Batter's Box Interactive Magazine Batter's Box Interactive Magazine Batter's Box Interactive Magazine
David Pinto has begun to release his defensive ratings for 2005. These ratings are based on a probabilistic model of range, comparing the number of outs recorded based on zone information with the number that would be expected from league averages and visiting player performance. The Blue Jay middle infielder statistics were very interesting. Thanks to Wildrose for the tip.

Orlando Hudson put up whopping numbers, recording between 76 and 83 more outs than would be expected (depending on which model is used). The second best regular second baseman was Chase Utley who recorded between 46 and 51 more outs than expected. At the other end of the spectrum, Robinson Cano recorded 55 fewer outs than expected.

These figures do seem a little high all the way around. A ground ball converted to an out rather than not converted is worth approximately .6 runs (.5 for the single and .1 for the out) in a 4.5 run per game context according to Tango's linear weight figures. This would mean Orlando saved between 46 and 50 runs with his glove against league averages. For comparison, he was 37 fielding runs above replacement and 17 fielding runs above average, according to Baseball Prospectus. At shortstop, John McDonald was one of the best defensive shortstops in the league. He was the best on ground balls, and signficantly above average if balls in the air were included. Russ Adams was 20 outs, or 12 runs worse than expected. This was below average, but not the worst in the league as David Gassko's Range and BP's fielding stats had him.

Interestingly, David Pinto's worst defensive shortstop rating went to Derek Jeter, by far. This contrasts mightily with BP's account, which has him as an above average defensive shortstop last year. I suspect that we have not heard the last of that debate.

Defensive evaluation remains a particularly difficult analytical topic. We will wait for the UZRs, which will hopefully be released soon. These provide a very important benchmark.
Pinto defensive ratings | 35 comments | Create New Account
The following comments are owned by whomever posted them. This site is not responsible for what they say.
Gerry - Wednesday, January 25 2006 @ 02:44 PM EST (#140223) #
Anecdotal observation....Orlando would catch any ball he could even if it meant calling off Rios, Adams or the first baseman du jour. If Rios and Adams were not so passive and young, respectively, Hudson would not have made as many outs. I am not denigrating )-Dog, just saying some of his numbers are inflated.

Mike Green - Wednesday, January 25 2006 @ 03:23 PM EST (#140228) #
That might be part of the issue with the whopping number for Hudson. Pinto will be posting ground-ball only numbers shortly. He also now indicates that there were errors, which may have contributed to Jeter's standing, and promises to correct them and re-post numbers.
Jim - Wednesday, January 25 2006 @ 03:28 PM EST (#140229) #
We will wait for the UZRs, which will hopefully be released soon


I thought with MGL working with St. Louis they dno't get released in bulk any longer. Every once in a while he'll give a few in a discussion, but I don't think they are available in bulk.
Pistol - Wednesday, January 25 2006 @ 03:34 PM EST (#140230) #
I was under the impression UZRs were not publically available as well.

And I would agree with Gerry. Hudson caught everything he could, even if it was an easier play for Rios (so perhaps Rios is underrated, even when he grades out high defensively). I used to get a good chuckle out of that.
Michael - Wednesday, January 25 2006 @ 03:57 PM EST (#140232) #
Be interesting to see what defense a Koskie-Adams-Hudson-Overbay (or even Hinske instead of Overbay) would have had instead of the Glaus-Adams-Hill-Overbay. Does the defense and millions of $ and Batista make up for the offense difference with the Glaus trade?
Mike Green - Wednesday, January 25 2006 @ 03:57 PM EST (#140233) #
MGL and Tango are publishing a book soon. UZR numbers might be part of it. We will see.

