PDA

View Full Version : The Ultimate Defense Thread



Pages : 1 [2]

dougdirt
06-16-2010, 01:32 PM
If we go on Doug's premise, there is no need to deal with positioning, just the result of the play. UZR and the like operate on this premise, due to the lack of positioning data if not out of principle.

However, if we go on lollipop's premise, we need to isolate positioning so as to not credit the player for something beyond his control. In this model, the difficulty of all plays are comprised of not just the batted ball characteristics but also the initial positioning of the player. If a manager tells the CF to shade in to LF and then the CF gets to a ball that he may not have reached otherwise, it would be silly to assert that the CF is better than one who was directed to play it straight and did not reach that ball.

But Rick, do we not add in the times the player 'hit it to the right side' because the coach called for it? What about those weak fly balls or weak grounders on the hit and run where he had to protect the runner on a terrible pitch? What about the single he gets that would have gone directly to the 2B had he not been covering on the hit and run? All of those are called by the manager, but the player gets dinged in offensive stats for those (or gets credit in the one example). Why does the same not apply when referring to value on defense? Again, and I know you know this Rick, but some others seem to keep looking past it, defensive metrics aren't measuring skill. They are measuring results. The two things are not always the same.

RedsManRick
06-16-2010, 01:39 PM
Again I ask, how often does poor positioning lead to a missed play?

I'm thinking it's not very often.

When position data becomes available, it will be fascinating to cross reference it against all balls hit in to a given zone. I don't think we really have a good idea on just how much variation there is.

It's been found that OF UZR is more consistent than IF UZR. While there are all sorts of possible explanations for that, one possible explanation is that positioning matters less for OF than it does for IF.

lollipopcurve
06-16-2010, 01:41 PM
Now I took the routes and put them side by side in the lower right corner. The Blue player clearly ran more to make the play, but why is his play better? Both guys got the ball in the same exact spot, one guy just positioned himself better for the batter at hand. Don't worry about the zones at all. Worry about why you seem to think the blue player is the better defender because he had to run further when both guys made the play, because it honestly doesn't make any sense at all.

You completely misunderstand what I'm saying, somehow.

This example of yours, despite the big ole visual aid, means almost nothing. We don't know if the play was easy or difficult, compared to all other possible plays. Maybe both players had to jog over and camp under the ball. Maybe it was a near sprint for both. Maybe it was much tougher for the blue guy than the red guy. It's quite possible that, using the kind of information I'm talking about, the play would be considered more or less identical for both players, even though one ran farther to make it than another.

What you also don't understand is that just because the ball was hit closer to one position than to another it does not mean that the fielder in the closer position was in a strategically better position. (BABIP should have taught you that the direction/trajectory of batted balls is quite random.) It may have been far likelier that the batter would have hit it closer to the other position, based on any number of variables, like the hitter's hit chart.

If you think certain fielders will be able to consistently place themselves closer to where balls are hit than other fielders, based on some defensive skill (visioning the flight path of balls before they are hit?), you're wrong. Fielders may be able to advantageously position themselves in certain spots based on their skillsets (one guy may be good at going back on balls, another may have a strong arm, whatever...), but the ability to position himself well vs. individual hitters is, for the most part, a product of knowing a hitter's tendencies and the strategy the pitcher will use against that hitter. And that knowledge is created by scouting departments and coaches, not individual players.

The next generation of defensive metrics will be based on the kind of information I'm talking about, Doug.

Over an extended period of time, they'll code all plays by their difficulty, based on things like how far a player has to go to make a play in how much time, and how hard a ball is hit (if it's right at an infielder, it makes a difference). The ability to make the more difficult plays will be correlated with information about fielders -- how fast is the guy? how direct are his routes? what kind of jump does he get? how strong is his arm? -- and they will find strong, positive correlations between those things (some, fascinatingly, I think, stronger than others). The players who show themselves to have the traits that correlate most strongly with getting outs will be the better defenders. The teams that consistently place their fielders closer to batted balls -- if that happens -- will have the better scouting and coaching staffs.

RedsManRick
06-16-2010, 01:45 PM
But Rick, do we not add in the times the player 'hit it to the right side' because the coach called for it? What about those weak fly balls or weak grounders on the hit and run where he had to protect the runner on a terrible pitch? What about the single he gets that would have gone directly to the 2B had he not been covering on the hit and run? All of those are called by the manager, but the player gets dinged in offensive stats for those (or gets credit in the one example). Why does the same not apply when referring to value on defense? Again, and I know you know this Rick, but some others seem to keep looking past it, defensive metrics aren't measuring skill. They are measuring results. The two things are not always the same.

Doug, you're right -- in an ideal world, we'd control for situations such as the one you've described when looking at offensive performance. However, the goal is not to make defensive metrics have the same problems as offensive ones. The goal is to have defensive statistics that are as accurate and reliable as possible.

While offensive statistics are definitely more reliable than defensive ones, we should not treat them as perfect. They still suffer from limited information, if less than defensive metrics do. We should not artificially place ourselves under informational restraints just because other statistics don't have access to comparable information.

To the extent that we can measure player performance and ability isolated from their teammates, coaches, ballparks, etc. -- we should.

nate
06-16-2010, 01:46 PM
When position data becomes available, it will be fascinating to cross reference it against all balls hit in to a given zone. I don't think we really have a good idea on just how much variation there is.

It's been found that OF UZR is more consistent than IF UZR. While there are all sorts of possible explanations for that, one possible explanation is that positioning matters less for OF than it does for IF.

One of the articles that OUReds linked to talked about that and how one of the other systems was more consistent WRT infield while UZR was more consistent WRT outfield.

lollipopcurve
06-16-2010, 01:50 PM
When position data becomes available, it will be fascinating to cross reference it against all balls hit in to a given zone. I don't think we really have a good idea on just how much variation there is.

It's been found that OF UZR is more consistent than IF UZR. While there are all sorts of possible explanations for that, one possible explanation is that positioning matters less for OF than it does for IF.

Of course, because infielders have less time to get to the ball.

nate
06-16-2010, 01:51 PM
OK, as someone who thinks UZR is "good," here are the questions I have:

1. If a three year sample is what would be considered the minimum sample size, why not just display "current" UZR as the past 3 years?

1a. Alternatively, why not have a UZR/450 which shows what UZR is for the past 450 defensive innings? One could also make that kind of a sliding window compared against a player's career UZR number to see if he might be getting better, worse or staying the same.

2. I wonder if UZR would be more palatable if auto-regressed.

RedsManRick
06-16-2010, 01:58 PM
OK, as someone who thinks UZR is "good," here are the questions I have:

1. If a three year sample is what would be considered the minimum sample size, why not just display "current" UZR as the past 3 years?

1a. Alternatively, why not have a UZR/450 which shows what UZR is for the past 450 defensive innings? One could also make that kind of a sliding window compared against a player's career UZR number to see if he might be getting better, worse or staying the same.

Conceptually, I agree 100%. However, Fangraphs does provide UZR broken out by year and position. It's not convenient, but we can easily do the math ourselves.

A recent article on Tango's blog suggests the sample size requirement for comparable reliability to OPS is more like 2.5 times (in games). So we should be looking at 150+150+75 (375 games) * 9 Inn = 3375 defensive innings. Call in 3,000 if we want a round number. UZR3000 sounds pretty cool to me.


2. I wonder if UZR would be more palatable if auto-regressed.

I think it would, but people would still want access to the un-regressed data as well.

dougdirt
06-16-2010, 02:03 PM
You completely misunderstand what I'm saying, somehow.

This example of yours, despite the big ole visual aid, means almost nothing. We don't know if the play was easy or difficult, compared to all other possible plays. Maybe both players had to jog over and camp under the ball. Maybe it was a near sprint for both. Maybe it was much tougher for the blue guy than the red guy. It's quite possible that, using the kind of information I'm talking about, the play would be considered more or less identical for both players, even though one ran farther to make it than another.

What you also don't understand is that just because the ball was hit closer to one position than to another it does not mean that the fielder in the closer position was in a strategically better position. (BABIP should have taught you that the direction/trajectory of batted balls is quite random.) It may have been far likelier that the batter would have hit it closer to the other position, based on any number of variables, like the hitter's hit chart.
Where players hit the ball is not random over a long haul. Most guys have clear tendencies to hit the ball to certain places.



If you think certain fielders will be able to consistently place themselves closer to where balls are hit than other fielders, based on some defensive skill (visioning the flight path of balls before they are hit?), you're wrong.
Am I? What gives you this knowledge? Guys who study the opposition's spray charts are going to have more insight on where to play than guys who don't.



Fielders may be able to advantageously position themselves in certain spots based on their skillsets (one guy may be good at going back on balls, another may have a strong arm, whatever...),
Agreed with this 100%. A shortstop with average range but an elite arm should play deeper than a shortstop with excellent range and an average arm.



but the ability to position himself well vs. individual hitters is, for the most part, a product of knowing a hitter's tendencies and the strategy the pitcher will use against that hitter. And that knowledge is created by scouting departments and coaches, not individual players.Created by those guys yes, but not always implemented by them. It is up to the players to make it happen (or coaches in some cases... who aren't positioning the players on EVERY play, but the player himself can indeed position himself on every play). But again it comes back to what does happen versus what could happen. In terms of applying value to a specific season, what could happen is meaningless. What does happen means a lot.



The next generation of defensive metrics will be based on the kind of information I'm talking about, Doug. And they will still not tell us much about what did happen that we don't already know. It may tell us who could be the best guy if they played to their full potential, but that doesn't give us any actual value as to what did happen.



