Also, how does UZR work that having "fewer balls hit his way" would help Jeter? That doesn't sound right. Isn't that like saying a person's batting average is helped by having fewer At-Bats?
Forums | FanGraphs Baseball » Baseball Statistics
Buck and McCarver dis Jeter fielding critics...
(46 posts)-
Zachary Klaas
MemberPosted 1 month ago # -
Jamesian
MemberYes. That's not all that bad a comparison actually. If you go 2-for-4, you have a .500 batting average. If you have 500 at-bats, you aren't going to hit .500. It's a lot easier hitting .400 over 200 at-bats than it is over 600.
If you read Tango's explanation with me of how UZR works, you know that a UZR can fluctuate by as much as 20 runs with no change in skill level. Just by a change in one year to the other in the way that balls are hit at a fielder.
If you know anything about UZR, you also know that a negative fielder can have a positive UZR and vice-versa. UZR doesn't tell you how good a fielder is. It tells you how their defense impacted the team that year. Usually, that means that the best defensive players have the best UZRs. But not always.
So Jeter's positive UZR doesn't necessarily mean that his defense is one bit better than it has been in the past.
If Jeter gets hit fewer balls, there are fewer chances for him to cost his team a run by his lack of range. That helps his UZR. The more balls hit at him, the more balls that can get through.
On the other hand, if you hit a lot of balls at Elvis Andrus, you'd probably only be doing him a favor with his UZR because that just gives him more chances to steal hits and save more runs for his team.
Theoretically, Adam Dunn could play SS for an entire year and if he didn't get hit any balls, he'd probably have a 0 UZR. That wouldn't make him an average SS. He just didn't cost his team any runs that year because no balls were hit to him.
The number of balls hit to a fielder and the difficulty of those balls affect their UZR. Fielding numbers can fluctuate from year to year due to those factors.
Posted 1 month ago # -
Kincaid
MemberThe range component of UZR measures range. There are other components (error rating, arm/DP ratings) in the total UZR on FanGraphs that give a more full view of fielding ability and value, but if you only care about range for whatever reason, you can just look at the range component.
The range component of UZR will give you a better measure of a fielder's range than range factor because of how each measures opportunities. Both measure plays made. UZR then looks at the distribution of batted balls when a player was in the field and estimates the number of plays an average fielder at the position, then compares the number of plays the fielder actually made to that estimate. RF just assumes that knowing the number of innings played in the field tells you enough to guess at the opportunities, but in some cases, such as comparing Lugo and Jeter in 2005, that assumption is pretty far off. So as long as the frequency and distribution of batted balls are normal when a fielder plays, UZR and RF will both tell you more or less the same amount of information, but when either the frequency or the distribution of balls in play for a fielder changes (say, he gets fewer ground balls in general, or just more balls hit to the opposite side of the field, for example), then UZR will tell you a lot more and RF will be biased either for or against the fielder. The problem is, you can't tell when the frequency and distribution are normal or abnormal when you are looking at RF, so you have a stat that may be fine, or it may be biased for or against the fielder, but it can't tell which it is. Then, in UZR (or other similar defensive stats like plus/minus) you have a stat that identifies the ones RF is biased for or against and accounts for the bias. So the latter will give you a better measure for a given fielder.
To get RF to tell you anything useful once you already have another stat that also measures plays made but addresses its biases, you would have to address the biases in RF, such as by correcting for pitchers and batters and parks (particularly for outfielders). Once you start making the adjustments, though, you no longer have RF. This is basically what Tango was talking about in the other thread with his WOWY technique. If you do that, then you have another good stat to compliment stats like UZR.
There will still be some measurement error in UZR, just not the systematic biases you see in RF. UZR lumps balls into different bins based on where each was hit and what type of batted ball it was and assigns each bin a value. So you might have a bin with an average value of +.5 plays where some balls in the bin will actually be worth +.55 and some will be +.45 (I'm just making these numbers up for illustration purposes), but to make sure the bin has enough balls in it to accurately determine a value, you have to lump these together. So you might end up with someone who gets more balls than normal in that bin that easier than average so that they are worth +.47 each but you think they are +.5, so that UZR will overestimate the difficulty of these balls.