Eyeballing it, it looks as though Orlando had roughly 40 more putouts than one might have expected. Undoubtedly some of those were balls that no one else could have caught and some of those were "choice of fielders". That would still put him at roughly 60 balls converted into outs or 36 runs above average. That is a lot. We'll wait for Pinto's revised data.
Wildrose - Wednesday, January 25 2006 @ 07:45 PM EST (#140247) #
Here's a good article summing up some of the current research done on defence.
Wildrose - Wednesday, January 25 2006 @ 11:58 PM EST (#140255) #
Here's a link that builds upon Pinto's work and translates the outs into actual runs cost/gained. The best translation is probably the ground ball only figures pro rated over 2000 ground balls or what would be an average full season. Russ Adams reflects poorly on these new numbers.
Wildrose - Thursday, January 26 2006 @ 11:57 AM EST (#140266) #
The ground ball only numbers are up at Pinto's sight, and as some intiutive Bauxites predicted, Hudson's numbers dropped by a certain degree. Hudson by this measure is still good, 24-25 outs (not runs) above league second base average, but not off the charts.

The new table factors out pop-ups and line drives, some feel that you need to do this because a veteran player like Hudson will "take charge" and wave off his more inexperienced teamates on high pop ups that either fielder could catch. Line drives are factored out because it's felt that it is not a repeatable skill, basically blind luck or good positioning by the coaches. It would be nice if this metric could be refined to the degree that discretionary fly balls could somehow be weeded out.

Unfortunately we don't have Hill's second base PMR's due to sample size cut offs, so we can't answer properly the impact of the Glaus trade. Still this is pretty instructive stuff.

Mike Green - Thursday, January 26 2006 @ 01:10 PM EST (#140268) #
If you add in a share of the balls in the air, Pinto's numbers for Hudson in 2005 are fairly comparable to BP's, but probably somewhat higher.

Hudson was, according to BP, 60 runs above average during 2003-5 or 20 per year. That is probably a fairer estimate of his defensive value, than the 40+ runs you would get if you allotted all of the caught flies to him.

Incidentally, Adams' standing (28 outs below expected) is the same using groundballs only or including outs in the air. It seems as though Hudson was not taking more pop-ups than one might have expected from Adams, or else Adams was making up for it by getting a lot of balls in the air down the line and back of short. It'll be interesting to see Rios' standing on the PMR.
Mike Green - Thursday, January 26 2006 @ 01:57 PM EST (#140271) #
And, returning to Derek Jeter. He was, according to PMR, 25 outs worse than expected in 2005. Below average, but not the worst in the league. The best everyday shortstop, according to PMR, was Rafael Furcal, who was 19 outs better than average.

My own view is that Jeter remains weak on ground balls up the middle, and that he's still a little below average overall. If you pick the mid-point between PMR's 15 runs below average and BP's 7 runs above average in 2005, that's my best estimate.
Michael - Thursday, January 26 2006 @ 09:48 PM EST (#140282) #
Other positions are out, including CF where Vernon is a slightly below average center fielder.
Mike Green - Thursday, January 26 2006 @ 10:31 PM EST (#140283) #
Pinto described the dying trade talks between Indians and Red Sox as "Cold Coco". Bueno.

I've got some doubts about the Wells rating, but any system that has Andruw Jones, Edmonds and Rowand at the top and Bernie Williams and Ken Griffey Jr. at the bottom meets the basic standard of plausibility.
Glevin - Thursday, January 26 2006 @ 11:17 PM EST (#140285) #
"I've got some doubts about the Wells rating, but any system that has Andruw Jones, Edmonds and Rowand at the top and Bernie Williams and Ken Griffey Jr. at the bottom meets the basic standard of plausibility."

If we came across an offensive stat had Pujols and Vlad at the top and Deivi Cruz at the bottom, but thinks Arod is a below average hitter, it would be rejected totally. I have yet to see a defensive stat, or any defensive system that doesn't have things that are totally out of whack. To me, it's like a unified theory of physics. People can do it, but I just don't see how you can take any of these seriously when there are such massive flaws in all of them. IMO , by a long-shot, the value of a players defensive contribution is best told through what baseball people (and my own eyes) think.
Wildrose - Friday, January 27 2006 @ 12:37 AM EST (#140287) #
Glevin, I think defensive stats are still rudimentary, but to call them "totally out of whack" is not accurate. Generally players who are thought as good fielders measure well across several metrics, and on occasion you'll find an odd discrepancy, but not very often.