Over an extended period of time, they'll code all plays by their difficulty, based on things like how far a player has to go to make a play in how much time, and how hard a ball is hit (if it's right at an infielder, it makes a difference).
Again, how far the player has to go means nothing in terms of the value of recording that specific out. Assuming the ball is hit on an extremely similar trajectory and speed, as long as the out is recorded, its of the same value for both players regardless of how far one had to go to make it.



The ability to make the more difficult plays will be correlated with information about fielders -- how fast is the guy? how direct are his routes? what kind of jump does he get? -- and they will find strong, positive correlations between those things (some, fascinatingly, I think, stronger than others).
The things you are talking about will help us know who has the best raw skills, not who is the best player. Jay Bruce absolutely 100% has the best raw skills of anyone on the Reds and its probably not that close. Would you argue that he is the Reds best player? No, because his raw skills aren't at the same point of his actual skills at this point in time. In terms of actual value of what does happen on the field in a season, what matters is the outs recorded or not recorded. Not how far someone had to run to make the play, just where the ball landed (and to your point, the trajectory/speed does matter) or was caught. If player B only had to run 45 feet to make the same play on the ball that player C had to run 60 feet for doesn't make player C's play any more or less valuable.

dougdirt
06-16-2010, 02:05 PM
To the extent that we can measure player performance and ability isolated from their teammates, coaches, ballparks, etc. -- we should.

Why should we not punish a players defensive value if he was positioned poorly, be it his fault or not? It either happened or it didn't. It doesn't tell us his skill, it tells us what happened.

lollipopcurve
06-16-2010, 02:13 PM
And they will still not tell us much about what did happen that we don't already know.

Oh yes they will.

A whole new set of stats will be created to tell us what kinds of plays players made/didn't make. It will be based on the degrees of difficulty inherent in the chances players receive. Just like UZR tells us a bunch of new stuff in its own language.

The difference is, the new generation will do a much better job of isolating players' skillsets, explaining why some players had better results than others and, importantly, predicting why some players are likely to defend better in the future than others. I think it's highly likely that the new generation of stats will provide data that can be considered significant in significantly less time than UZR now can.

lollipopcurve
06-16-2010, 02:22 PM
If player B only had to run 45 feet to make the same play on the ball that player C had to run 60 feet for doesn't make player C's play any more or less valuable.

We're not arguing about the value of the play, we're talking about the value of the player. (And, accordingly, the value of one player vs. another.) The player who can cover more ground, take better routes, get better jumps/first steps, get more on his throws -- in the right mix, all of these things will correlate to one degree or another with getting more outs -- is the better defender. And in the future they'll be able to collect that information about players, and use it, in far less time than it takes UZR to give you anything you feel is trustworthy. It's going to take a while, but it's going to happen.

nate
06-16-2010, 02:34 PM
Conceptually, I agree 100%. However, Fangraphs does provide UZR broken out by year and position. It's not convenient, but we can easily do the math ourselves.

Right. I'm just saying when looking at a guy like Jeter who has these wildly fluctuating values, it might be more approachable when displayed as "the last 3000 innings."


A recent article on Tango's blog suggests the sample size requirement for comparable reliability to OPS is more like 2.5 times (in games). So we should be looking at 150+150+75 (375 games) * 9 Inn = 3375 defensive innings. Call in 3,000 if we want a round number. UZR3000 sounds pretty cool to me.

UZR3k...sweet!


I think it would, but people would still want access to the un-regressed data as well.

Yes.

RedsManRick
06-16-2010, 02:49 PM
Why should we not punish a players defensive value if he was positioned poorly, be it his fault or not? It either happened or it didn't. It doesn't tell us his skill, it tells us what happened.

It depends on the question we're trying to address. If we just want to record what happened as a record of team performance, fine. But if you want to use the stat as a measure of player skill, presumably for the purpose of planning/projection, you'd want to know how much of his defense was his responsibility.

Are you suggesting that we shouldn't be using UZR type metrics to estimate player skill?

dougdirt
06-16-2010, 02:50 PM
We're not arguing about the value of the play, we're talking about the value of the player. (And, accordingly, the value of one player vs. another.) The player who can cover more ground, take better routes, get better jumps/first steps, get more on his throws -- in the right mix, all of these things will correlate to one degree or another with getting more outs -- is the better defender. And in the future they'll be able to collect that information about players, and use it, in far less time than it takes UZR to give you anything you feel is trustworthy. It's going to take a while, but it's going to happen.

The value of a player to a team in a given season is only calculable by what he did or didn't do. No one has chances that are created equally. If Albert Pujols got to face the same exact pitches that Juan Pierre got, he might OPS 1.400 every year. He doesn't get those pitches though. So he only OPS's 1.000 or so.

You seem to be equating skill with production and that isn't reality. The guy with the better defensive tools isn't always the better defender. Willy Taveras had outstanding defensive tools, but was merely average in the field. Chase Utley has slightly above average defensive tools and is the best defensive second baseman in baseball. What matters is how you use your tools, not how good those tools are.

dougdirt
06-16-2010, 02:56 PM
It depends on the question we're trying to address. If we just want to record what happened as a record of team performance, fine. But if you want to use the stat as a measure of player skill, presumably for the purpose of planning/projection, you'd want to know how much of his defense was his responsibility.

Are you suggesting that we shouldn't be using UZR type metrics to estimate player skill?
I think we shouldn't use UZR to measure tools. Some guys have the toolset to be great defenders and just aren't for different reasons. Some guys have average toolsets, but are good defenders for different reasons. UZR doesn't always show us those tools. Sometimes it does, just like sometimes offensive stats can show us. But sometimes those tools don't always translate into actual skills.

Lollipop seems to want stats to measure ones tools. How far a guy ran to make a play for example. That tells us very little in terms of his defensive ability compared to another player who may not have had to run as far to make the same play. He seems to think otherwise. I 100% disagree with that assessment.

RedsManRick
06-16-2010, 03:05 PM
The value of a player to a team in a given season is only calculable by what he did or didn't do. No one has chances that are created equally. If Albert Pujols got to face the same exact pitches that Juan Pierre got, he might OPS 1.400 every year. He doesn't get those pitches though. So he only OPS's 1.000 or so.

You seem to be equating skill with production and that isn't reality. The guy with the better defensive tools isn't always the better defender. Willy Taveras had outstanding defensive tools, but was merely average in the field. Chase Utley has slightly above average defensive tools and is the best defensive second baseman in baseball. What matters is how you use your tools, not how good those tools are.

Should we not even try to be isolating player performance from context in any way? By your logic, we shouldn't control for park effects or anything else.

You don't seem to be willing to concede that the method of measurement needs to be tailored to the purpose. Knowing exactly what happened on the field is one thing. Trying to figure out how players are likely to perform in the future is another.

dougdirt
06-16-2010, 03:09 PM
Should we not even try to be isolating player performance from context in any way? By your logic, we shouldn't control for park effects or anything else.

You don't seem to be willing to concede that the method of measurement needs to be tailored to the purpose. Knowing exactly what happened on the field is one thing. Trying to figure out how players are likely to perform in the future is another.

For park effects, in terms of dimensions, absolutely. It is the old 'Fenway and left field' issue. But in terms of adding value to a players 'worth', what he is likely to do in the future is irrelevant. Defensively, if I want to know what a guy could do in the future, I go to my scouts and ask if we can improve his defense. Once I get that answer, I then adjust an aging curve to figure out where he is going to likely be defensively moving forward.

lollipopcurve
06-16-2010, 03:29 PM
The value of a player to a team in a given season is only calculable by what he did or didn't do.

So, you calculate stuff like "how often did he get a good jump" and "how often did he take a direct route" and "how fast did he close on balls hit a long but catchable distance from his original position?" It's a whole new set of data.

Remember that a year's worth of UZR data is not considered very reliable.


No one has chances that are created equally.

Your point?


You seem to be equating skill with production and that isn't reality.

You really do not understand at all. I'm saying that with certain precise information about the velocity and trajectory of batted balls, the original position of the fielder, and precise information about the movements of the fielder (including "jump" and route), we will be able to establish correlations between characteristics belonging to the fielder (tools, skills, whatever you want to call them) and outs. Did he miss the ball because he was too far from it (not his fault), or did he get a bad jump (his fault)? Questions like that will become much more answerable.


The guy with the better defensive tools isn't always the better defender. Willy Taveras had outstanding defensive tools, but was merely average in the field.

I think you need to define what you mention by tools. Taveras has speed, but there's a lot more to defense than speed, and if you've read my posts with any understanding, you know that I've listed several tools/skills that need to be (and will be) measured, in addition to speed. Jumps, routes, "accuracy" in leaving one's feet, for example.


What matters is how you use your tools, not how good those tools are.

Of course. But the argument you're making is similar to the one Dusty Baker makes when, advocating for an attacking approach on offense, he says "the point of the game is to score runs." Of course, the point of defense is to record outs, but what are the ingredients necessary for a defender to be good at recording outs? If you don't know what those things are, you're going to have a hard time finding good defensive players for your team.

lollipopcurve
06-16-2010, 03:36 PM
Why should we not punish a players defensive value if he was positioned poorly, be it his fault or not?

First, how do you know if a player is positioned poorly?

Is it possible for a poorly positioned player to have a ball hit directly at him?

bucksfan2
06-16-2010, 04:38 PM
Doug you keep mentioning skill. Isn't that a little misleading? A 5 tool outfielder (one of Jimbo's favorites) often has more skills than your run of the mill outfielder, but it isn't for certain that that toolsy, skillful, outfielder is a better defender. Its not what skills you have, its how you put to use those skills, correct?

dougdirt
06-16-2010, 04:55 PM
Doug you keep mentioning skill. Isn't that a little misleading? A 5 tool outfielder (one of Jimbo's favorites) often has more skills than your run of the mill outfielder, but it isn't for certain that that toolsy, skillful, outfielder is a better defender. Its not what skills you have, its how you put to use those skills, correct?