This is much less of a problem than with RF for a couple reasons. One, all that has to happen for RF to be off on the opportunities is for a fielder to get more balls in play or ground balls or whatever, because it doesn't even try to count the number of balls hit toward them. It doesn't care if a ball is hit right at a fielder or if it is a strikeout, it counts both as the same opportunity. So it is going to be off quite a bit, and it will be way off a lot of times. With UZR, it is just when the balls within each bin are distributed oddly that you will have this problem, which is both less likely and much less extreme when it happens. Two, these errors in UZR are mostly random. The likelihood of error in the measure is more or less the same for everyone, it decreases as the sample increases, and you can estimate what it is. For RF, there are systematic biases. A shortstop in front of a low-strikeout ground ball staff will always be overvalued relative to a shortstop in front of a high-strikeout fly ball staff. It won't go away no matter how large your sample gets as long as the biases are still there, and in order to tell how big the error might be, you pretty much have to start turning RF into something more UZR-like anyway. Plus, you'll still have the random error on top of that, only moreso because you are not trying to identify the distribution of balls in play, which adds an extra element of randomness. You can also reduce your chances of being thrown off by error if you look at something else like plus/minus that does more or less the same thing but has its bins divided up differently because the chances of the balls being distributed oddly within your bins in both stats is lower than in just one.
Having more or fewer balls hit to you won't make a good UZR bad or vice versa, it will just tend to make UZR a bit less extreme. A fielder might go from +5 to +4 with a pitching staff that doesn't yield a lot of opps, but it won't make him appear bad, and it won't make a sub-par fielder appear above average. Random variation can, but that is random as far as we can tell, and having more or fewer opportunities doesn't tell you anything more about how someone is likely to be affected than just how it changes the sample size. This is why FanGraphs also presents expected outs, defensive games and UZR/150. Defensive games is based entirely on the number of opportunities, so you can see who got more or fewer opportunities, by how much, and what they're UZR would be normalized to a typical full season. Or you can just look at UZR; it's completely up to you how you want to look at the numbers, which is a nice feature of how FanGraphs publishes the stat.
UZR tells you how good a fielder in more-or-less the same way as other stats do. A player's stats are never his true talent level, and how well the stat estimates true talent level depends on how well the stat reflects value in real life, how stable the statistic is, and how big the sample is. For example, wOBA is good for estimating how good a hitter is because it reflects actual value. Batting Average is not as good because it has a lot of biases. UZR does well in this, so as long as we know how stable it is and what the sample size is, we can have an idea of how well it measures true talent, which is the important thing. So, in the same sense that no stat tells you exactly how good a player is, UZR doesn't tell you how good a fielder is, but it does about as good a job of estimating fielding ability as any stat as long as you know how to handle the sample size.
As far as Jeter goes, chances of his UZR being skewed high are likely greater than normal because we have so much data that says his true talent is probably below average, but his UZR is also an indication that he probably did actually field better this year than he has in the past.
Posted 1 month ago # -
Jamesian
MemberOK. I'm not really saying that RF really has any value when compared to UZR. I'm just kind of pointing out that it might lead you in the right direction. With Jeter, it has reported him as being one of the worst SS in baseball in most years so obviously it hasn't been too far off base and the more years you look at, the more likely you are to come to the same conclusion as UZR reports.
So if I look at Jeter and he has a flat UZR and I look at TZ and he has a -4. And then I look at RF and he is at the bottom of the league. I might conclude that he didn't have an average year in the field.
Or a guy like E. Aybar last year had a +10 UZR and a -10 TZ but a pretty good RF. If I didn't know anything about Aybar, I might have concluded that he might have been a closer to plus 10 rather than a minus 10 or just an average fielder. It might have been offbase but since RF and UZR/TZ have some amount of correlation, it might have lead me in the right direction more times than not.
The way I look at it, the more things I look at, the better assessment I can get. RF does not tell you what UZR or TZ but it can tell you something. It's not that far off base most of the time.
Posted 1 month ago # -
lee
MemberDeparting from "Zone Expectation" based evaluations and Range Factor inputs, I still have Jeter, rated purely as a defensive SS, averaging @ -9-10 DR/seasons over his fourteen year career as a full-time regular. I do have him above average in the three GG seasons; and it's easy to argue that he deserved at least one, maybe two of those GG's. Tejada may well have deserved the other.
In 2009, my raw defense rating, unadjusted for pitching staff K's = .912, -11.99 DR. However, the NYY staff racked up 160 K's above League average; so adjusted for equivalency, Jeter = .932, -4.49 DR. Here are my arguments to support the position that Derek Jeter is NOT the worst SS in either the A.L. or all of MLB:
1. The things that Jeter actually does, the things that make him "The Captain," seldom go into defense statistical data:
The Subtle
-unique sense of position in game moment context, to go to the right place at just the right time and to execute with a tag, a throw, a running feint to/at the perfect place on the field [two instances so far in the current LAA series];
-uncanny sense of showing up in unexpected places in critical defensive situations as backup or relay man, that innate sense like Pete Maravich, Bob Cousy or Michael Jordan exhibited in the NBA;
-commanded respect of teammates for being the "go-to" guy in the field;
The Positional; Mechanical:
-It can be argued that a SS will be co-dependent on his 2B teammate; and in his first two seasons, that was a revolving door with the NYY:
-an aging Mariano Duncan alternating with Sojo;
-three rather stable seasons with an average 2B in Chris Knoblauch;
-three full seasons with Alfonso Soriano at 2B
-one season with Miguel Cairo; [Jeter 1 GG]
-2005-present with a stable Robinson Cano; [Jeter 2 GG]
Team~Pitching Interdependent:
If a MLB team's defense is weak or has weaknesses at critical positions, such seive-like tendencies should be revealed in things that might hurt the pitching staff, like WHIP vs. League. Conversely, if a team's pitching staff racks up lots of K's above League average, the position players behind them are deprived of fielding opportunities to other players with low K staffs. Sometimes, highest to lowest K's can range up to 400 or more.
