this guy's Google spread sheets- the MVP race is between Pujols and Wright

at least in the outfield, the out of zone plays appear to be mostly discretionary balls, as they seem to be the driver behind the huge discrepancies in some players' ratings (e.g. david wright, ichiro, andruw jones, etc).

FWIW I've noticed that most metrics have Pedro Feliz as an outstanding defensive 3B, does this mean hes not "The worst player in baseball" as BPro insists on labeling him?

Still, a team that is full of players that rate very well using UZR or another system should also tend to have a high DER, and so on. Is this actually the case? Has anyone ever tried to compare the sum of individual defensive statistics to some team stat? Do the teams that look especially good or bad using individual UZRs in 2007 also look bad using other measures?

I would also like to know this.

Isn't a Zone Rating a decimal? Isn't this something more like Fielding Runs?

It's pretty easy to convert ZR into runs, Dial did an article a couple years ago on how to do it.

at least in the outfield, the out of zone plays appear to be mostly discretionary balls
Do we know this (the mostly part)? I'm sure that's a lot of it, but I don't know how much and I'd think the non-discretionary balls might have higher the normal associated run values.
***
AROM, where's the link in #19?

FWIW I've noticed that most metrics have Pedro Feliz as an outstanding defensive 3B, does this mean hes not "The worst player in baseball" as BPro insists on labeling him?

Their own metrics consider him a fine defender too: 12 FRAA. They have him at -15 BRAA this year. So, just using their own numbers, he is marginally below average.

So is the outfield in Citizens Bank Ballpark tilted?

#25

For Bonds, Ichiro, Andruw, the differences appear to be between STATS and BIS.

FWIW I've noticed that most metrics have Pedro Feliz as an outstanding defensive 3B, does this mean hes not "The worst player in baseball" as BPro insists on labeling him?

Yes. Worst players in baseball, 2007: Richie Sexson, Josh Barfield, Ray Durham, Nick Punto. Add in part-timers and Alexi Casilla, Shea Hillenbrand, and Andy Gonzalez become good candidates. These are not projections, just 2007 totals. Barfield and Casilla are young enough to improve. Punto should be OK as a utility infielder but not a starting 3B. Gonzalez either improves or goes back to AAA. Durham was very good in 2006, maybe its just an aberration. Biggio is done.

So I'd go with Sexson and Hillenbrand, with contracts telling us Sexson will continue to get a chance to suck next year.

Still, a team that is full of players that rate very well using UZR or another system should also tend to have a high DER, and so on. Is this actually the case? Has anyone ever tried to compare the sum of individual defensive statistics to some team stat? Do the teams that look especially good or bad using individual UZRs in 2007 also look bad using other measures?

I have the numbers for the Rays from 2003-06 using ZR, I haven't input the 2007 final numbers yet. Average for DER tends to be somewhere around .700 each year.

2003: .714 DER, +24 ZR runs
2004: .711 DER, +27 ZR runs
2005: .691 DER, -30 ZR runs
2006: .680 DER, -37 ZR runs

Decent correlation. ZR runs probably isn't the best way to do this though since they're based on how far above or below average the player is compared to others at the position that year while DER is simply based on whether or not a ball is converted into an out.

AROM, where's the link in #19?

Just to confuse you, I posted #19 in this thread and on Tango's blog. Go to Tango's blog, and in comment #19 I posted a link to my spreadsheet.

So is the outfield in Citizens Bank Ballpark tilted?

Nah, it's just an illusion; Burrell always looks like he's running uphill.

Pat Burrell, circa 2040: It was so tough when I was a kid. I had to run 50 feet uphill for any fly ball -- no matter where it was hit.

It isn't just OF defensive metrics which don't agree.
If you compare the JinAz spreadsheet based on RZR and UZR, there are lot of significant differences. Is this methodological or just where your PBP data comes from?

FWIW I've noticed that most metrics have Pedro Feliz as an outstanding defensive 3B, does this mean hes not "The worst player in baseball" as BPro insists on labeling him?

He is an outstanding fielder. The worst player in baseball is some guy in rookie ball. Feliz may well be the worst position player in MLB who plays every day. Most other guys that bad have to at least share time.

Feliz may well be the worst position player in MLB who plays every day.

Based on 2007 I have some other names for you:
Punto; Barfield; Jose Lopez; M. Giles; all can give him a stiff run for that title.

Feliz may well be the worst position player in MLB who plays every day. Most other guys that bad have to at least share time.

Whether you use UZR, RZR or FRAA, after combining offense and defense, Feliz is around slightly below average.

Has anyone ever tried to compare the sum of individual defensive statistics to some team stat? Do the teams that look especially good or bad using individual UZRs in 2007 also look bad using other measures?

There are problems with this, as have been mentioned. Effects of pitching vs defense, ball distribution to different positions, and ballpark effects. Still, in general teams with good DER should have highly rated fielders, and there really isn't any other way to validate defensive metrics. UZR is ballpark adjusted, so you should park adjust your DER to make a comparison. ZR and RZR are not park adjusted, so comparisons are a little easier.

Correlations to DER, 2007:
ZR .652
RZR .575
averaged ZR: .684

Another thing we can do is compare to THT defense rating, which attempts to separate DER into pitching and defense based on the types of batted balls allowed. And we get:

ZR: .769
RZR: .696
averaged ZR: .818

And that is why I went through the trouble of matching up players and averaging zone ratings for everyone. Two heads are better than one.

Is this methodological or just where your PBP data comes from?

I'm not sure if it really matters. At the end of the day when two systems give us vastly different answers to the same question - we have no way of elaborating which is reasonable: For example MGL has ICHIRO! as a very bad CF'er, the guy in the link I provided ICHIRO! as a very good centerfielder. What good would batting stats be if EQR had Prince Albert as a bad hitter and BSR had him as a good one. Or Julio Lugo as good?

There needs to be some type of directional correlation between numbers that proport to estimate the same things. When it it's there, the data is mostly useless, as we have no way to discriminate between what might be accurate and what might not be accurate. Until someone comes up with a better way to estimate opportunity I don't think we be making much progress.

This is comming from the guy who ALWAYS wants more data. The problem is I don't know what this data tells me.