Correct. Which is why I could care less about how far a guy has to run to catch the ball compared to someone else as long as both guys get the same ball. One play doesn't make one guy or the other better because we don't know based off of that play.

dougdirt
06-16-2010, 04:58 PM
First, how do you know if a player is positioned poorly?
Based upon scouting reports and where the guy is initially lined up. If a guys spray chart suggests he sprays the right center gap 50% more than the left center gap, if the guy isn't shading to the right center gap, he is positioned poorly.



Is it possible for a poorly positioned player to have a ball hit directly at him?

Of course its possible for a poorly positioned player to have a ball hit directly at him. Why wouldn't it be possible?

lollipopcurve
06-16-2010, 05:12 PM
Correct. Which is why I could care less about how far a guy has to run to catch the ball compared to someone else as long as both guys get the same ball. One play doesn't make one guy or the other better because we don't know based off of that play.

What if one guy gets to the ball and another, on more or less the exact same play, doesn't? Don't you want to collect that information? And wouldn't it tell you something important?


Based upon scouting reports and where the guy is initially lined up. If a guys spray chart suggests he sprays the right center gap 50% more than the left center gap, if the guy isn't shading to the right center gap, he is positioned poorly.

Exactly. It is a function of the hitter (and the approach teams take in pitching to him). It is not a function of the defender applying his skill. That is, you don't tell a young third baseman, "Position yourself like Rolen does, he's always in good position." You tell him, "Play where we tell you -- we've got charts on the hitters and know their tendencies." Proper positioning is a product of information collected by scouts and implemented by coaching staffs. It is not a product of a player's skill.


Of course its possible for a poorly positioned player to have a ball hit directly at him. Why wouldn't it be possible?

Just checking.

dougdirt
06-16-2010, 05:41 PM
What if one guy gets to the ball and another, on more or less the exact same play, doesn't? Don't you want to collect that information? And wouldn't it tell you something important?
We already get that information with UZR and other similar zone rating systems.



Exactly. It is a function of the hitter (and the approach teams take in pitching to him). It is not a function of the defender applying his skill. That is, you don't tell a young third baseman, "Position yourself like Rolen does, he's always in good position." You tell him, "Play where we tell you -- we've got charts on the hitters and know their tendencies." Proper positioning is a product of information collected by scouts and implemented by coaching staffs. It is not a product of a player's skill.
We seem to be on different things here. To me, knowing where to position yourself is a part of your defense. Whether it is a 'skill' or not, it counts. If it makes you better or worse in terms of how you are able to use your range, that goes on you, the player because you either did or did not make the play because of it. Much like the scouting reports for pitchers.... if your scouting department sucks, you are going to suffer at the plate against some pitchers. That is still looked at as on the player is it not? Yes, it is.

TheNext44
06-16-2010, 05:58 PM
YouTube - Lamb chops play-Along (http://www.youtube.com/watch?v=HNTxr2NJHa0&feature=fvw)

Scrap Irony
06-16-2010, 10:32 PM
We seem to be on different things here. To me, knowing where to position yourself is a part of your defense. Whether it is a 'skill' or not, it counts. If it makes you better or worse in terms of how you are able to use your range, that goes on you, the player because you either did or did not make the play because of it. Much like the scouting reports for pitchers.... if your scouting department sucks, you are going to suffer at the plate against some pitchers. That is still looked at as on the player is it not? Yes, it is.

There's not an OF in baseball that positions himself. Positioning is never "on the player"-- it's on the team and the scouting department.

dougdirt
06-16-2010, 10:34 PM
There's not an OF in baseball that positions himself. Positioning is never "on the player"-- it's on the team and the scouting department.

Where do you get this information from?

jojo
06-16-2010, 10:43 PM
There's not an OF in baseball that positions himself. Positioning is never "on the player"-- it's on the team and the scouting department.

No one tells Ichiro where to stand. Its in his contract.

lollipopcurve
06-17-2010, 10:52 AM
We already get that information with UZR and other similar zone rating systems.

No, we don't. Not with the precision I'm talking about. There's a lot of variability in defensive chances that isn't being captured in the current metrics.


To me, knowing where to position yourself is a part of your defense.

Knowing the scouting reports, you mean. Not a skill inherent to defensive play.


If it makes you better or worse in terms of how you are able to use your range, that goes on you, the player because you either did or did not make the play because of it.

Again, where a player positions himself is not a skill. You could stick Orlando Cabrera in exactly the same place on every play, and he would have the opportunity to make some plays and not make others simply because he is in the SS position. It's his other skills that actually allow him to execute plays.


Much like the scouting reports for pitchers.... if your scouting department sucks, you are going to suffer at the plate against some pitchers. That is still looked at as on the player is it not? Yes, it is.

Who cares how it's looked on? Try to think for yourself. Hitting and defense are not the same. And we're talking about defense here.

Ghosts of 1990
06-17-2010, 11:28 AM
My buddy asked me a question the other day I didn't know the answer to. We were talking about Range Factor as a statistic and how it shows the value of an outfielder defensively with the amount of putouts.

He asked how is that a fair statistic when someone plays half their games in the OF of say, Great American while someone he is in competition with might play half of their games in a much more spacious outfield like Dodger Stadium or Petco?

RedsManRick
06-17-2010, 11:36 AM
Again, where a player positions himself is not a skill. You could stick Orlando Cabrera in exactly the same place on every play, and he would have the opportunity to make some plays and not make others simply because he is in the SS position. It's his other skills that actually allow him to execute plays.

But this isn't completely true either. The pattern of batted balls is not completely random. To a degree, there is a unique pattern of probability associated with each batter and circumstance. To the extent that a fielder is positioning himself (and I would think we can agree positioning is the result of the combined input of the player and coaches), he can put himself in a position to have a greater number of opportunities.

Sure other skills determine whether or not he can move 11 feet laterally to his right in 0.46 seconds. But if the player, through his own awareness of the likelihood of a ball to be hit in that direction, positions himself such that he's only 7 feet away, does he not deserve credit for that?

One could make a comparison with a better who, through his superior knowledge and video review, knows to sit on an outside change-up on a 1-2 count and thus is able to smack a single to RF where a player with a similar batting eye, plate discipline, hand-eye-coordination, and power would not have done been able to do so. Maybe he watched some video with his batting coach before the game to get that piece of knowledge. Should we not credit the hitter for that because of that?

We care about the batters ability to get a hit and don't attempt to isolate the influence of his coaches. Nor do we control for the quality of the pitchers he faces nor the quality of pitches those pitchers throw.

I agree that we don't need to have the exact same processes in evaluating both offense and defense. But I do find it interesting that we seem to be holding defensive evaluation to a higher standard. As somebody who does impact measurement for a living, I would argue that there are some things resulting from complex systems which just cannot be reliably measured.

Given both the noise in the data, the variability in performance, and (a point which people too often ignore) the constantly changing talent level of players over time, we may have reached a point of diminishing returns on defense.

I wouldn't not be surprised if 5 years from now we look on UZR the way we look at OPS. Sure, it's not the most accurate measure. We'll come up with more accurate (and complicated) ways of assessing and projecting defensive performance. But we'll have reached a point at which we're more or less close enough to make informed decisions and the greater gains are to be found elsewhere.

lollipopcurve
06-17-2010, 12:05 PM
But this isn't completely true either. The pattern of batted balls is not completely random. To a degree, there is a unique pattern of probability associated with each batter and circumstance. To the extent that a fielder is positioning himself (and I would think we can agree positioning is the result of the combined input of the player and coaches), he can put himself in a position to have a greater number of opportunities.

True. But, in my view, it is the information collected by scouting departments and staffs that primarily determine where the best positioning is. They have access to the data that can lead one to safe conclusions in this regard. For the most part, major league players just follow instructions about positioning.


Sure other skills determine whether or not he can move 11 feet laterally to his right in 0.46 seconds. But if the player, through his own awareness of the likelihood of a ball to be hit in that direction, positions himself such that he's only 7 feet away, does he not deserve credit for that?

Reams of data will eventually be able to determine if one player can consistently position himself more closely to batted balls than his peers (including his own teammates, who will be implementing the same approach to positioning as he will). I suspect this will go the way of the clutch hitter -- a myth. But, if not, sure -- he'll deserve credit.


One could make a comparison with a better who, through his superior knowledge and video review, knows to sit on an outside change-up on a 1-2 count and thus is able to smack a single to RF where a player with a similar batting eye, plate discipline, hand-eye-coordination, and power would not have done been able to do so. Maybe he watched some video with his batting coach before the game to get that piece of knowledge. Should we not credit the hitter for that because of that?

The difference here is that a hitter is in a cat-and-mouse with a pitcher. The pitcher is armed with a scouting report, too. You can't really separate to what extent the pitcher and hitter are successfully/unsuccessfully doing what the scouting report tells them to, and oftentimes the exigency of the situation (count, score, etc) may override a scouting report. With defense, these issues do not arise.


Nor do we control for the quality of the pitchers he faces nor the quality of pitches those pitchers throw.

Advances will be made for hitters in a similar vein as they will be made for defense.


I agree that we don't need to have the exact same processes in evaluating both offense and defense. But I do find it interesting that we seem to be holding defensive evaluation to a higher standard. As somebody who does impact measurement for a living, I would argue that there are some things resulting from complex systems which just cannot be reliably measured.