Season...WHIP/LG WHIP.......K's > LG
1996......1.444/1.505........+135
1997......1.359/1.443........+134
1998......1.251/1.432........+ 56
1999......1.377/1.486........+117
2000......1.429/1.490........+ 38
2001......1.305/1.391........+236
2002......1.270/1.383........+134
2003......1.291/1.385........+138
2004......1.369/1.416........+ 22
2005......1.369/1.362........- 2
2006......1.357/1.414........- 9
2007......1.431/1.412........- 55
2008......1.364/1.391........+ 75
2009......1.352/1.403........+160Total.......................+1,179
Yankee pitching WHIP exceeded League only twice, 2005 and 2007; and the worst ranking out of 14 teams was 9. WHIP for the NYY average top 4-5 over most of Jeter's career. This would not betray a serious leak at a critical position.
It's interesting to note that in the three GG seasons for Jeter, his raw defense rating and adjusted rating for K's were virtually unchanged.
Over 14 seasons, pitcher K performance has reduced balls in play for fielders by 1,179. That averages 84.2 per season; and if we assign a distribution of 18% to SS, that's 15/season @ SS.
2004, 2205 and 2006 were very nearly League average, so neither rating would bias for or against Jeter.
Just to cap it off, let's consider Jeter OVERALL over 14 seasons:
-ten time AS;
-3 time GG
-3 time Silver Slugger
-7 times, top 12 MVP voting
-career Stolen Base success rate = 79.2%
-2009 [age 35] Stolen Base success rate 85.7%Posted 1 month ago # -
Zachary Klaas
Member"-unique sense of position in game moment context, to go to the right place at just the right time and to execute with a tag, a throw, a running feint to/at the perfect place on the field [two instances so far in the current LAA series]"
I was going to mention that myself, but somehow never got around to it...especially the feint issue. Does UZR even measure it when a player looks somebody about to score back to third base? Both Jeter and Mark Teixiera were doing a _lot_ of this during the ALCS games so far.
Posted 4 weeks ago # -
gkiger1620
MemberI'll agree that Jeter is a smart player - he makes the best of his talents. But last night's game was telling in that he barely missed getting to a couple of balls that would have made the difference in the game. If the Yankees had a shortstop with better range they would be in the Series this morning.
Posted 4 weeks ago # -
dorsal
MemberI thought the same thing gkiger1620. Especially on Vlad's single at the end. In the replay's I've seen you can't tell if Jeter was cheating in the hole. How can we quantify (and maybe we already are and I'm ignorant) where the infielder stands at point of contact of the batted ball when determining range? Had Jeter been shading middle it would have looked routine.
Posted 4 weeks ago # -
Zachary Klaas
MemberDorsal, I don't know the answer to that either, though I have seen these sites, which pertain to the issue you've raised...I've brought these to the attention of the Fangraphs faithful here before:
http://otherfifteen.blogspot.com/2008/04/can-we-measure-fielder-range.html
http://otherfifteen.blogspot.com/2008/04/derek-jeter-vs-troy-tulowitzki.html
The first of these sites points out that no one has a collected set of measurements of where the player was when the ball was hit, and that this would be a good thing to have when measuring range.
However, the second site does a good job of showing where players were when the ball was fielded. In what you've just posted, you're taking Jeter to task for not "shading middle"...but this is what he normally did, at least in the pre-training period. Ironically he wasn't in position to field the ball you're talking about because he's following the saberists' advice to position himself _farther_ from second.
By following that advice this year, Jeter apparently kept the UZR enthusiasts happy and people seem provisionally willing to possibly suggest that, if he keeps this up, he will be considered as "having range".
Of course, he would have made the play if he were Ozzie Smith. Someone could hit the ball behind the catcher and Ozzie would reach it. :) Okay, that's a little bit of hyperbole, but the guy was good.
Posted 4 weeks ago # -
dorsal
MemberBrilliant. I'll have a read.