I don't discount observation regarding defence, but then how do you ascribe actual value to your opinions.
The value of Pinto's work, while far from perfect, gives you an idea how much value a player actually has,( e.g. we know John Macdonald is " good defensively", but now we have a measurement of roughly how good, 22-23 runs above the average shortstop.)
NDG - Friday, January 27 2006 @ 08:42 AM EST (#140288) #
I've got some doubts about the Wells rating, but any system that has Andruw Jones, Edmonds and Rowand at the top and Bernie Williams and Ken Griffey Jr. at the bottom meets the basic standard of plausibility.

I've always considered Wells to be an average to above-average fielder. If he were below average that would not surprise me. I've said in the past that Wells won his gold gloves on offense not defense.

Mike Green - Friday, January 27 2006 @ 09:34 AM EST (#140291) #
For what it's worth, Baseball Prospectus has Wells at 12 runs above average over the last 3 years, or 4 runs per year above average. To my untrained eye, he has lost some speed over the last couple of years, but so far he makes up for it by getting good reads and taking direct routes to the ball.
Wildrose - Friday, January 27 2006 @ 09:50 AM EST (#140293) #
It should be noted the Pinto tables account for only "range", it does not account for a players value/shortcoming in regards to his throwing. I think Wells would probably gain back some value in this area.
Wildrose - Friday, January 27 2006 @ 10:03 AM EST (#140295) #
In 2004 Pinto had had Wells doing a little better in regards to his range than 2005.
Mike Green - Friday, January 27 2006 @ 11:14 AM EST (#140297) #
It is generally advisable to look at several years of defensive statistics.

In the case of offensive statistics, we know that all statistics fluctuate over the years (for reasons as varied as injuries, developmental or personal issues, and quality of pitching faced), but batting average fluctuates more than power or plate discipline. The reason is that the element of luck is present, in, for instance, the flare and the Baltimore chop, and for a single season, this can be particularly important.

In the case of defence statistics, we are essentially measuring batting average, and our measures remain fairly unrefined. Russ Adams may have fielded 10 Baltimore chops in 2005, where he had no play and Juan Uribe may have fielded 5. These things do tend to become less important over a period of several years of full-time play.

I wish that Baseball Prospectus published how they calculate their fielding runs above average. I find that their scale corresponds with my intuitive sense of the importance of various defensive positions, and the runs that can be lost or gained by defensive play. A great, great centerfielder such as Willie Mays in his youth saves 20 runs per season above average, although his skill deteriorates somewhat fairly early on. A great, great middle infielder such as Mazeroski or Ozzie save perhaps 25 runs in a great season, and this skill deteriorates more slowly.

Hudson's 60 fielding runs above average over 3 years is, by BP's measure, among the greatest fielding performances ever. The great defenders from Maz to Ozzie to Rabbit Maranville to Honus Wagner have garnered between 60 and 65 fielding runs above average in their best 3 consecutive seasons. The best 3 year number that I could find was, to my surprise, Dave Concepcion's 74 runs above average in the mid-70s. Concepcion's six year peak number of 128 FRAA is exceptional. My own observation was that while Concepcion was great, Ozzie was a little better. However, I do respect the statistical evidence to the contrary.
Gitz - Friday, January 27 2006 @ 02:55 PM EST (#140301) #
IMO , by a long-shot, the value of a players defensive contribution is best told through what baseball people (and my own eyes) think.

I agree with this. There are some obvious statistics you can rely on -- i.e. a guy who makes 54 errors is probably a bad defender, no matter the subjectivity of the error rule -- but in general, this comment is bang on. The Internet baseball community knows more than baseball people care to acknowledge, but that works both ways: baseball people know more than the Internet baseball community care to acknowledge.

Mike Green - Friday, January 27 2006 @ 03:37 PM EST (#140302) #
Many baseball people, including Branch Rickey, found the need to attempt to quantify these things. Long before Rickey, boxscores contained all kinds of defensive statistics.

The problem is that some people's eyes say Vernon Wells is the best centerfielder in the league, while others (as you can see from this thread) say he is an average centerfielder. The spread is even greater for Derek Jeter, even among those who watch 100 Yankee games a year.
Mike D - Friday, January 27 2006 @ 06:02 PM EST (#140309) #
The problem is that some people's eyes say

That's a problem, Mike G, as our eyes are subjective and unreliable. But it's not the only problem.