You're right that they can't be measured now. But you do need to understand that defense and offense are really, really different.


I wouldn't not be surprised if 5 years from now we look on UZR the way we look at OPS. Sure, it's not the most accurate measure. We'll come up with more accurate (and complicated) ways of assessing and projecting defensive performance. But we'll have reached a point at which we're more or less close enough to make informed decisions and the greater gains are to be found elsewhere.

I think the potential gains for defensive metrics are huge. Less so for offense. My guess? Peak defensive performance happens at a younger age than peak offensive performance.

dougdirt
06-17-2010, 12:18 PM
My buddy asked me a question the other day I didn't know the answer to. We were talking about Range Factor as a statistic and how it shows the value of an outfielder defensively with the amount of putouts.

He asked how is that a fair statistic when someone plays half their games in the OF of say, Great American while someone he is in competition with might play half of their games in a much more spacious outfield like Dodger Stadium or Petco?

Range factor isn't a very good defensive stat unless you are comparing two players at the same position in the same season from the same team with no real pitching difference and have a decent sample size on both guys.

dougdirt
06-17-2010, 12:27 PM
True. But, in my view, it is the information collected by scouting departments and staffs that primarily determine where the best positioning is. They have access to the data that can lead one to safe conclusions in this regard. For the most part, major league players just follow instructions about positioning.
What proof do you have of this actually happening? You keep saying it as if it were a fact. You also continue to ignore that everyone has scouting reports for everything from the scouts. Pitchers do on hitters. Hitters do on pitchers. Managers do on the other managers. Should we just stop crediting the players and managers and give credit to the scouts?



The difference here is that a hitter is in a cat-and-mouse with a pitcher. The pitcher is armed with a scouting report, too. You can't really separate to what extent the pitcher and hitter are successfully/unsuccessfully doing what the scouting report tells them to, and oftentimes the exigency of the situation (count, score, etc) may override a scouting report. With defense, these issues do not arise.
Yes, they do. Shift is on versus a lefty. The lefty drops a bunt down the 3B line. Single. Shift is on versus a lefty, he bails out on a ball to hit a grounder toward 3B. Single. It happens.



I think the potential gains for defensive metrics are huge. Less so for offense. My guess? Peak defensive performance happens at a younger age than peak offensive performance.We already know this to be true. Defensive players peak around 24/25. Offensive players around 27-29.

westofyou
06-17-2010, 12:32 PM
What proof do you have of this actually happening? You keep saying it as if it were a fact. You also continue to ignore that everyone has scouting reports for everything from the scouts. Pitchers do on hitters. Hitters do on pitchers. Managers do on the other managers. Should we just stop crediting the players and managers and give credit to the scouts?


Yes, they do. Shift is on versus a lefty. The lefty drops a bunt down the 3B line. Single. Shift is on versus a lefty, he bails out on a ball to hit a grounder toward 3B. Single. It happens.

We already know this to be true. Defensive players peak around 24/25. Offensive players around 27-29.
http://msn.foxsports.com/mlb/story/pittsburgh-pirates-need-to-make-decision-on-manager-john-russell-061710

Huntington even is involved in strategy, telling the Pittsburgh Post-Gazette earlier this season that he and his statistical analysts play significant roles in forming the team’s defensive shifts.

dougdirt
06-17-2010, 12:36 PM
http://msn.foxsports.com/mlb/story/pittsburgh-pirates-need-to-make-decision-on-manager-john-russell-061710

Significant in what way? Significant to never doing so before they showed up? Significant to we do it on every play? Significant to we do it to 3 or 4 guys per team? There is a giant gap between what could be meant in that quote.

westofyou
06-17-2010, 12:41 PM
Significant in what way? Significant to never doing so before they showed up? Significant to we do it on every play? Significant to we do it to 3 or 4 guys per team? There is a giant gap between what could be meant in that quote.

For further proof that your point might be wrong will you require a notarized statement?

lollipopcurve
06-17-2010, 12:47 PM
What proof do you have of this actually happening? You keep saying it as if it were a fact.

Ever seen a hit chart?


Yes, they do. Shift is on versus a lefty. The lefty drops a bunt down the 3B line. Single. Shift is on versus a lefty, he bails out on a ball to hit a grounder toward 3B. Single. It happens.

You're completely misunderstanding the exchange between Rick and me, and I can't figure out what you're trying to say anyway. You can't be arguing that some hitters have the ability to "hit away from" the defensive positioning on a consistent basis, can you? Show me a hitter with no discernable pattern in his hit charts?

Other than Wee Willie Keeler, of course.

Ghosts of 1990
06-17-2010, 12:47 PM
Range factor isn't a very good defensive stat unless you are comparing two players at the same position in the same season from the same team with no real pitching difference and have a decent sample size on both guys.

Thanks Doug

lollipopcurve
06-17-2010, 12:51 PM
We already know this to be true. Defensive players peak around 24/25. Offensive players around 27-29.

I'll guess it could be younger for defense. Just put the kids in major league environments.

And there's a whole lot more to be learned about defense. A whole lot.

dougdirt
06-17-2010, 12:58 PM
For further proof that your point might be wrong will you require a notarized statement?

I don't think that quote tells me at all how much a team positions a player. Its a broad statement. It could mean "before we came in, the coaching staff never positioned its defenders and now we do it more than ever, 5 or 6 times a game we will position our fielders accordingly" or it could mean "on nearly every play we position our fielders according to the batters batted ball tendencies". There is a giant gap in what could be meant there.

And lets say that they do it on every play, it still comes down to whether the play was made or not when determining the value of the player in that season. If a player gets poor scouting reports on every pitcher in the league and struggles at the plate because of it, it is still on that player and no one questions it was the teams fault. Its a giant double standard we are having in regards to offensive and defensive values within this conversation.

dougdirt
06-17-2010, 01:02 PM
Ever seen a hit chart?
Of course I have. Heck, I have even made them. Not sure how this is 'proof' though. Ever seen a scouting report on a pitcher? Does that mean that every player absolutely 100% knows the report? Does it mean that they absolutely 100% follow it every time?




You're completely misunderstanding the exchange between Rick and me, and I can't figure out what you're trying to say anyway. You can't be arguing that some hitters have the ability to "hit away from" the defensive positioning on a consistent basis, can you? Show me a hitter with no discernable pattern in his hit charts?

Other than Wee Willie Keeler, of course.

You said that defenders don't face a cat and mouse game like hitters do at the plate. I suggested that they do because when teams shift to a large degree we see guys slap it where they aren't. Thus making a cat and mouse game for defenders.

Almost all hitters have a tendency to hit the ball to spots. No argument at all there. But if you shift too much and the 3B is essentially playing shortstop or the 2B is playing directly behind the 2B bag, guys have been able to hit the ball where they aren't to get hits. On purpose (you can tell by the lackluster swings on some of them).

westofyou
06-17-2010, 01:03 PM
I don't think that quote tells me at all how much a team positions a player. Its a broad statement. It could mean "before we came in, the coaching staff never positioned its defenders and now we do it more than ever, 5 or 6 times a game we will position our fielders accordingly" or it could mean "on nearly every play we position our fielders according to the batters batted ball tendencies". There is a giant gap in what could be meant there.

And lets say that they do it on every play, it still comes down to whether the play was made or not when determining the value of the player in that season. If a player gets poor scouting reports on every pitcher in the league and struggles at the plate because of it, it is still on that player and no one questions it was the teams fault. Its a giant double standard we are having in regards to offensive and defensive values within this conversation.

significance [sɪgˈnɪfɪkəns]
n
1. consequence or importance
2. something signified, expressed, or intended

dougdirt
06-17-2010, 01:04 PM
And there's a whole lot more to be learned about defense. A whole lot.

There is. But I also believe where we are at now gives us a very strong hold on what actually happens in the games now in terms of recording outs compared to your positional peers.

What we don't know much about is the scouting side of the players 'tools' or how those may allow the player to better perform one day when solely using the defensive statistics.

lollipopcurve
06-17-2010, 01:08 PM
You said that defenders don't face a cat and mouse game like hitters do at the plate. I suggested that they do because when teams shift to a large degree we see guys slap it where they aren't. Thus making a cat and mouse game for defenders.

Oh yeah, all the time. Griffey and Dunn vs. the shift? Stroked down the 3rd base line all day. Hell, hitters are wearing out the right and left field foul lines with regularity, not to mention the gaps.

Doug, consider for a moment the lessons of BABIP. Forget any specific statistic associated with it, and think about what it means.

RedsManRick
06-17-2010, 01:08 PM
I'll guess it could be younger for defense. Just put the kids in major league environments.

And there's a whole lot more to be learned about defense. A whole lot.

To be learned, sure? But in a way that will fundamentally shift our understanding and valuation of defensive contributions? Obviously we don't know what we don't know, but I think we're getting closer to a point where future gains are marginal. Time will tell. Positioning and the ability of hitters to change their hitting styles in reaction to changes in positioning is likely the big area of strategic growth.

For example, LH sluggers tend to only hit the ball in the air when and if they hit it to the left side. So maybe it would be more effective to shift the SS to left center, the LF closer to the line, and have the 3B cover the whole left side of the IF.

dougdirt
06-17-2010, 01:10 PM
significance [sɪgˈnɪfɪkəns]
n
1. consequence or importance
2. something signified, expressed, or intended

Thanks professor. Before then I had no idea what it meant.

0 to 5 is significant.
As is 30. The problem is, we don't know the baseline, so we can't tell just how significant the Pirates play is.

