I have a bookmark for the Fangraphs website.

Well, if you bookmark it, you must understand and appreciate baseball metrics. And speaking of fangraphs, they respond to Kepner although not specifically to this blog post.

http://www.fangraphs.com/blogs/index.php/seeing-and-uzr-and-teixeira

I'd be willing to be that some teams have systems that make UZR look primitive and do in fact put much credence in their own proprietary systems.

While I realize that this thread will be about defensive statistics and their efficacy, can I ask how it is that Teixeira has a negative UZR?

While I realize that this thread will be about defensive statistics and their efficacy, can I ask how it is that Teixeira has a negative UZR?

Remember that UZR doesn't include a lot of 1B defense. It says nothing about handling throws, and, IIRC, doesn't deal with pop-ups well.

A 1B can have a negative UZR and still be a plus fielder.

While I realize that this thread will be about defensive statistics and their efficacy, can I ask how it is that Teixeira has a negative UZR?


Sample size, for one. You need a larger sample size for defensive numbers. 4 months of defensive data is like a month or two of offense data and one can easily see the fluctuations there. Players are so close (thanks to being pre-sorted to positions) to each other that a few extra hard-hit balls or bad bounces can change things quite a bit in 2/3 of a season of defensive numbers.

While people will think of a "30 homer guy" as someone who will probably hit 20-40 home runs or so in a year, people also need to get used to a "+5 defense guy" sometimes being 0 and sometimes being +10. But when a 30-homer guy "only" hits 20 home runs, those missing 10 homers do hurt the bottom line, even if the player's better than that.

Alex Rodriguez sometimes hits 35 instead of 45 home runs and Mark Teixeira sometimes is sometimes a run below average instead of 5 above. I'm not sure what the big deal is. It's silly to think a +5 fielder is going to +5 every season.

The same way Yunel Escobar or Torii Hunter can have a negative UZR and still be a good SS. Sometimes the level of defensive play around the league is raised or lowered to such a degree that a person who steadily provides the same production wise on defense sees his numbers fluctuate.

I'm very much a novice when it comes to the defense metrics, but is there any sort of consensus as to what can be thought of as a good sample size? Do people generally wait for 150-game samples before drawing conclusions?

I'm very much a novice when it comes to the defense metrics, but is there any sort of consensus as to what can be thought of as a good sample size? Do people generally wait for 150-game samples before drawing conclusions?

A quick rule of thumb is about a 3:1 ratio - consider 3 years of defensive data worth a single year of offensive data. A +5 fielder can easily be -1 over the course of a single season.

I'm very much a novice when it comes to the defense metrics, but is there any sort of consensus as to what can be thought of as a good sample size?

I thought you needed at least two years of data before you can draw a confident conclusion about a player's defensive ability.

Do people generally wait for 150-game samples before drawing conclusions?

Some people find even that insufficient.

MGL who is something of a standard-setter in creating defensive metrics, has also discussed the problems with defensive metrics. One is that, even for regular players, the sample sizes are a lot smaller than for offensive metrics. The other is observational error -- what gets recorded as a catchable chance may not, in fact, be one. In less than a season, given already small sample sizes and fluctuations, it would not take much in the way of observational error to skew the numbers.

Then you have the problem that defensive ability is no more static than offensive ability. As you gain knowledge, your reflexes and speed may go down. An injury that slows you or affects your reflexes can have a big impact.

At the end of the day, the only thing the defensive metrics can tell you is whether, in that year, within a larger margin of error than for offensive statistics, you appeared to make more or less plays than others at your position. A consistent pattern over multiple years, confirmed by multiple metrics, is almost certainly significant. When the metrics disagree, and/or tend to fluctuate, then you have to take them with an even larger does of salt.

#8 & 9: thanks, makes sense to me

as to Teixeira, by my eyes his play this year has been above average, but not exceptional. His range doesn't wow you, but he certainly looks much better when put against Giambi.

And I don’t think I’ve met anyone in uniform who gives credibility to defensive statistics.

Sure you have. Unfortunately, there are tons of people in the game that give credibility to fielding percentage.

As for UZR, as MGL Will tell you it says little in a small sample. Tex is usually pretty good in UZR. This year he rates at -0.8, but last year was +10.6. He's likely a good to very good fielder and has likely been a good fielder this year as well, but for whatever reason hasn't had a great UZR so far. Could be measurement error, or small sample skewed by plays the system thinks were easy were actually hard, blah blah. We've been over this a million billion times.

