or: How I Learned to Stop Worrying and Love Advanced Defensive Metrics
We know that measuring defense is a better method of evaluation than relying on group opinion, and that data is a more reliable source of information than observation. But the data we have, and the tools at our disposal, aren't doing the job. They're flawed, and what we really need is a defensive statistic that does the same thing a good offensive statistic does: measure success in the context of opportunity.
Until we have that--and I believe we will in this decade--we're all best advised to avoid relying on any one way of measuring defense.
The quote above is taken from a June 2001 article by Baseball Prospectus author Joe Sheehan; the tools being examined are Zone Rating and Range Factor. In the two and a half years since Sheehan's article, better metrics have been developed and adopted. At the moment, advanced defensive metrics (ADM) like Clay Davenport's Rate2, Bill James's Win Shares, Mitchel Lichtman's Ultime Zone Rating, and David Pinto's Probabilistic Model of Range are where to go when defense is on your mind.
Each system takes objective fielding data and in its own way calculates a player's defensive value, attempting to "measure success in the context of opportunity." (For more on the particulars of each ADM, consult the list of links compiled by Sylvain Cognet at James Fraser's Sabermetrics Page.)
While defensive metrics have come a long way since Sheehan's article, his conclusion from a few summers past still holds: "we're all best advised to avoid relying on any one way of measuring defense." Words of wisdom if I've ever heard them. A problem occurs, however, if you take Sheehan's words to heart. Let's say you're excited by the prospect of Mike Cameron roaming the outfield in Flushing. You click on a few handy links and gather the following information:
| Rate2: | 117 |
| Win Shares: | 7.73 Fielding WS |
| UZR: | 31 UZR Runs |
| Prob. Model: | .00424 DER diff. |
Each system expresses itself in a unique way, and for all intents and purposes, you can't really do much of anything with these numbers alone. However, if there was some way to equalize the results, we'd be in business. So let's do that.
UZR
Not much to do here. UZR is already expressed in terms of runs saved or cost above an average fielder, and Mitchel Lichtman provides all the data in one spreadsheet. For the purpose of this article, I simply used UZR Runs per 162 games. In Mike Cameron's case, this is the same as his total UZR Runs: 31.
Rate2
Clay Davenport's metric, used by Baseball Prospectus, is defined as:
A way to look at the fielder's rate of production, equal to 100 plus the number of runs above or below average this fielder is per 100 games. A player with a rate of 110 is 10 runs above average per 100 games, a player with an 87 is 13 runs below average per 100 games, etc.
Since the UZR rate I'm using is expressed in terms of 162 games played, a few simple calculations need to be made. Multiply the amount above/below 100 (17) by 1.62 (162 games divided by 100 games = 1.62). Thus:
17 * 1.62 = 27.54.
Rounding to a single number, Mike Cameron saved 28 runs above average as measured by Rate2.
Win Shares
Dave Studemund, who runs Baseball Graphs.com, provides fielding win shares ranked by position in each league. He also provides fielding win shares per 1000 innings, which is exactly what it says it is. Here's what you do to turn FWS per 1000 INN into runs above/below average:
1. Calculate fielding win shares per 162 games. Since 162 games take about 1444 innings, multiply Cameron's WS per 1000 INN by 1.444:
6.02 * 1.444 = 8.69 FWS per 162 G
2. Calculate fielding win shares per 162 games for each position. Dave provided position averages expressed in terms of FWS per 1000 INN, which I multiplied by 1.444:
Position FWS/1000 INN FWS per 162 G 1B 1.67 2.41 2B 4.45 6.43 3B 3.29 4.75 SS 5.00 7.22 OF 2.59 3.74 C 5.20 7.51
3. Calculate fielding win shares above/below average per 162 games:
8.69 - 3.74 = 4.95
4. Calculate wins above/below average per 162 games. Three win shares equal one win, therefore:
4.95 / 3 = 1.65
5. Calculate runs above/below average per 162 games. I think it's safe to use 10 runs as equalling 1 win:
1.65 * 10 = 16.5
Again, rounding to a single number, Mike Cameron saved 17 runs above average as measured by Win Shares.
Probabilistic Model
For each player, David Pinto's model provides Expected Outs, as well as Actual Outs. For example, Mike Cameron was expected to register 467.5 outs. He made 484. The difference, 16.5 outs (484 - 467.5), is the first calculation that needs to be made. Next, this number needs to be converted into runs. Using Tangotiger's explanation, I went with 0.8 runs per out above/below:
16.5 * 0.8 = 13.2
Cameron saved 13.2 runs in 2003. However, to use this number against the other metrics, it needs to be expressed in terms of 162 games played. Divide 1444 by the number of innings played, and then multiply by the number of runs cost or saved:
(1444 / 1284) * 13.2 = 14.84
Cameron saved 15 more than an average fielder per 162 games. With this final calculation, we now have the following information about Mike Cameron's 2003 defensive value:
| Rate2: | 28 |
| Win Shares: | 17 |
| UZR: | 31 |
| Prob. Model: | 15 |
That's more like it. Each number represents the amount of runs above or below an average fielder per 162 games. While it's dangerous to use any one ADM to determine a player's defensive value, using all four might be the best course of action. Mike Cameron is an easy one. They all say he's very good, and the largest difference between the highest and lowest rating is 16 runs. Things get trickier with other players and positions.