Brutus
06-17-2010, 01:10 PM
I thought it was universally understood that teams position their defenders a large majority of the time, based on the scouting reports of the hitters. I'm almost shocked that this is even being debated.

westofyou
06-17-2010, 01:11 PM
Thanks professor. Before then I had no idea what it meant.

That's what I thought.

lollipopcurve
06-17-2010, 01:13 PM
To be learned, sure? But in a way that will fundamentally shift our understanding and valuation of defensive contributions? Obviously we don't know what we don't know, but I think we're getting closer to a point where future gains are marginal. Time will tell.

Fundamental shifts in understanding? Oh hell yes. That should be obvious. In terms of what gains are made in terms of valuation, I would say that defensive metrics will stabilize and be a more reliable tool on a year to year basis. Seems significant to me.

dougdirt
06-17-2010, 01:14 PM
Oh yeah, all the time. Griffey and Dunn vs. the shift? Stroked down the 3rd base line all day. Hell, hitters are wearing out the right and left field foul lines with regularity, not to mention the gaps.

Doug, consider for a moment the lessons of BABIP. Forget any specific statistic associated with it, and think about what it means.

All the time, no. But 5 times a year they did and that was 5 hits. Those hits all counted.

As for BABIP, I know what it means. Not sure which lesson you are trying to get me to know here though, so just spell it out for me so we can go from there.

RedsManRick
06-17-2010, 01:15 PM
I thought it was universally understood that teams position their defenders a large majority of the time, based on the scouting reports of the hitters. I'm almost shocked that this is even being debated.

But you can still have fielders take an extra step or two before the ball is hit. Andruw Jones was famous for breaking on the ball before contact was even made. Chase Ultey does the same.

We'll never know who's decision it is when a fielder is set up in a specific spot -- maybe the fielder is set up in the general spot the coach wanted but made another adjustment himself. But to Doug's point -- let's at least measure it and then we can decide who gets the credit.

dougdirt
06-17-2010, 01:15 PM
Fundamental shifts in understanding? Oh hell yes. That should be obvious. In terms of what gains are made in terms of valuation, I would say that defensive metrics will stabilize and be a more reliable tool on a year to year basis. Seems significant to me.

Valuation in terms of what? Valuation in terms of what has already happened? No. Outs are outs and hits are hits. You either made outs and prevented hits, or you didn't. We currently get all of this data.

Valuation in terms of who could one day be the best defender? Yeah, we don't statistically have data that supports this.

Scrap Irony
06-17-2010, 01:17 PM
For further proof that your point might be wrong will you require a notarized statement?

Heh.

pedro
06-17-2010, 01:17 PM
Thanks professor. Before then I had no idea what it meant.



Classy Doug.

You just can't ever accept that you don't know everything can you?

dougdirt
06-17-2010, 01:19 PM
Classy Doug.

You just can't ever accept that you don't know everything can you?

Yeah, it was real classy to respond with the definition of a word. Super classy.

lollipopcurve
06-17-2010, 01:19 PM
Valuation in terms of what? Valuation in terms of what has already happened? No. Outs are outs and hits are hits. You either made outs and prevented hits, or you didn't. We currently get all of this data.

You're making me tired.

You will start to unpack the outs and hits in terms of how difficult the play truly was. That an out or hit was recorded really tells us very little about to what extent a defender brought a significant level of skill to the play.

You just can't get out of that hit or out box, it seems.

westofyou
06-17-2010, 01:23 PM
Yeah, it was real classy to respond with the definition of a word. Super classy.
I'm styling too.

pedro
06-17-2010, 01:23 PM
Yeah, it was real classy to respond with the definition of a word. Super classy.

He was just trying to help you out "Professor".

dougdirt
06-17-2010, 01:24 PM
You're making me tired.

You will start to unpack the outs and hits in terms of how difficult the play truly was. That an out or hit was recorded really tells us very little about to what extent a defender brought a significant level of skill to the play.

You just can't get out of that hit or out box, it seems.

You seem to be of the idea that a players skills reflect a players value rather than the players production reflecting that value.

In a given season skill and production aren't always going to line up. As a hitting example, lets just look at a guy who carries a .240 BABIP on the season and only hits .250/.340/.455. His production that season wasn't up to par for his usual self. However the next season his BABIP is at .305 and he hits .290/.385/.520. Did his skill change? No, it probably didn't. But his production certainly did. It is the same with defense. Skills don't always lead to actual production. When it comes to current value in terms of what did happen on the playing field, skills don't matter. Results do. The difficulty of the play matters, but going back 13 pages or so, you seem to think running further for the ball makes a play more difficult and thus the player who ran further is the better defender despite another guy also making the same play who ran less because he was positioned better for the play. I strongly disagree with that premise because positioning is part of defense and if you gain an advantage by being positioned correctly it allows you to play your skills up.

RedsManRick
06-17-2010, 01:25 PM
Valuation in terms of what? Valuation in terms of what has already happened? No. Outs are outs and hits are hits. You either made outs and prevented hits, or you didn't. We currently get all of this data.

Valuation in terms of who could one day be the best defender? Yeah, we don't statistically have data that supports this.

Doug, the foundation of UZR is data pertaining to the likelihood of a given batted ball being turned in to an out. Currently that determination is made solely by zone. Lolliopop is suggesting that the fielder's initial position could be a factor in that as well -- at least to the extent that we can break the play in to the portion of it which is due to positioning and the portion due to the movement of the fielder once the ball was hit.

dougdirt
06-17-2010, 01:28 PM
Doug, the foundation of UZR is data pertaining to the likelihood of a given batted ball being turned in to an out. Currently that determination is made solely by zone. Lolliopop is suggesting that the fielder's initial position could be a factor in that as well -- at least to the extent that we can break the play in to the portion of it which is due to positioning and the portion due to the movement of the fielder once the ball was hit.

Certainly the position of the player at the start of the play has to do with turning a ball into an out or not. I never said it didn't. I am suggesting that is part of the value of a defender though. If player X and player Y start 15 feet apart, but both catch the ball in quadrant XYZ that neither one made a more or less valuable play because both made the out. He seems to be suggesting that one guys play is more valuable.

RedsManRick
06-17-2010, 01:36 PM
Certainly the position of the player at the start of the play has to do with turning a ball into an out or not. I never said it didn't. I am suggesting that is part of the value of a defender though. If player X and player Y start 15 feet apart, but both catch the ball in quadrant XYZ that neither one made a more or less valuable play because both made the out. He seems to be suggesting that one guys play is more valuable.

I think it will be interesting to see how much variation we see in positioning, especially between players on the same team playing the same position.

I tend to take the middle position that while positioning is significantly influenced by the coach/manager, it's not something we will ever be able to tease out. Thus, the player will be given credit for those plays. However, I do think we should break the value up in to positioning and movement components the same way we break out OBP from Slugging. You can have players who are good at one and bad at the other (or vice versa) and end up having similar overall defensive value. However, it would give us some insight in to how much a player's defense can be improved.

Brutus
06-17-2010, 01:40 PM
But you can still have fielders take an extra step or two before the ball is hit. Andruw Jones was famous for breaking on the ball before contact was even made. Chase Ultey does the same.

We'll never know who's decision it is when a fielder is set up in a specific spot -- maybe the fielder is set up in the general spot the coach wanted but made another adjustment himself. But to Doug's point -- let's at least measure it and then we can decide who gets the credit.

Pre-pitch positioning is a lot bigger deal than the select few elite fielders who are such a natural defensively that they're able to move before the ball is hit. What you describe is not something that happens very often.

Positioning by scouting reports is something that can make a huge difference. If a team moves the outfielders over 15 feet, it literally takes them into another "zone." That can make or break the difference into whether they get to a ball that UZR says they should get to. That's the point that I do agree is being severely underestimated.

dougdirt
06-17-2010, 01:45 PM
Positioning by scouting reports is something that can make a huge difference. If a team moves the outfielders over 15 feet, it literally takes them into another "zone." That can make or break the difference into whether they get to a ball that UZR says they should get to. That's the point that I do agree is being severely underestimated.

But here is where I am at with this....

Regardless of where the guy started at, the ball went into a zone where most guys make the play on. Be it the managers fault for the position or the players fault, most other guys make that play. That goes against the players value for that play. Be it fair or not, it is what it is.

That doesn't tell us how good of a defender the player could be on a different team. It just tells us how good he is on his current team compared to his positional peers.

Again I go back to how good Albert Pujols would hit if he got the exact same pitches Juan Pierre gets. We will never know because it won't happen. But we don't suggest Albert is better than he actually is because he doesn't get to see those same pitches. We suggest he is as good as he is based on the pitches he does get and what he does with those pitches.

lollipopcurve
06-17-2010, 02:11 PM
Regardless of where the guy started at, the ball went into a zone where most guys make the play on. Be it the managers fault for the position or the players fault, most other guys make that play. That goes against the players value for that play. Be it fair or not, it is what it is.

The key is, in the example you use, who didn't make that play (and plays of similar difficulty)? Conversely, who did make plays that most others didn't? And, in both cases, why (looking at precise data collected about the players)?

dougdirt
06-17-2010, 02:56 PM
The key is, in the example you use, who didn't make that play (and plays of similar difficulty)? Conversely, who did make plays that most others didn't? And, in both cases, why (looking at precise data collected about the players)?

UZR tells us who did make the play and who didn't. The why can't be answered by it. For that, we need scouting reports.

lollipopcurve
06-17-2010, 04:01 PM
UZR tells us who did make the play and who didn't. The why can't be answered by it. For that, we need scouting reports.

No, it doesn't, not anywhere close to the degree I'm talking about. And that's because UZR does not account for where the player was positioned.

