Sunday, February 05, 2006

Aggregating the Range Statistics

Over the course of the winter, I’ve presented four different fielding range metrics in the following posts:

Article 1 (Range Factor, Zone Rating)

Article 2 (David Pinto’s Probabilistic Model of Range or PMR)

Article 3 (David Gassko’s Range)

Two other respected measures – Ultimate Zone Rating (UZR) and Davenport’s Fielding Translations (DFT) are summarized in an article written for Hardball Times by David Gassko. I have not talked much about these measures because the data are either not available (UZR) or would need to be tediously data entered (DFT). So, I will continue to concentrate on the first four measures for now.

None of the range metrics are nearly as reliable as hitting metrics but much progress has been made in the elusive measurement of range in the past several years. I think we can safely say that players who do well on all of the metrics have good range and that players who do poorly on all the metrics have limited range. However, there is a lot of variation across measures.

So, rather than rely on just one of the measures, I thought it might be a good idea to get an average across all four measures. For example, Brandon Inge was 1st in Range Factor, 6th in Zone Rating, 1st using the Gassko Range and 17th using the Probabilistic Model of Range. This comes out to an average rank of 6.3.

I won’t give the ranks for the individual methods for all the players because I am not allowed to release Gassko’s ratings at this time. I will, however, give a best ranking, a worst ranking and an average.

This analysis is a work in progress. Ideally, I would like to add UZR and DFT to the mix and perhaps compute a weighted average of all the measures. For now, I’ll just do an unweighted average of the four measures which are available to me. For those who are interested in the details, note that I used Gassko’s runs saved above average prorated to 150 games and I used the smoothed PMR using ground balls only. The other two metrics are zone rating and range factor.

The table below presents the third basemen with 600 or more innings played. There were 26 qualifying third basemen in 2005. The table shows that Inge had the third best average rank in baseball behind David Bell and Eric Chavez. I’ll present other positions as the week goes along.

Player

Team

Innings

Best Rank

Worst Rank

Avg

Bell

Phi

1297

1

8

4.0

Chavez

Oak

1348

1

12

6.0

Inge

Det

1400

1

17

6.3

Nunez

StL

721

2

13

6.8

Ensberg

Hou

1286

5

12

7.5

Burroughs

SD

657

3

22

10.5

Crede

CWS

1120

7

16

10.5

Beltre

Sea

1326

8

15

10.8

Mora

Bal

1290

9

14

11.0

Boone

Cle

1250

6

15

11.3

Wright

NYM

1404

4

18

11.5

Koskie

Tor

674

2

26

12.8

Atkins

Col

1162

4

20

13.5

Gonzalez

TB

780

11

19

13.5

Mueller

Bos

1209

3

22

13.5

Glaus

Ari

1264

4

24

13.8

Jones

Atl

830

2

24

14.5

Lowell

Fla

1127

9

19

15.5

Cuddyer

Min

816

14

18

16.8

Teahen

KC

1068

3

25

17.0

Rodriguez

NYY

1385

4

26

18.5

Castilla

Was

1171

14

26

19.3

Randa

Cin/SD

1210

7

25

20.3

Ramirez

ChC

1020

14

26

20.5

Blalock

Tex

1374

21

24

22.5

Alfonzo

SF

813

22

25

23.3

4 comments:

  1. Interesting aggregation, although I don't necessarily know about a straight averaging, since the different methodologies are not necessarily equal in scope or credibility.

    That said, I took a half-assed hack at something like this back in 2003. I took the three key fielding metrics of the time -- fielding average, range factor, and zone rating -- and averaged them out on different scales depending on the position. Here's the original post:

    http://groups.google.com/group/rec.sport.baseball/browse_frm/thread/e970cca169754184/56154fd2e69aa9eb

    ReplyDelete
  2. I see the URL got truncated. I'll reproduce that entire post here:

    Been thinking about some of the things said around here lately regarding our
    up the middle regulars -- specifically, we know their hitting sucks, but
    we're doing it for the defensive advantage.