I'm more inclined to agree with Gitz and Glevin on this point, primarily because the various metrics themselves disagree so much. Both defensive analyses are objective -- i.e., they crunch numbers and don't trust anyone's eyes -- yet they come to wildly disparate results for the likes of Vernon and Jeter. If two smart analysts can put together models on the same season's worth of data and one can conclude that Jeter was above-average and played by far the best defensive shortstop of his career, and the other says that he was the worst in the league by far, then I am frankly unpersuaded by either.

Getting back to Glevin's point, if you were to take three offensive metrics like, say, OPS, Adjusted OPS+ and Win Shares Above Bench, there might be some adjustments in relative rankings for offensive players based on park factors, durability, etc., but you would never have a situation where a batter would grade out above-average in one metric and comically bad in another.

A lot of this has to do with the nature of the offensive game versus the defensive game. In any given game, batters will get a regular turn every ninth at-bat. Once they're at-bat, they'll see hittable pitches more or less in the strike zone. If they see nothing but unhittable pitches, then good players will walk while bad players will flail away. But more or less, everyday players will have similar opportunities to hit similar pitches, and thus we can feel confident that, over time, we can compare them to each other and announce who performed better.

But defensively, you get balls that go all over the field, and reach their destination at wildly varying speeds and trajectories. On Opening Day, Vernon Wells might have five balls hit to centre that Willie Mays couldn't have caught. He might have five balls that I could have caught. Or he might not even have five balls hit to him. We can assume that this will all net out by the end of the year, but it frankly may not. In the more orderly world of hitting, the "breaks" are far more likely to come out in the wash.

The other problem is that hitting is an individual act, while there is a team component to fielding. As a hitter, there is no downside to being "greedy" about how often you can get on base or drive runners in. But defensively, the ground you cover is related to and impacted by the players around you. An illustration: My friend Jeff and I were at a Twins-White Sox game in Chicago, and the Sox hitter lofted a fly ball into left centre with two outs. Jacque Jones camped under it, called it, and at the last second Torii Hunter jumped in front of him and snagged the ball. Hunter and Jones were laughing and play-fighting as they jogged off the field, but Jeff and I commented at the time that there will be a very well-intentioned computer program that will credit Hunter's glove and demerit Jones' based on that meaningless hot-dog play.

I used to say, in a casually hyperbolic way, that it's impossible to overstate how great Orlando was with the glove. Apparently not impossible! I mean, had the Jays employed a perfectly competent league-average second baseman in the 130 games played by the O-Dog at second, they would have allowed fifty more runs? You might as well say that Hudson saved a zillion runs. He was tremendous with the glove, but 50 saved runs is beyond the limits of what a second baseman can physically do.

I suspect that it will continue to prove difficult to nail a definitive defensive metric. But I know that we're not there yet.
Glevin - Friday, January 27 2006 @ 07:51 PM EST (#140314) #
"Generally players who are thought as good fielders measure well across several metrics, and on occasion you'll find an odd discrepancy, but not very often."

Generally is not good enough. It needs to be all the time or at least 99% of the time. It isn't and the systems don't even agree with one another.


"I don't discount observation regarding defence, but then how do you ascribe actual value to your opinions.
The value of Pinto's work, while far from perfect, gives you an idea how much value a player actually has,( e.g. we know John Macdonald is " good defensively", but now we have a measurement of roughly how good, 22-23 runs above the average shortstop.)"