Scouts can't see with the precision the next wave of data will have.

Brutus
06-17-2010, 04:01 PM
But here is where I am at with this....

Regardless of where the guy started at, the ball went into a zone where most guys make the play on. Be it the managers fault for the position or the players fault, most other guys make that play. That goes against the players value for that play. Be it fair or not, it is what it is.



How so? If a manager positions a left fielder 15 feet toward the left-center gap, then a ball is hit down the left field line in a zone that ordinarily is caught 60% of the time, if the ball drops didn't the positioning wind up causing the player's UZR to suffer? After all, that is a play, in that case, that he otherwise had a great chance of making but by being shifted over, he could not make a play on the ball in a zone he was supposed to make.

I like UZR, as I'm sure most here know. But the positioning of fielders by teams are a major flaw as the zones don't take that into account. I do believe that positioning evens out and the advanced scouting is good enough not to hurt fielders much, but it's a valid flaw in the system.

dougdirt
06-17-2010, 04:36 PM
How so? If a manager positions a left fielder 15 feet toward the left-center gap, then a ball is hit down the left field line in a zone that ordinarily is caught 60% of the time, if the ball drops didn't the positioning wind up causing the player's UZR to suffer? After all, that is a play, in that case, that he otherwise had a great chance of making but by being shifted over, he could not make a play on the ball in a zone he was supposed to make.
For the same reason we still ding a player for hitting a weak ball on a hit and run called by the manager.... because that is what happened. What happened doesn't always reflect ones talent, just ones production.



I like UZR, as I'm sure most here know. But the positioning of fielders by teams are a major flaw as the zones don't take that into account. I do believe that positioning evens out and the advanced scouting is good enough not to hurt fielders much, but it's a valid flaw in the system.
It is only a flaw if we are trying to measure skill rather than production.

dougdirt
06-17-2010, 04:38 PM
No, it doesn't, not anywhere close to the degree I'm talking about. And that's because UZR does not account for where the player was positioned.

Scouts can't see with the precision the next wave of data will have.

Production doesn't care where a guy was positioned. It cares whether he made the out. Whether it is his fault or not, he either made the play or didn't. That is what matters in terms of production.

lollipopcurve
06-17-2010, 04:48 PM
Production doesn't care where a guy was positioned. It cares whether he made the out. Whether it is his fault or not, he either made the play or didn't. That is what matters in terms of production.

I don't need UZR to tell me if a guy catches a ball or not.

What you seemingly have trouble understanding is that you are arguing the merits of a framework (UZR) that is designed to allow us to compare fielders. Yet, even though I am advocating collecting data that would allow us to compare fielders with greater precision -- and therefore to make an even better comparison -- you dismiss it as superfluous. Can't you see that what I am talking about is an extension/improvement on what UZR has accomplished?

dougdirt
06-17-2010, 04:54 PM
I don't need UZR to tell me if a guy catches a ball or not.

What you seemingly have trouble understanding is that you are arguing the merits of a framework (UZR) that is designed to allow us to compare fielders. Yet, even though I am advocating collecting data that would allow us to compare fielders with greater precision -- and therefore to make an even better comparison -- you dismiss it as superfluous. Can't you see that what I am talking about is an extension/improvement on what UZR has accomplished?

I see that you want more information than what UZR offers. I am on board for that. Where we continue to disagree at is the comparisons you want to make and how valuable they are towards defensive values of what happened.

As for you don't need UZR to tell you if someone caught the ball or not... well no, but you do likely need it to tell you if someone caught it in a place where others also catch the ball.

lollipopcurve
06-17-2010, 05:12 PM
As for you don't need UZR to tell you if someone caught the ball or not... well no, but you do likely need it to tell you if someone caught it in a place where others also catch the ball.

What does the end location of the play tell you about 1 player vs. another?

dougdirt
06-17-2010, 05:38 PM
What does the end location of the play tell you about 1 player vs. another?

Who made outs in certain parts of the field.

Brutus
06-17-2010, 07:04 PM
For the same reason we still ding a player for hitting a weak ball on a hit and run called by the manager.... because that is what happened. What happened doesn't always reflect ones talent, just ones production.


It is only a flaw if we are trying to measure skill rather than production.

That seems like a runaround from saying the reason we measure production is to get an idea of skill. Production is supposed to be reflective of talent when given a large enough sample. I agree with you that if we're truly only looking at UZR as a measure of what happened, then fine - no big deal.

Problem is, you and I both know that is absolutely not the case.

If it's only a measure of what happened, why is it a part of a formula to measure players' worth? If they are a product of their positioning, at least in part, isn't it silly to use UZR as part of the WAR formula, thereby measuring their value? I don't agree that UZR is only used by the masses as a measure of "what happened" anymore than we use offensive statistics to only measure "what happened." We use them, as much as we can, to determine a pattern that reflects skill.

dougdirt
06-17-2010, 07:21 PM
If it's only a measure of what happened, why is it a part of a formula to measure players' worth? If they are a product of their positioning, at least in part, isn't it silly to use UZR as part of the WAR formula, thereby measuring their value? I don't agree that UZR is only used by the masses as a measure of "what happened" anymore than we use offensive statistics to only measure "what happened." We use them, as much as we can, to determine a pattern that reflects skill.
For the same reason that a player in a down season because of a poor BABIP has his 'worth' valued at what he actually did on the field regardless of how his true skill is actually 'worth'.

And while I do agree that a lot of the time we can use stats to determine skill, that isn't always the case. UZR when used in a WAR format is about what was produced on the field.

kaldaniels
06-17-2010, 09:42 PM
So what's the point of UZR when you boil it down. I see it determines what a player WAS worth...but isn't the point of metrics to help predict future results, or is it not. It seems it is dandy to use to help determine the 2009 gold glove winner, but I want something to help decide what player to sign in free agency for 2011.

jojo
06-17-2010, 09:59 PM
So what's the point of UZR when you boil it down. I see it determines what a player WAS worth...but isn't the point of metrics to help predict future results, or is it not. It seems it is dandy to use to help determine the 2009 gold glove winner, but I want something to help decide what player to sign in free agency for 2011.

In order to project a player's performance you have to first estimate his true skill. In order to do that you need to be able to measure his performance and then regress it back to a population made up of similarly skilled players. UZR represents the best method available for measuring defensive performance.

All metrics measure what happened and can also be used to inform what might happen in the future.

TheNext44
06-17-2010, 10:27 PM
In order to project a player's performance you have to first estimate his true skill. In order to do that you need to be able to measure his performance and then regress it back to a population made up of similarly skilled players. UZR represents the best method available for measuring defensive performance.

All metrics measure what happened and can also be used to inform what might happen in the future.

Very well put.

And when you get down to it, the defensive metrics currently being used are not that much better or worse than the offensive metrics being used to determine a player's worth going forward. They both share similar limitations.

dougdirt
06-18-2010, 10:24 AM
So what's the point of UZR when you boil it down. I see it determines what a player WAS worth...but isn't the point of metrics to help predict future results, or is it not. It seems it is dandy to use to help determine the 2009 gold glove winner, but I want something to help decide what player to sign in free agency for 2011.

If you want something to help decide what player to sign in free agency in 2010, talk to your scouts. If they have a large sample of UZR, look at it and apply an age factor in it (if its a free agent, project the defense to decline each and every year because guys aren't free agents at age 25 and that is when defense starts to go the other direction for most players).

nate
06-24-2010, 12:34 PM
I had a thought about this whole "if his defense led to a hit but didn't give up a run, who cares?" thing. That thought is this: offensively, we measure many things that don't directly lead to runs. In fact, the only statistic that directly leads to a run is HR. Indirectly, players fortunate enough to hit safely with men on base as well as those men on base can share in the run creation process. However, stats like BA, OBP, SLG, SB and other "base-acquisition with no scoring" offensive metrics are largely moot if viewed in the same light as "he didn't directly allow any runs to score so it doesn't matter."

To me, the whole problem with traditional defensive metrics is you can't "see" the day to day progression of the core. The outliers like dreadful errors and exciting leaping catches, yes. Not getting to a ball on the fringe of one's range, no.

dougdirt
06-24-2010, 06:04 PM
My favorite way of evaluating defense is a way that Justin, now at RedReporter.com (http://redreporter.com) came up with a few years ago.

What he did was take every play per position in its Revised Zone Rating and figure out the 'average' zone rating for that position. Then to apply that number to each players 'Balls in Zone' to figure out the +/- plays a guy has made. Then to look at the out of zone plays for that entire position and divide it by innings played to get the average number of out of zone plays per inning and then place that against the innings for each player to figure out a +/- for a player in their out of zone plays, then to add them up for each player. I went ahead and did this for right field for this season so far. Here are the results for all 121 players who have played RF in the majors this season, where IZVAL is the in zone value compared to average, OOZVAL is the out of zone value compared to average and the TotVAL is those two combined. All are measured in +/- plays made.