But here, we're not trying to determine a player's defensive ability. We're trying to determine a player's 2009 defensive performance. We don't need to wait until 2011 to know how Mark Teixeira actually performed in 2009.

I'm certain that (a) what the defensive stats tell us about Teixeira in 2009 aren't necessarily going to line up with what they've said in prior years, and (b) UZR doesn't capture everything a 1B does on defense. What that means regarding his 2009 performance, I don't know. It certainly means you shouldn't extrapolate from his 2009 numbers to judge his ability. But that's not what people are trying to do.

But here, we're not trying to determine a player's defensive ability. We're trying to determine a player's 2009 defensive performance.

Right, but the UZR for 2009 doesn't necessarily tell you about his defensive performance, for the reasons outlined above. Best we can do is infer his 2009 performance, based on his recent UZR (including 2009 UZR) and his reputation. Given that:

- his 2008 UZR was so high.
- his reputation is very good.
- his 2009 UZR is basically average.

I am comfortable guessing that his actual defensive value this year is pretty good. Probably in the +5 zipcode(*) if I had to give a number to it.

As has been written before, when it says his UZR is -1 or whatever, it's really saying: "His defensive value this year is probably between -6 and +4" or some such.

* hehehe.

as to Teixeira, by my eyes his play this year has been above average, but not exceptional. His range doesn't wow you, but he certainly looks much better when put against Giambi.

And that's probably an additional reason Yankee watchers are having trouble reconciling their opinions of Teixeira with the numbers. They're comparing him to the awkward-looking Giambi, which may skew their view of Teix's contributions or how he truly rates against others at the position.

So basically, the author of TFA both does and doesn't have a point. His good point is that a few months' worth of UZR should be taken with a shovelful of salt. His bad point is that since you need the salt, and since he once saw so-and-so make a good throw, all statistical efforts to measure defense are worthless.

Right.

It reminds me of a column I read on ESPN some years ago. The author was talking about goaltending statistics in the NHL. His points were:

1) GAA (goals against average,) is flawed because it doesn't take into account the number of shots the goalie faced or the quality of his defense. He is 100% correct and I nodded as I read along.

2) Save% is flawed because it doesn't take into account the quality of shots the goaltender faces (eg the quality of his defense.) Again he is 100% correct and I nodded as I read along, anticipating his conclusion.

His conclusion? Throw both out the window and judge the goalie based on wins and losses. Now there's a does-not-compute...

But I digress. Obviously UZR is flawed. Every metric is flawed. But just because the best tool we have isn't perfect, doesn't mean we should use some other tools that are even less perfect, simply because they've been around longer. A lot of people are of the opinion that we should not switch from one tool to the other unless the new tool is flawless. I don't understand that logic.

But here, we're not trying to determine a player's defensive ability. We're trying to determine a player's 2009 defensive performance. We don't need to wait until 2011 to know how Mark Teixeira actually performed in 2009.

And people seem to have a problem recognizing that good defensive players can underperform their talent in a limited number of chances, even though we know that players go through similar things on offense. It is very possible that both the observers and UZR are right - Teixeira may well be a talented defensive player who simply didn't get to as many balls as his peers did over a short period of time.

This year he rates at -0.8, but last year was +10.6.

Those decimal points are part of the problem.

Hey #17. Your handle screws up the mainpage when you are on hot topics. Make that stop.

Those decimal points are part of the problem.

You mean in the sense that it presents a level of precision that isn't there?

[20] It doesn't for me, but sure.

It didn't for me either Shock. I did notice if you had the last post that your handle name ran off the page, but it's never impacted the page width for me.

Hey #17. Your handle screws up the mainpage when you are on hot topics. Make that stop.

You should change it to 53686F636B.

And I shouldn't know that.

And that's probably an additional reason Yankee watchers are having trouble reconciling their opinions of Teixeira with the numbers. They're comparing him to the awkward-looking Giambi, which may skew their view of Teix's contributions or how he truly rates against others at the position.

Is Coco Crisp the reason many Sox fans don't accept Ellsbury's -11 UZR this year (or -6 per 150 for his career)? A lot of these numbers are screwy compared to what one sees on the field on lots of teams and I think it has little to do with any team's previous players.

well, its makes the hot topics sidebar very wide in IE. though I think i am using an old version (at work). Is there one after 7?