My point is that you can't say that. I can say roughly how many runs John Mcdonald would cost you on offense with great precision, but a stat that says "Macdonald is 22-23 runs better than average" I find absurd. If you want to be so exact, you need a perfect, or near-perfect system. One of the many absurdities of the system is this...(using BP's defensive method, but all the ones I have seen have this flaw)

Cal Ripken FRAA

1986- 23
1987- -7
1988- -6
1989- 21

Are we to believe that Ripken was a great defensive SS when he was 25 then bad when he was 26 and 27, then great again when he was 28? Even Ozzie Smith has periods where he goes from being great to being OK to being below average to being great. There are so many problems with every system, that the unified stats saying "this player was worth.." are useless. Defensive stats individualy are useful IMO. if Andruw Jones is making 50 more POs than any other CFer, well that says something. If B.J. Upton is making 50 errors, that says something.
Wildrose - Friday, January 27 2006 @ 09:49 PM EST (#140318) #
Houston we have a problem. Unfortunately apples and oranges are being incorrectly compared. In other words comparing Pinto's tables to that of BP'S, assuming they are the same, is just plain wrong ( in hitting terms it would be like comparing on base % to VORP , and then wondering why the two don't match up).

Pinto's numbers essentially measure only one thing , range factor. Actual play by play data is analyzed to determine how a particular player compares to his peers regarding range , nothing more.

The BP'S numbers however, attempt to measure OVERALL defensive value. Range is just one factor in overall defensive ability. BP attempts to look at ALL the various components of defence and tries to roll them into one number. Such things as sure handedness (not making errors),the ability to turn double plays, outfielders throwing out runners, and range factor are believed, ( I say believe, because Davenport has never publically released his proprietary formulas) to be melded together to come up with an overall defensive mark.

Unfortunately the sabermetric community has not been very clear on differentiating and explaining how various defensive systems work... confusion is frequent and somewhat understandable.
Mike D - Friday, January 27 2006 @ 10:47 PM EST (#140319) #
Wildrose, I apologize if I fell into the trap you just described. Thanks for clarifying...but if that's true, then how can a system that omits several facets of defensive value then purport to specifically state how many outs above or below average a player recorded?
Wildrose - Friday, January 27 2006 @ 10:50 PM EST (#140320) #
Pinto does not say, as incorrectly asserted, that " Jeter is by far the worst in the league " defensively, rather by only using groundball data ( filtering out the noise of discretional flyballs) , he states, Jeter is 14 runs below average in regards to RANGE.

Perusing Primer the consensus seems to be that this is accurate. It's felt Jeter would gain back defensive value if such things as catching shallow line drives in the outfield, making good relay throws, turning a good double play, and not making a lot of errors was to be measured. Unfortunately Pinto's work does not capture this type of data. Much debate remains as how to measure these other defensive factors. Davenport gives Jeter a 106 rate measurement( a little above average), so obviously he feels that he does well in these other areas, again as I've said, and as MG has stated, we lament that BP's methadology is not public knowledge, so some question its validity.

Hudson is not 50 runs ( or as Mike says a zillion) runs over the league average range factor, rather when you read the data properly, Pinto has him at 15 runs over league average range factor ( using groundball data) for 150-155 games. For me that seems about right.

Regarding measuring defence , were definitely not there yet, but to throw out all defensive measures as a group, I think is premature.
Wildrose - Friday, January 27 2006 @ 11:01 PM EST (#140321) #
Mike this stuff is terribly complex, so many factors, so many things affecting the outcome, I barely understand all of this myself. I do somewhat like Pinto's measurements for infielders, as for outfielders, I'm not as sure, you've intuitively noted in your comments, many of the same concerns that others have expressed on Primer regarding discretional fly balls and blind luck skewing outfield numbers.

We still have a long ways to go, but what a fascinating area.
Craig B - Friday, January 27 2006 @ 11:01 PM EST (#140322) #
Cal Ripken FRAA

1986- 23
1987- -7
1988- -6
1989- 21

Yeah, and look at this

Cal Ripken BRAA

1986- 21
1987- 4
1988- 27
1989- 1

Are we to believe that Ripken was a great offensive player when he was 25, then average when he was 26, then great when he was 27, then average again when he was 28? There are so many problems with every system, that the unified batting stats saying "this player was worth.." are useless too, dammit!

Seriously, turning the sarcasm off for a sec, remember that because a statistic doesn't do everything you want it to do doesn't make it worthless. Statistics, ratings and methods don't just come in two flavours (perfect and worthless). You can't ask the stat to do too much (which is why saying that 'PMR says so-and-so was the worst defender at his position' is sloppy, sloppy, sloppy) but on the other hand just because the statistic doesn't pour your beer and scratch your ass for you doesn't make it "worthless".