Name IZVAL OOZVAL TotVAL
Ichiro Suzuki 5.90 11.81 17.71
Ben Zobrist 0.79 14.05 14.83
Ryan Ludwick 4.23 7.77 12.00
Jay Bruce 3.79 7.64 11.43
Cody Ross 3.90 5.73 9.62
Nelson Cruz 1.89 6.89 8.78
Garrett Jones -1.99 9.67 7.68
Jason Heyward 5.90 0.83 6.73
N Schierholtz 4.67 1.52 6.19
Hunter Pence 5.90 -0.33 5.57
Gabe Kapler -1.22 5.91 4.69
Michael Stanton 1.22 3.33 4.55
Will Venable 1.79 2.62 4.41
Tyler Colvin 1.22 2.58 3.81
David Murphy -1.00 4.69 3.69
Willie Harris -1.00 4.56 3.56
J.D. Drew 0.23 3.14 3.38
Mike Morse 0.78 2.26 3.04
Cory Sullivan 0.11 2.69 2.80
Nick Stavinoha -0.67 3.23 2.57
Jose Bautista 0.45 2.07 2.52
Ryan Sweeney 2.57 -0.08 2.48
Dar McDonald 0.89 1.32 2.21
Corey Hart -1.43 3.51 2.08
Jack Cust -0.11 2.14 2.03
Jose Guillen 0.45 1.52 1.97
Joe Inglett 0.44 1.29 1.73
Chris Dickerson 0.11 1.60 1.71
Jerry Hairston 0.67 0.60 1.27
Jon Van Every 0.11 1.09 1.20
Reggie Willits 0.33 0.86 1.19
Brett Carroll 1.67 -0.53 1.14
Ross Gload 0.67 0.40 1.07
Wil Bloomquist 0.78 0.29 1.07
Jason Kubel -1.22 2.21 0.99
Ryan Langerhans -0.44 1.32 0.87
Justin Upton 7.01 -6.16 0.85
Jim Edmonds 0.67 0.18 0.85
Cristian Guzman -0.33 1.17 0.84
Jayson Nix 0.22 0.60 0.82
Jesus Feliciano 0.33 0.49 0.82
Chris Heisey 0.33 0.46 0.79
Chris Denorfia 1.67 -0.88 0.79
Aar Cunningham 0.22 0.52 0.74
Greg Golson 0.22 0.49 0.71
Justin Maxwell -0.66 1.32 0.66
Chad Huffman 0.22 0.37 0.60
Gabe Gross 1.56 -0.96 0.59
Melky Cabrera -0.67 1.12 0.46
Garret Anderson 0.44 -0.05 0.39
Allen Craig 0.44 -0.14 0.30
Josh Reddick -0.67 0.86 0.19
John Bowker 0.56 -0.36 0.19
Brennan Boesch 0.56 -0.44 0.11
Kosuke Fukudome 3.01 -2.96 0.05
Don Kelly 0.00 0.00 0.00
Jeff Baker 0.33 -0.34 -0.01
Adam Stern 0.00 -0.06 -0.06
John Raynor 0.11 -0.17 -0.06
Reed Johnson 1.00 -1.11 -0.11
Reid Brignac 0.00 -0.11 -0.11
Ramiro Pena 0.11 -0.23 -0.12
Shelley Duncan 0.33 -0.46 -0.12
Eugenio Velez 0.00 -0.17 -0.17
Robb Quinlan 0.11 -0.29 -0.17
Cole Gillespie -0.89 0.71 -0.17
Jeff Francoeur -5.21 5.01 -0.20
Andres Torres -2.44 2.19 -0.25
Hank Blalock 0.00 -0.29 -0.29
Mike McCoy 0.00 -0.29 -0.29
Chris Carter 0.33 -0.63 -0.29
Jeremy Reed -0.11 -0.20 -0.31
Craig Gentry 0.00 -0.34 -0.34
Greg Dobbs 0.00 -0.34 -0.34
Jeremy Hermida 0.22 -0.57 -0.35
Matt Carson 0.00 -0.40 -0.40
Casper Wells 0.00 -0.40 -0.40
Joe Mather 0.00 -0.46 -0.46
Last Milledge -2.67 2.21 -0.46
Jayson Werth -0.21 -0.31 -0.52
Lou Montanez 0.00 -0.57 -0.57
Jody Gerut 0.33 -0.91 -0.58
Xavier Paul 0.56 -1.22 -0.66
Randy Winn 1.33 -2.14 -0.80
G Matthews Jr. -0.67 -0.25 -0.92
Juan Rivera -0.22 -0.74 -0.96
Nolan Reimold -0.89 -0.11 -1.00
Gerardo Parra 0.33 -1.34 -1.00
Vlad Guerrero -0.33 -0.68 -1.01
Mitch Maier -0.56 -0.51 -1.07
Jason Michaels -0.67 -0.51 -1.18
Fred Lewis 0.22 -1.45 -1.23
Bryan Petersen -0.78 -0.46 -1.23
Ben Francisco -1.67 0.35 -1.32
Oscar Salazar 0.11 -1.45 -1.34
Aubrey Huff -0.44 -1.05 -1.49
Andruw Jones 1.00 -2.78 -1.78
Ryan Raburn -1.56 -0.31 -1.87
Bill Hall -1.44 -0.43 -1.87
Scott Hairston -2.33 0.46 -1.87
Willy Taveras 0.33 -2.25 -1.92
Nick Swisher -3.10 1.03 -2.07
Ryan Spilborghs -0.11 -1.96 -2.07
Jonathan Jay -0.33 -1.94 -2.27
Ryan Church -3.33 1.02 -2.32
Marcus Thames -0.78 -1.83 -2.60
Mark Kotsay -3.00 -0.17 -3.17
David DeJesus -1.55 -2.03 -3.58
Roger Bernadina -2.11 -1.79 -3.90
Shin-Soo Choo -1.55 -2.62 -4.17
Xavier Nady -2.55 -1.87 -4.42
Carlos Gonzalez 0.45 -5.07 -4.63
Delwyn Young -2.55 -2.08 -4.63
Travis Snider -3.78 -1.69 -5.47
Nick Markakis -2.77 -7.68 -10.44
Magglio Ordonez -2.66 -7.92 -10.58
Andre Ethier -0.44 -11.89 -12.33
Bobby Abreu -2.43 -12.08 -14.51
Carlos Quentin -4.77 -10.30 -15.07
Michael Cuddyer -0.77 -15.54 -16.31
Brad Hawpe -3.10 -14.01 -17.11


The difference between the best (Ichiro) and the worst (Hawpe) is 34 plays, or about 3.5 plays per week. That passes the smell test for me. All of the math makes sense. The way that the plays are recorded, while not perfect, is pretty good to me as well. Just thought I would add this to the conversation.

Scrap Irony
06-24-2010, 06:20 PM
This, too, inflates some "scores" and lessens others as it relates to those plays that could be made by either a CF or RF and the CF takes the ball.

Let's say Drew Stubbs happens to have a ton of range (hypothetically) and Dusty Baker tells him to catch any ball he can get to. Let's also say Ryan Ludwick calls off CF Colby Rasmus on those same plays because Rasmus is a step slower as to range.

Ludwick would get credit and Bruce would not.

Is that a fair representation of their ability?

dougdirt
06-24-2010, 06:28 PM
This, too, inflates some "scores" and lessens others as it relates to those plays that could be made by either a CF or RF and the CF takes the ball.

Let's say Drew Stubbs happens to have a ton of range (hypothetically) and Dusty Baker tells him to catch any ball he can get to. Let's also say Ryan Ludwick calls off CF Colby Rasmus on those same plays because Rasmus is a step slower as to range.

Ludwick would get credit and Bruce would not.

Is that a fair representation of their ability?

It isn't about ability. Its about what was made. Do we make notations and adjustments to offensive value when the manager calls for the hit and run causing the batter to swing at crappy pitches and thus make weak contact? Do we take away the hits on the hit and run that would have been a routine grounder to 2B/SS that now is a hit because they went to cover the bag?

How often does Drew Stubbs make a legit catch in a left fielders zone?

Scrap Irony
06-24-2010, 06:30 PM
Often, I'm guessing.

nate
06-24-2010, 06:31 PM
This, too, inflates some "scores" and lessens others as it relates to those plays that could be made by either a CF or RF and the CF takes the ball.

Let's say Drew Stubbs happens to have a ton of range (hypothetically) and Dusty Baker tells him to catch any ball he can get to. Let's also say Ryan Ludwick calls off CF Colby Rasmus on those same plays because Rasmus is a step slower as to range.

Ludwick would get credit and Bruce would not.

Is that a fair representation of their ability?

Yes. Plays that both fielders could get to are likely to be marginal.

And...

Not all defense "happens" between two defenders.

dougdirt
06-24-2010, 06:35 PM
Often, I'm guessing.

You think its often that Drew Stubbs catches a ball in left field? As someone who has watched 90% of the games he has started, I have to ask.... what? The guy has range. But no one has the range to make catches at an entirely different position on the field, much less to do it often. Its awesome when a center fielder makes a play in the gaps. It would be downright jaw dropping if one were to make plays in left or right field.

Scrap Irony
06-24-2010, 06:39 PM
Sigh.

Obviously, I meant in the LFers "area" of the field.

Stubbs takes many of those plays away from Gomes and from Bruce. This costs both Bruce and Gomes UZR OOZ "points". (All other things being equal, just three such plays would make Bruce the best RF in baseball and Gomes adequate, if my math is right.)

dougdirt
06-24-2010, 06:43 PM
Sigh.

Obviously, I meant in the LFers "area" of the field.

Stubbs takes many of those plays away from Gomes and from Bruce. This costs both Bruce and Gomes UZR OOZ "points". (All other things being equal, just three such plays would make Bruce the best RF in baseball and Gomes adequate, if my math is right.)

I don't think Stubbs is taking those plays away from Bruce at all. And 3 such plays for Gomes would still make Gomes the 2nd worst left fielder in baseball, a fraction of a play ahead of the guy in last. Given the innings, Gomes would need to be at 24 plays out of zone to be average. He is at 11. Not sure where adequate lies between terrible and average though, but either way, he has next to no range no matter how its sliced.

mth123
06-24-2010, 06:45 PM
Yes. Plays that both fielders could get to are likely to be marginal.