#21 Yup.

When the standard error of your metric is ~7 runs it annoys me to see it presented in the #.# format.

Of course the same objection applies to any offensive stat.

Is Coco Crisp the reason many Sox fans don't accept Ellsbury's -11 UZR this year (or -6 per 150 for his career)?

I doubt it. Ellsbury's been playing CF in Boston for parts of three seasons. But you'd probably find people in Boston who would insist that Bay is better than his numbers for this reason.

More important, I didn't say it was the only reason, just a possible factor. Teixeira is following a guy that by most accounts is a legitimately bad defender. Even if Teix struggled by his standards, he'd likely still look considerably better than the guy he replaced, possibly skewing the perception of his performance.

Is that really a controversial position CP?

But you'd probably find people in Boston who would insist that Bay is better than his numbers for this reason.

No. Bay is terrible. He can run, and he tries hard, but he's always getting very late jumps and taking bad angles to balls, and he doesn't throw all that well. He's really only better than Manny because he gives a better effort.

Of course the same objection applies to any offensive stat.

Joe Mauer leads the AL with a 0 batting average.

Obviously an objective measure isn't useful if the only time you ever give it credence is when it confirms your subjective opinion.

I think at the very least UZR should cause people to consider the possibility that their subjectively formed opinion might be wrong. Given the nature of defensive metrics, that's about as far as you can take it. It's a part of the picture, and nothing more.

The main problem with evaluating a player's defense relative to another's has always been opportunity. How many times do you get to see the 30 starting first basemen for each team play? If you're a scout, that's one thing.

But schlubs like us just don't get to see 30 first basemen play 100 times a year. So really, we don't even have a subjective basis for opinions we form with our eyes. That's why we need the 'objective' (objective/subjective is a matter of degree and not kind) data in the first place to form any opinions backed up by evidence.

No. Bay is terrible. He can run, and he tries hard, but he's always getting very late jumps and taking bad angles to balls, and he doesn't throw all that well. He's really only better than Manny because he gives a better effort.

You're speaking for all of Beantown now, huh tjm? ;-)

Is there one after 7?

Unfortunately.

I have long contended that the skill of the first baseman at scooping throws in the dirt makes a huge difference in throwing errors, and therefor in the evaluation of the other infielders' defense. It seems to me that infielders are being taught to keep the ball way down, to the point where we see several bounced throws per game and rarely an airmailed delivery. Of course, it makes sense to be too low rather than too high, since the bounced throw can most often be scooped, but the ball winged over the 1B's head (or far to the left or right) has no hope. However, any bounced throw that's missed, no matter how easy, is still and E-6 rather than an E-3.

Right, but the UZR for 2009 doesn't necessarily tell you about his defensive performance, for the reasons outlined above.

The reasons outlined above are:

1. Not everything a 1B does is captured in UZR.

2. Sample size produces large fluctuations. ("4 months of defensive data is like a month or two of offense data and one can easily see the fluctuations there.")

3. It's not that Teixeira got worse; rather the rest of the league might have gotten better.

4. Observational error.

I concede 1, but Teixeira's prior years are subjected to the same issue. His drop in UZR this year is a reflection of differences between 2008 and 2009, and UZR didn't account for this stuff in either year. I suppose one could argue that he's doing even better than usual in all the ways UZR doesn't measure, but I haven't seen anyone make that argument.

On 2, as Szym notes there's a lot of fluctuation between hitting stats over a month and hitting stats in the long haul. But if a hitter has a .120 average in a month, nobody cites "flaws with batting average" as the reason. If you're measuring that hitter's performance, .120 is an accurate representation of his results. Maybe he hit a lot of "right at 'em" shots, or maybe he was up against abnormally good pitching/defense that month. And maybe it'll be the only month under .280 in his career. But a .120 average is a .120 average; he didn't really hit .280 that month. In short, Teixeira's 2009 UZR could be authentically low.

Regarding 3, if the entire league except Teixeira got better at 1B defense in 2009, that works against him, no?

One could chalk up the entirety of Teixeira's low UZR this year to the effect of 4, I suppose. As you point out, based on past performance it seems reasonable that there must be some of this. It seems to me that some folks are very quick to judge the effect of #4 as being the whole story. It's likely not.

Is that really a controversial position CP?