More thoughts on the subject at Tyblog.

Mike D - Friday, January 27 2006 @ 11:09 PM EST (#140323) #
Thanks again, Wildrose. The refinements and adjustments make Pinto's new analysis far more realistic.

Since you are clearly well-versed (and I am definitely not) in the contours of the defensive metric debate, I was wondering if the two following questions have been addressed. (Then I have to head out, I'm afraid, though I'll try to rejoin the debate tomorrow.)

First, have defensive analysts debated the issue of distribution of balls in play? In other words, let's assume that Vernon Wells and Torii Hunter have absolutely identical defensive performances over the course of a season, except an extra 200 perfectly routine fly balls get hit to Wells -- does Vernon reap the benefit statistically? Or do they control for that?

Second, have they debated the issue of (for lack of a better word) responsibility? In other words, strikeouts are the batter's "fault" in the sense that the batter either fanned on strikes or chased bad pitches, and walks are the pitcher's "fault." But what about a crisp single to centre that no CF in history would be able to catch? Do they count that against the CF? Ignore crisp singles? Assume that every CF has a roughly equal number of crisp singles?

I would like to be educated.
Wildrose - Saturday, January 28 2006 @ 02:31 AM EST (#140326) #
Craig since the beginning of time man has tried to quantify and measure the world around him, to create a construct to better understand the world.

If you spend a good portion of the day thinking about baseball, trying to understand defence and creating a construct about this topic can be both frustrating and stimulating ( what in tarnation is a prolegomena?).

Mike, I'm not nearly as well versed as you may think regarding the contours of understanding defence, but I'll think about your questions. Following the discussions in Primer, I'm in awe of some of the brilliance in regards to unlocking the riddle of measuring defence.

Pinto does apparently factor in errors regarding his range measurements, I gave the impression earlier that he didn't.

Mike Green - Saturday, January 28 2006 @ 02:40 PM EST (#140335) #
Actually, Pinto originally posted Jeter as worst in the league by far, taking into account both balls in the air and balls on the ground. This was due to an error, which has now been fixed. He now has Jeter at 30 outs below predicted, which is better than Adams among regular shortstops and no others. There is a large discrepancy between Pinto's rating of Jeter, and BP's.

Pinto's system, according to Studes in THT, takes into account exact location and speed of each batted ball. There are issues that result from positioning of the fielders, particularly for infielders, which, to my knowledge. So, in answer to Mike D's question, Pinto's system (and UZR) do attempt to ascertain the difficulty of the play in evaluating fielder opportunities. There is little question though that our defensive statistics are less accurate than our offensive statistics. Defence is a more difficult thing to measure.

Jim - Saturday, January 28 2006 @ 04:53 PM EST (#140336) #
This current thread at Primer:

http://www.baseballthinkfactory.org/files/numbers/discussion/baseball_musings_pinto_probabilistic_model_of_range_2005_center_fielders/


Is one of the more interesting I've read in a long time. I've always thought that MGL's personality held back the acceptance of UZR in the past. I'm very persuaded by some of his comments in this thread though.

Mike Green - Tuesday, January 31 2006 @ 11:23 AM EST (#140422) #
Here are the rightfielders PMRs from David Pinto. Rios shows as 10 runs below expected (despite being way above average on Gassko's Range and BP's Fielding Runs Above Average). This does tend to support the anecdotal observations made by Gerry and Pistol above that Hudson was taking more than his share of "choice of fielder" short flies to right, and that PMR's figures give too much credit to Hudson and not enough to Rios on account of these plays.
Mike Green - Wednesday, February 01 2006 @ 09:46 AM EST (#140475) #
The PMR for leftfielders was released yesterday. Reed Johnson had the highest rating, while Catalanotto's was a little above average. BP's statistics show Johnson as 9 fielding runs above average as a leftfielder over the last 3 seasons (155 games). That is a lot for a leftfielder.
Pinto defensive ratings | 35 comments | Create New Account
The following comments are owned by whomever posted them. This site is not responsible for what they say.