And...

Not all defense "happens" between two defenders.

But it probably happens at least two or three time per week between say the LF and CF. When a guy is only getting 3 or 4 plays per game, making one more play per week would probably move the numbers in a fairly significant way and if they did, many would view all those other plays that aren't between the fielders a lot differently.

westofyou
06-24-2010, 06:48 PM
Sigh.

Obviously, I meant in the LFers "area" of the field.

Stubbs takes many of those plays away from Gomes and from Bruce. This costs both Bruce and Gomes UZR OOZ "points". (All other things being equal, just three such plays would make Bruce the best RF in baseball and Gomes adequate, if my math is right.)

The Kelly Leake effect

dougdirt
06-24-2010, 07:04 PM
But it probably happens at least two or three time per week between say the LF and CF. When a guy is only getting 3 or 4 plays per game, making one more play per week would probably move the numbers in a fairly significant way and if they did, many would view all those other plays that aren't between the fielders a lot differently.

In about 50% of the time that Gomes has been -11 plays out of the zone, the rest of the Reds left fielders are -1 plays out of zone. Stubbs doesn't seem to be taking away the out of zone play from them.

Scrap Irony
06-24-2010, 07:04 PM
Heh. Does that make Gomes Abdullah?

(He was my favorite.)

jojo
06-24-2010, 09:11 PM
Is the argument really that Stubbs makes Gomes' defensive numbers look bad? Really, its Stubbs that limits Gomes' OOZ plays?

RedsManRick
06-25-2010, 01:11 AM
Given how many OOZ plays Bruce has (only Ichiro has more), doesn't that really throw some water on the "Stubbs is stealing his OOZ plays" argument? It seems there most be plenty of balls in the gaps for them both.

Scrap Irony
06-25-2010, 01:42 AM
Does UZR tell us where each out is made? In what quadrant, I mean. Can we see where Stubbs (or any OF, for that matter) makes each catch on the field?

I'm guessing that's info you'd have to pay BIS for, right? Because I'd love to see that study.

And I'd love to see how that affects both Dewan's and UZR as it relates to OOZ plays, and, in turn, overall WAR.

dougdirt
06-25-2010, 02:36 AM
Does UZR tell us where each out is made? In what quadrant, I mean. Can we see where Stubbs (or any OF, for that matter) makes each catch on the field?

I'm guessing that's info you'd have to pay BIS for, right? Because I'd love to see that study.

And I'd love to see how that affects both Dewan's and UZR as it relates to OOZ plays, and, in turn, overall WAR.

Yes you can see that, but you are going to have to pay for it. Now what you could do for free and a little bit of skill (and a TON of time) would be to look up and chart the plays as charted by Gameday. While its not perfect, it would at least be a rough idea.

Scrap Irony
06-25-2010, 02:52 AM
We've all seen these types of hits. They're a ton of them, in fact. Balls skied between OFs where either one could make the play. I've seen a ton of those (too many, in fact) just in the West Coast swing the Reds took.

This is, to me, a major problem with UZR. Others insist it's insignificant.

How insgnificant is it, do you think?

It wouldn't affect "run" determination, as the out has been made. It wouldn't affect the reliability of the numbers either, or their validity as they relate to runs scored.

It could, however, affect quite a few player UZR and WAR values by quite a bit.

dougdirt
06-25-2010, 03:16 AM
We've all seen these types of hits. They're a ton of them, in fact. Balls skied between OFs where either one could make the play. I've seen a ton of those (too many, in fact) just in the West Coast swing the Reds took.

This is, to me, a major problem with UZR. Others insist it's insignificant.

How insgnificant is it, do you think?

It wouldn't affect "run" determination, as the out has been made. It wouldn't affect the reliability of the numbers either, or their validity as they relate to runs scored.

It could, however, affect quite a few player UZR and WAR values by quite a bit.Why is it not affecting Jay Bruce or the other left fielders and only Jonny Gomes of the Reds outfielders? I think its very likely because the other guys all have range and Gomes simply doesn't. Bruce, Dickerson, Heisey and Nix can all run. They don't seem to have balls 'stolen' by Stubbs. I am also pretty sure that if the ball was caught by someone and turned into an out, it doesn't go against someone else. Gomes UZR would not be affected in any way by Stubbs catching a 'Gomes ball'.

For the record, Stubbs is slightly below average right now on plays made out of his zone.

Scrap Irony
06-25-2010, 03:24 AM
Who said I'm talking about Gomes? I'm questioning the statistic and its effectiveness as it relates to all players, not just one random guy.

But, playing devil's advocate, I could see Stubbs shading to the left as per orders from the bench. ("Bruce has better range than Gomes," says Dusty. "Shade to left." Stubbs does so. With Heisey, Dickerson, or Nix, Dusty says to Stubbs, "Play straight up.") That shading to left could have a significant effect on Gomes' OOZ, n'est-ce pa?

Scrap Irony
06-25-2010, 03:29 AM
Why is it not affecting Jay Bruce or the other left fielders and only Jonny Gomes of the Reds outfielders? I think its very likely because the other guys all have range and Gomes simply doesn't. Bruce, Dickerson, Heisey and Nix can all run. They don't seem to have balls 'stolen' by Stubbs. I am also pretty sure that if the ball was caught by someone and turned into an out, it doesn't go against someone else. Gomes UZR would not be affected in any way by Stubbs catching a 'Gomes ball'.

For the record, Stubbs is slightly below average right now on plays made out of his zone.

I'm not talking about "stealing" balls or "poaching" them. Well, I guess I am. A little. I'm talking about those lazy fly balls that are hit between OFs (OOZ for both of them) that could be caught by either player. UZR gives extra "points" for these types of catches, yes?

dougdirt
06-25-2010, 03:34 AM
Who said I'm talking about Gomes? I'm questioning the statistic and its effectiveness as it relates to all players, not just one random guy.

But, playing devil's advocate, I could see Stubbs shading to the left as per orders from the bench. ("Bruce has better range than Gomes," says Dusty. "Shade to left." Stubbs does so. With Heisey, Dickerson, or Nix, Dusty says to Stubbs, "Play straight up.") That shading to left could have a significant effect on Gomes' OOZ, nec pa?

I will play devils advocate too.... is this something that has happened to Gomes throughout his career? He has never had range to make out of zone plays. He also tends to be below average on plays made in his zone, which certainly aren't going to be made by another fielder.

Now, to the point that someone could shade over and have an effect on Gomes out of zone plays.... yeah, it could happen some. It isn't likely for Gomes because he doesn't have the range to make the plays anyways which we can tell by his atrocious in zone plays made rate compared to his counterparts (both this year and in his career).

The guys who are hurt on their out of zone plays are the guys who don't have range, not the guys with the selfish fielders next to them.

For Jonny Gomes to be average on his out of zone plays, he would need to have 24 of them. Right now he has 11. Given that Stubbs is also currently below average (by 1 play), I just can't buy into Stubbs stealing Gomes out of zone plays from him. An out of zone play for Gomes is an out of zone play for Stubbs as well (we know that Gomes surely isn't running down balls that center fielders are supposed to be catching in their zones). If Stubbs were stealing a large amount of plays from Gomes out of zone area by shading as you say, then his true range would be well below average.

dougdirt
06-25-2010, 03:35 AM
I'm not talking about "stealing" balls or "poaching" them. Well, I guess I am. A little. I'm talking about those lazy fly balls that are hit between OFs (OOZ for both of them) that could be caught by either player. UZR gives extra "points" for these types of catches, yes?

No, they don't. Those plays are expected to be made and aren't awarded much at all.

Scrap Irony
06-25-2010, 04:22 AM
No, they don't. Those plays are expected to be made and aren't awarded much at all.

But they are given points for it.

This, then, is a problem with the statistic, is it not? It could very well be a meaningful problem with the statistic, couldn't it?

I looked it up on Fangraphs (thanks, texasdave). They said:


Of course, with this method, a ball hogger will get slightly more credit than he deserves, but as long as his ball-hogging is done on easy fly balls, he isnít going to get much credit anyway. For example, if a certain type of ball and location is caught 90% of the time, whoever catches it is only going to receive .1 (1 minus .9) times .83, or .083 runs in credit.

Now, .083 doesn't sound like much until you add up how many there could be. Let's say there's three a week. (That's a WAG there; I truly don't know how many there are.) And the CF catches them all. (Likely, if both CF and LF are used to their roles.)

There are 27 weeks in the baseball season (give or take), from April through September. That's more than six and a half UZR credits right there.

I'd call that pretty serious, especially considering the average LF gets between 300 and 400 balls per season.

What does all this prove?

Nothing.

I know Gomes isn't a great fielder. No one on this board thinks that. (Or, if someone does think it, they've kept it quiet.) Last season, UZR said that Gomes was a pretty good LF (1.1 in limited action). This season, he's among the bottom dwellers. In his career as a LF (138 games), he's at - 18.2.

FanGraphs says to take as much data as you can. If there's less than three full years of data available (paraphrasing here), the true number is between the two, moving more toward zero than not.

I'm comfortable, in other words, that he's a bad defender. I get that. (That passes the smell test.)

What makes me uncomfortable is the amount of run UZR says he is (and, by extension, other players are) worth. The subjectivity of the data at the beginning (covered elsewhere) and the amount of padding available to ball hoggers clouds UZR to the point that it's a nice tool to see, but I wouldn't depend on it to help tell me who's valuable and who isn't.

Again, I liken UZR to batting average. (Two statistics that are similar in their correlation to runs, if other posters are correct.) You wouldn't want value to be judged by batting average, would you?