I don't think it's controversial but I don't think it's right. The Yanks have had more than their share of bad defenders over the course of the years but I still think that anyone who watches enough games will get a good sense of what is approximately the standard of play at each position by watching the other guys on the field. Because you get to see so many other guys play the field while watching your team I don't think that being subjected to a horrible defensive player like Giambi or Soriano or Manny has much of an effect on a fans' subjective view of what passes for good and bad defensively.

Sorry, I'm cranky, between the defensive metrics conversations going on and the MVP conversation, I'm reading too many people certain of things that aren't clear and it's bugging me. I shouldn't have responded to you like that. I should probably take a day or two off from the boards.

Joe Mauer leads the AL with a 0 batting average.

Slacker. Pujols leads all of baseball in OBP (0), SLG (1) and OPS (1).

In short, Teixeira's 2009 UZR could be authentically low.

It could be, yes. But it could be flaws in the system.

With batting average, we have two very rigid components over which there is little quibble: hits and at-bats. Simple. No real way that that can get screwed up unless someone has a bizarre run of a luck where he keeps ROE'ing and the official scorer keeps giving him hits for some reason.

In UZR, the components are "actual outs" and "estimated outs that he should have had" or some such wording. The first of these components is pretty rigid. The second is most definitely not.

Suppose Tex gets a sharp liner shot at him and he has to make an extraordinary dive to catch it. Maybe UZR doesn't realize how difficult a play it was (because it doesn't capture batted ball speed, or positioning.) UZR "thinks" it's a play that is made 99% of the time and gives Tex little credit. In reality, it's a play that is made only 10% of the time and Tex deserved more credit.

Over the long haul, we expect these sorts of things to "even out." But over the small samples we have, a few plays like this can easily skew the numbers. That's the problem.

One could chalk up the entirety of Teixeira's low UZR this year to the effect of 4, I suppose. As you point out, based on past performance it seems reasonable that there must be some of this. It seems to me that some folks are very quick to judge the effect of #4 as being the whole story. It's likely not.

The problem isn't simply measurement error. It's that unlike offensive stats in which the denominator is a known quantity with defensive stats the denominator is an estimate. Thus, and MGL would say this too, UZR unequivocably does not say what a player's defensive value for a given year is.

I maintain that all of this is a false dilemma. Over the long haul UZR and your eyes should be consistent. The problem isn't with our ability to make observations the problem is with the recording of those observations. If you want to know whether Teixiera has been good this year go back and watch the condensed versions of every single Yankee game this year. If you do this I am quite certain you'll be able to come up with a good assessment of how good he is. Then since defensive numbers should be banded anyway you could give a very educated evaluation of whether he is zero, +5, or + 10. This number will be representative of Teixiera's actual value this year. The problem is that nobody is doing this.

Edit] I'll add that if you think Teixiera is + 10 defensively it should be possible to point out with your eyes exactly where this value came from. If you can't do this then you don't know that Teixiera is +10 defensively.

Joe Mauer leads the AL with a 0 batting average.

Too true. Of course what I was talking about was expressing the results of that batting average in runs. Since the standard error would be something close to 13 runs, I'd probably be annoyed if the results weren't rounded to the nearest 5 runs. (not of course that anybody would bother)

BA itself? I ... accept the 3 digit presentation even if it does give a false sense of large differences. (I know that's written poorly but I can't figure out a simple way to explain. But try listing the BA leaders to 2 decimal places. My sense is that they look more tightly bunched when displayed to 2 decimal places even though there no actual difference)

Imagine if we did not have "at-bats." Instead we merely had hits, games played and batting order. We could do a pretty good job of estimating a players at-bats (and thus, his batting average,) by creating a model of expected at-bats based on games played and batting-order, but we could no longer say that a player hit .300. Only that he probably hit somewhere between .280 and .320 (or something.) That is where we are with defensive metrics.

It seems to me that some folks are very quick to judge the effect of #4 as being the whole story.

It could be (not saying that it IS, mind you). He's in front of different scorers now, and when you are talking about a relatively small number of balls in play in the 1B vicinity, moving a handful of boundary chances from a zone further away to a zone closer, or having a more restrictive definition of what constitutes a hard-hit ball, could very easily make a big difference.