    That got me to wondering how true that is and how it could be corroborated.


    The idea I got was to measure how the Tigers are fielding at their
    respective positions versus the rest of MLB.


    The hard part is figuring out how to do it. Basically, there are three
    measurements to consider, and they apply at eight of the nine positions:


    Fielding Percentage (FPCT): Errors divided by Total Chances. A weak
    measurement because it does not take range into account, nor does it account
    for the difference between fielding errors and throwing errors, but it does
    have limited value directionally in determining who has a better glove --
    very limited.


    Range Factor (RF): (PO + A) divided by 9 innings. This is supposed to
    measure a fielder's ability to get to more batted balls, which is more
    important measure of defensive value than raw number of errors or even error
    rate. After all, if Player A has 500 chances and 25 errors (.950 FPCT)
    versus Player B with 400 chances and 10 errors (.975 FPCT) in the same
    number of innings, Player A may have booted 15 more balls -- but
    theoretically, he’s also turned in 85 more outs, too, and that makes A more
    valuable in the field than B, even with a lower FPCT. However, this
    statistic is heavily influenced by the team's pitching staff. Fly ball
    pitching staffs create higher range factors for the outfield because more
    balls are hit there, and conversely lower factors for infielders. Vice
    versa for ground ball staffs. So this number is not foolproof, but does
    represent an upgrade from raw counting stats like errors or the FPCT rate
    stat.


    Zone Rating (ZR): The percentage of balls fielded by a player in his typical
    defensive "zone," as measured by STATS, Inc. This does take range into
    account, and assigns a player responsibility only for those balls deemed to
    have been hit into his area, negating the fly ball staff/ground ball staff
    split. But it does not separate errors from successful chances, treating
    them equally, which leaves this statistic a bit deficient. But this is
    actually a better measure of range factor than actual Range Factor is, I
    think, because it is more team dependent (on pitching staffs, and also on
    whether there are great range players at adjacent positions), while ZR is
    more individually focused (what percent of balls hit into your zone you
    actually convert into chances handled yourself). This is also a subjective
    stat as well as it relies on the judgment of observers working for STATS to
    determine just whose zone a ball fell in, but they use a redundant multiple
    stringer system which helps iron those subjectivities out.


    So in the end, how do we measure? Here's what I'm going to do: I'm going to
    do a rate ranking system. I'm going to rank all the teams in all three of
    the measures, then weight the categories based on their relatively
    importance to each other, than attempt to adjust for position based on
    relatively importance of that category.


    So -- for the P, 2B, SS, and all the OF positions, I've weighted the
    categories as FPCT=1/9, RF=3/9, and ZR=5/9, given the importance of range to
    these positions. For 3B, where range is less of a factor, I've dialed back
    the weighting to 1/5, 2/5, 2/5. For 1B, I’m flipping it around and putting
    more emphasis on fielding percentage (1/2-1/4-1/4), since most of the
    chances are taken on throws, making error percentage much more relevant here
    than range.


    By the way, I'm completely open to suggestions on the relative weighting of
    each category to the others. I’m using a gut feel to assign weights based
    on what I know of these positions – if someone can do a more empirical
    analysis of weights should be applied to each position, please, have at it.
    I’m only here to introduce the concept.


    So enough of the setup -- how do Tiger fielders rank?


    1B: FPCT: 20 (of 30 teams); RF: 3; ZR: 11; Overall: 10th. Carlos Pena has
    been fairly decent here, but his RF is actually lower than the team’s (10.31
    vs. 10.35), suggesting that the backups of Kevin Witt, Craig Paquette, Shane
    Halter, and others have actually been better. Same with ZR (.833 vs. .857),
    and Carlos has committed every 1B error on the team. But Carlos has played
    75% of the innings at 1B, so he drives the overall team ranking. Bottom
    line: 1B has been fairly decent for us defensively.