Posted 4 weeks ago # -
Kincaid
MemberHit F/X and Fielder F/X systems are in development that aim to track batted balls and fielders much like the Pitch F/X system tracks pitches, so those would provide information on fielder positioning in the future. For now, it could possibly be done the same way as hit locations (basically, a stringer watching the game marks the observed location on a screen), but it isn't done. This wouldn't be feasible for MLB's Gameday data because that data is all input in real time at the game, but for BIS, which uses video of the game, it might be possible. I'm guessing they don't because either:
-They only have the actual telecast of the game and don't have access to non-televised camera angles, so they don't have visuals of every fielder at the start of the play (I don't know if this is true or not, it's just a possibility for why they wouldn't track this)
of
-They can track it, but don't feel the benefit is worth the extra time and money they would have to put into it.
Scatter plots don't really tell you anything about where a fielder was positioned, they just tell you where he fielded the ball, which is exactly what the advanced zone-based metrics tell you. There is no additional information, it just translates to a visual better so it looks like you can guess where a fielder was positioned (even though you could do the same thing from the metrics if you broke them down by location). They actually give you less information about positioning than the advanced metrics because they don't differentiate between different situations. The metrics will look at a double play situation or an infield in situation or a likely extreme defensive shift and make adjustments for the difference in positioning, whereas everything is just lumped into one data set on a scatter plot. So where the metrics will see a fielder moving in or shading toward the bag and realize that the range on different plays is centered around different points, the plot will just pick up the points where the ball was fielded and inflate the range whenever the fielder changes positioning. Including pop ups in the plot further obscures the measure of range. The plot also can't tell you as much about how good a fielder is because it has no sense of opportunities. A fielder who just randomly slid his positioning back and forth all over the left side of the infield would have a huge range in the scatter plot even though he would get to way fewer balls. The metrics would say, correctly, that he has no clue what he is doing.
If you wanted to guess positioning based on the metrics like you would with a scatter plot, you could look at where the balls he got to and didn't get to were. Plus/minus breaks up ratings into plays to the left and plays to the right (and straight on) for this purpose. If you have the input data, you could run it to get whatever breakdowns on locations of fielded balls you want. Just like looking at the scatter plots, it wouldn't actually tell you where the fielder was positioned, but you could make guesses about it.
There is a debate about whether knowing the actual positioning is even important. As long as you adjust for the fact that fielders are positioned differently in different situations, when you have enough data, you don't necessarily add anything by including positioning. If you only have a week to evaluate a player, that would be a great thing to know, because maybe he is just positioned oddly over that week for whatever reason and his range is either inflated or deflated because you caught an aberration in the sample. If you've got a fielder who is consistently positioned to get to fewer balls or more balls over a few seasons of data, though, then that's likely because he is doing something wrong or something right. Sort of like if you want to see who can hit balls on the outside corner better, you don't care if one hitter stands 9 inches further back from the plate than another. You wouldn't say you need to look at how the former hits balls in the middle of the plate to compare to how the latter hits balls on the outside corner because each hitter is responsible for positioning himself in a better position to reach as many pitches as possible. It is the same with a fielder as long as you have enough data and account for the different situations. If someone has to position himself further up the middle or further in because he has little range to one side or doesn't have the ability to make a quick, strong, accurate throw from as deep in the hole as someone else, what most people really care about is how that affects how many balls he can get to, which is what you're already measuring with the metrics. The positioning data doesn't necessarily tell you more about how good a fielder is as much as it tells you more about why a fielder is as good or bad as he is.
Posted 4 weeks ago # -
Zachary Klaas
MemberI like the scatterplots better because the information it gives is straightforward and not the product of an unnecessarily complex measurement logic. What you see is what you get.
Maybe the advanced metrics really are advanced. All I know is that the things I've heard about them don't convince me that they are, because, frankly, I'm never sure what these metrics are actually measuring. The scatterplots may convey less info, but they do this, at least, in a clear manner.
Posted 4 weeks ago # -
dorsal
MemberIn English soccer there is a company that determines the distance each person runs during a match. They are a performance analysis service for teams.
http://www.prozonesports.com/product-prozone3.html
I think this type of thing....cameras in the stadium with computer software to analyze film would give us the true range.
Posted 3 weeks ago # -
dorsal
MemberExample http://www.refworld.com/features/16/article/how-is-prozone-used-to-facilitate-referee-analysis/44
Posted 3 weeks ago # -
Kincaid
MemberThat looks a lot like the fielding F/X system Sportsvision (the company that developed pitch F/X) was working on this year. They had it set up in AT&T Park and were experimenting with it there. I remember seeing a short demonstration on one of the nationally broadcast games from San Fran, so it seems like it is working at least to some extent already. I don't know how long it will be before they have it ready for the same level of use as pitch F/X, though.
Posted 3 weeks ago # -
dorsal
MemberSportsvision looks like they are doing alot of good stuff.
Posted 3 weeks ago #
Reply
You must log in to post.