-- MWE

Can't you put an upper bound on 1B scooping ability by looking at throwing error rates of infielders? I can't believe it's that important. But maybe that's my anti-JT Snow bias showing.

just to answer myself +/- 2 runs between the best and worst 1B: First baseman scoops (fangraphs)

I don't think it's controversial but I don't think it's right. The Yanks have had more than their share of bad defenders over the course of the years but I still think that anyone who watches enough games will get a good sense of what is approximately the standard of play at each position by watching the other guys on the field. Because you get to see so many other guys play the field while watching your team I don't think that being subjected to a horrible defensive player like Giambi or Soriano or Manny has much of an effect on a fans' subjective view of what passes for good and bad defensively.

I disagree, people standards are based upon what they see the most, and what they see with their "heart" when they are dealing with the hometeam. Cardinal fans bad mouth Duncans defense and yet Holliday isn't any better, and his arm is vastly inferior(and that is saying something as Duncan had an average arm) yet the ability to get any Cardinal fan to admit this is almost impossible. Anyone outside of New York who has watched Jeter play defensively for any amount of time, can flat out admit his defense is poor, how long did it take New Yorkers to even begrudgingly admit he may be only average.(heck you still can't get a lot of them to admit that. Just watch him play for one series and you will find at least one play that he missed than an average guy gets too, rarely you'll also find a play where his arm makes up and allows him to get an out where most shortstops didn't) Tex is getting mad props, because the fans are arguing with their hearts, and he is genuinely a very good defender, and in comparison to previous first baseman, he has a lot more mobility.

Even if you believe that defensive stats with the error bars are accurate and the bugs are out and we're at the gates of Nirvana, there's still a significant issue of people gaining a feel for them, being able to use them. We know BA is bad at .220, OK at .280 and good at .340, and we even have a notion of what sort of variance to expect (catcher Mauer at .370 is historic.) But any stat measured in runs or games is not intuitive for the average fan. And most of the time any expressed confidence interval has to more do with the numbers that went in than sensitivity to any historical context. I have no feel for how much UZR moves from year to year. It's sort of like the early days of DIPS when they were publishing the lists of hits allowed by pitchers and how surprising it was that guys swung from the good end to the bad end year after year with no pattern. Is this how fielding numbers look, are they this mercurial, or do we just need a better feel for how they behave? I prefer to use defensive numbers in a list to rank and compare rather than as some absolute measure of value. I'm just not comfortable enough with their behavior or their accuracy to move beyond that yet.

But any stat measured in runs or games is not intuitive for the average fan.

that and it's also not accurate either, so you are just adding estimates on top of estimates. Fans know that if a fielder with two outs man on third makes a diving stop to get the force out at first, that is one run saved by the fielder (of course that misses how the guy got into position to score in the first place or it relys on the out situation, in a one out situation it wouldn't have saved the run) But this is what we as fans feel we know while watching the game, and when you see Albert Pujols on 5 different occasions play in and throw the guy out at the plate with no outs, telling me that it's only 1.4 runs saved(or whatever) is not intuitive and just seems wrong. It's why people make the comment Andruw Jones probably saves 100 runs a year with his glove, the situation doesn't matter, it's the result that they remember and credit to the guy(unless he made a bad play to set it up)

and anyone that trumpets their system has to acknowledge the emotional impact of what a particular play feels like and explain why that isn't accurate when describing their system, or describe how their system takes it into account.

and anyone that trumpets their system has to acknowledge the emotional impact of what a particular play feels like and explain why that isn't accurate when describing their system, or describe how their system takes it into account.

MGLs response to this would be amusing. I'm sure he'd say doing so would be like trying to teach nuclear physics to dogs. Though maybe he could come up with something more polarizingly offensive. The man has a gift.

Either way, given the nature of defensive metrics, I think convincing those people is pretty far down the line in terms of progress. If you're trumpeting the value of your own defensive metric, you have a hard enough time convincing the people who are speaking your language that it has merit, much less the kinds of folks you reference.

There's no damn way Ellsbury is a below-average fielder.

It could be, yes. But it could be flaws in the system.

Great discussion. It seems to me that every player is subject to those flaws. I mean, the system wasn't designed to uniquely job Tex. So if the 2009 numbers say he is below average or this or that, then I agree that's what the numbers say. You just have to acknowledge that they say that within some kind of margin of error. But to say, well, there are flaws in the system and to uniquely adjust Tex without adjusting anybody else would seem to me to be adding error, not subtracting it.

Defensive stats are more Monte Carlo analysis than the discrete events measured by offensive statistics. I do believe the "saved 100 runs" though if you look at it using a sheer number of opportunities. I imagine having a SS saves 1000 runs versus not having one at all.