    2B: 28/10/28 = 24th overall. This is where Ramon Santiago is hurting us –
    he’s a very bad hitter (.233/.298/.270), but he’s also a bad fielding 2B,
    especially in terms of zone rating, which is a killer at this position.
    Even Shane Halter is an upgrade to Ramon as he has put better numbers on the
    board in all three categories in a little over 20% of the innings there.


    SS: 26/1/7 = 5th overall. I know those 11 errors he has look awful bad (and
    he should have had his 12th in the 1st inning yesterday), but fact is, Omar
    Infante is one of the best fielding shortstops in the majors now. His RF is
    far and away the best among regulars (5.56; Omar Vizquel is second at 5.24).
    The question is – how important is his fielding to withstand carrying his
    .199/.261/.236? Yikes.


    3B: 29/20/26 = 26th overall. No other way to put it: Tiger fielding at 3rd
    base has been absolutely awful, and Eric Munson is the main miscreant (in ZR
    and RF, even though Dmitri has a lower FPCT). I know we’ve all seen Eric
    make a play or two and have called it “progress”, but people say the same
    exact thing about Derek Jeter, who’s been among the worst defensive SS in
    the majors for some time now. Luckily for us, 3B is not considered a
    critical defensive position.


    LF: 4/5/8 = 4th overall. LF is been a very good position for the Tigers as
    well, and Craig Monroe is to thank here. He has generated as many chances
    as Dmitri Young (57) in 41 fewer innings (193 vs. 234). Everyone else has
    been fairly pedestrian here.


    CF: 10/16/10 = 12th overall. Pretty middle of the pack here. Torres has
    somewhat better numbers than Kingsale. Alex Sanchez has not been a stiff
    out there, so far, although his reputation in the field can generously be
    described as “bad”.


    RF: 9/8/20 = 16th overall. One of Bobby’s strengths has actually turned
    into somewhat of a liability. He’s getting plenty of chances and he’s
    converting them into outs for the most part, but he’s not covering all the
    ground he could. Bobby is 12th of 18 ZR among all RF who qualify (at least
    he’s ahead of 55-gallon drum imitators Shawn Green and Richard Hidalgo).
    But this is the consequence of getting old.


    P: 10/10/20 = 16th overall. Those rankings read like a 10-10 phone
    commercial. The Tigers get excellent range and sure-handedness out of Mike
    Maroth and Gary Knotts, but fairly pedestrian if not steady work out of the
    others.


    You will notice I have left catchers out of this equation. Catcher is
    probably the most difficult position to which to assign value statistically.
    Zone Rating and Range Factor obviously mean nothing. Raw error counts per
    inning caught therefore probably means a lot more here. Brandon Inge has
    been somewhat OK behind the plate. He has given up 28 steals, fifth most in
    baseball, is below average on CS % (31%), and has withstood the fifth most
    attempts with 41. But then, runners steal more on the pitcher and less on
    the catcher than commonly discussed. He has allowed only one passed ball,
    so his glove seems fairly sure. And he has committed no errors. He may
    have gotten fewer opportunities to make an error than average, but there’s
    no way to tell from the stats at hand. The $64 question, of course, is:
    does his defensive value necessitate carrying his .161/.221/.302 line into
    the lineup every day? (And what the hell is Trammell thinking by batting
    him above Santiago at all? That’s just dumb.) I don’t think anyone is
    surprised that Matt Walbeck is absolutely brutal both behind the plate and
    at bat, and he may in fact be the most harmful player being carried on a
    major
    league roster today.

    ReplyDelete
  3. Thanks Chuck. I like to try to keep things relatively simple on this blog but I agree the straight average is probably not the best route. When I get around to it, I'll probably play around with different weights and see how it changes the final ranking.

    ReplyDelete
  4. Also, thanks for showing me your old post. I figured somebody had tried something like this before but I hadn't seen anything anywhere.

    ReplyDelete

Twitter

Blog Archive

Subscribe

My Sabermetrics Book

My Sabermetrics Book
One of Baseball America's top ten books of 2010

Other Sabermetrics Books

Stat Counter