I'll post something MGL posted on "The Book" blog during the discussion on Poz and Betancourt (link is 5 posts up in newsblog, (The Book - Poz gets it wrong again) I hope he won't mind:

MGL:

Anyway, getting back to these silly discussions about whether one year samples of metrics are “reliable” or “accurate..”

They are silly because they are what they are. A sample is a sample is a sample is a sample. The best way to explain it is that if a metric is perfect. Absolutely perfect. I mean perfect as in G-d perfect. It better and I mean it BETTER be wrong for a certain exact percentage of players, otherwise someone is cheating.

That is the answer to the, “Why does Teixera have a league average UZR when we observe him to be a great defensive first baseman (assuming that is true, which it may not be).” And until people grasp the concept in the above paragraph, these discussion are always going to be silly and meaningless.

Let me repeat that for ALL the fans and MSM writers who point to Teixera or any other player that UZR MAY have gotten wrong (not that many of them are reading this...):

Let’s start with the premise that a certain metric - any metric - is perfect. Again, I mean STONE COLD PERFECT. We are starting with that premise. That metric MUST BE WRONG for a certain percentage of players over any finite period of time! Again, for that perfect metric, how much wrong and for what percentage of players, depends on the sample size. But, it MUST BE WRONG for some players. So how could the fact that someone picks a player for whom it appears to be wrong (and I’ll concede that it is dead wrong for this hypothetical exercise) be evidence that the metric is anything less than perfect? I’ll answer that question. It can’t be! There has to be some other method of analysis to determine how good that metric is. It CAN’T be by picking one or a few players for whom the metric appears to be wrong. Why? Because the PERFECT METRIC HAS TO BE WRONG FOR A CERTAIN PERCENTAGE OF PLAYERS. THAT IS AN INTRACTABLE LAW OF NATURE OR MATHEMATICS OR WHATEVER YOU WANT TO CALL IT.

Let’s start with the premise that a certain metric - any metric - is perfect. Again, I mean STONE COLD PERFECT. We are starting with that premise. That metric MUST BE WRONG for a certain percentage of players over any finite period of time! Again, for that perfect metric, how much wrong and for what percentage of players, depends on the sample size. But, it MUST BE WRONG for some players. So how could the fact that someone picks a player for whom it appears to be wrong (and I’ll concede that it is dead wrong for this hypothetical exercise) be evidence that the metric is anything less than perfect? I’ll answer that question. It can’t be! There has to be some other method of analysis to determine how good that metric is. It CAN’T be by picking one or a few players for whom the metric appears to be wrong. Why? Because the PERFECT METRIC HAS TO BE WRONG FOR A CERTAIN PERCENTAGE OF PLAYERS. THAT IS AN INTRACTABLE LAW OF NATURE OR MATHEMATICS OR WHATEVER YOU WANT TO CALL IT.

This is all true, of course. The question, though, is what the typical errors are. If 5% of the time, UZR will be off by 5 runs, then we'd expect about one starter in the majors to be off by that much. Or does it have even larger errors? These sorts of things are extremely difficult to estimate for this kind of metric.

I imagine having a SS saves 1000 runs versus not having one at all.

If you watch a bit of cricket, as I do, you'll realize that DL is making an important point here. The mere presence of a player actually adds defensive value in that zone. In cricket, you can add or subtract defensive value by moving your fielders around. Since a batsman can hit into the 360 degree arc, there's always a gap somewhere to exploit. In baseball, most players hit into a much more narrow arc, and over a hundred years of experience has created the optimum fielding positions, which players occupy almost all the time, except in the cases of extraordinary shifts like that against David Ortiz.

Thinking about the contrasts and similarities in defensive actions in baseball and cricket, I increasingly find myself wondering whether UZR and similar complex defensive statistics actually add a whole lot of understanding we couldn't actually achieve with simpler methods. Is something along the lines of the Win Shares defensive system actually close to the optimum trade-off between complexity and accuracy?

On another tack, I would really love to see someone carry out this experiment. I don't have the skills to do it myself. What if we start evaluating fielders in pairs? 1b-2b, 2b-ss, ss-1b, ss-3b, 3b-1b, lf-cf, cf-rf. Has anyone ever tried that?

I am still trying to grasp the notion that Holliday is deemed as poor as Duncan in the outfield.

Because Chris was a bear swatting bees with a honeypot on his hand.