Below you will find the top thirty players at each position (except catcher), as measured by games started, ranked by what I call their Cedeno Number, found by adding the four equalized ADM numbers and dividing by four. In Cameron's case: [(28 + 17 + 31 + 15) / 4] = 22.
It's obviously named after every Met fans favorite outfielder, Roger Cedeno, who inspired this article. I was searching for any system that proved he's as bad as he looks out there. To my surprise, they all say he's bad, just not that bad. With every Mets fan refusing to believe Cedeno is anything less than the worst right fielder in the history of organized baseball, I thought I would try to stick up for Cedeno. This article is the result of that labor.
SHORTSTOP
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Jose Valentin, CWS | 8 | 8 | 34 | 14 | 16 |
| Julio Lugo, Hou/TB | 11 | 1 | 15 | 24 | 13 |
| Alex S. Gonzalez, ChC | 23 | 11 | 14 | 3 | 13 |
| Adam Everett, Hou | 10 | 0 | 18 | 20 | 12 |
| David Eckstein, Ana | 16 | 0 | 22 | 8 | 12 |
| Juan Uribe, Col | 2 | 4 | 12 | 27 | 11 |
| Angel Berroa, KC | 5 | -3 | 10 | 20 | 8 |
| Rey Sanchez, Sea/NYM | 11 | 7 | 4 | 8 | 8 |
| Orlando Cabrera, Mon | 6 | -1 | 8 | 16 | 8 |
| Alex Gonzalez, Fla | 6 | 10 | 7 | 4 | 7 |
| Cesar Izturis, LA | 8 | 9 | 0 | -2 | 4 |
| Jimmy Rollins, Phi | 2 | 4 | 5 | 3 | 3 |
| Chris Woodward, Tor | -2 | 1 | 0 | 14 | 3 |
| Royce Clayton, Mil | 0 | 2 | 7 | 4 | 3 |
| Carlos Guillen, Sea | 3 | 5 | -6 | 8 | 3 |
| Alex Rodriguez, Tex | 5 | -1 | 5 | -3 | 1 |
| Jose Reyes, NYM | 16 | 6 | -13 | -10 | 0 |
| Edgar Renteria, StL | -3 | -6 | 10 | -2 | -1 |
| Nomar Garciaparra, Bos | -8 | -3 | 3 | 2 | -1 |
| Rafael Furcal, Atl | -16 | -7 | 7 | 9 | -2 |
| Cristian Guzman, Min | -3 | 3 | -15 | -6 | -5 |
| Deivi Cruz, Bal | -3 | -3 | 0 | -14 | -5 |
| Jack Wilson, Pit | 3 | -5 | -14 | -6 | -5 |
| Miguel Tejada, Oak | -10 | -1 | -4 | -12 | -7 |
| Alex Cintron, Ari | -5 | -5 | -18 | -5 | -8 |
| Jhonny Peralta, Cle | 6 | -1 | -14 | -25 | -8 |
| Rick Aurilia, SF | -3 | 1 | -15 | -17 | -9 |
| Ramon Santiago, Det | 2 | -5 | -22 | -18 | -11 |
| Ramon Vazquez, SD | -21 | -11 | -26 | -18 | -19 |
| Derek Jeter, NYY | -31 | -18 | -40 | -14 | -26 |
SECOND BASE
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Orlando Hudson, Tor | 32 | 20 | 20 | 12 | 21 |
| Placido Polanco, Phi | 11 | 6 | 49 | 10 | 19 |
| Mark Ellis, Oak | 11 | 11 | 26 | 27 | 19 |
| Adam Kennedy, Ana | 19 | 1 | 27 | 22 | 17 |
| Brandon Phillips, Cle | 13 | 1 | 11 | 22 | 12 |
| Mark Grudzielanek, ChC | 3 | 5 | 26 | 9 | 11 |
| Brian Roberts, Bal | 5 | 2 | 22 | 6 | 9 |
| Jeff Reboulet, Pit | 8 | 1 | 7 | 17 | 8 |
| Marcus Giles, Atl | 6 | 5 | 2 | 14 | 7 |
| Luis Castillo, Fla | 8 | 4 | 0 | 13 | 6 |
| Junior Spivey, Ari | 6 | 5 | 14 | -1 | 6 |
| Ray Durham, SF | 19 | 11 | -2 | -8 | 5 |
| Bret Boone, Sea | 13 | 7 | 10 | -13 | 4 |
| Abraham O. Nunez, Pit | 3 | 0 | -7 | 20 | 4 |
| Alex Cora, LA | 13 | 12 | -11 | -6 | 2 |
| Mark Loretta, SD | -3 | -2 | 4 | 7 | 1 |
| Alfonso Soriano, NYY | 0 | 2 | 5 | -3 | 1 |
| Warren Morris, Det | 3 | -5 | -4 | 9 | 1 |
| Jeff Kent, Hou | 0 | -2 | 10 | -8 | 0 |
| D'Angelo Jimenez, CWS/Cin | 2 | -6 | -1 | -9 | -4 |
| Desi Relaford, KC | -5 | -6 | -15 | 6 | -5 |
| Jose Vidro, Mon | -15 | -5 | 7 | -10 | -5 |
| Bo Hart, StL | -10 | -8 | -7 | 2 | -6 |
| Marlon Anderson, TB | -15 | -14 | 6 | -7 | -7 |
| Eric Young, Mil/SF | -6 | -9 | -8 | -7 | -8 |
| Michael Young, Tex | -13 | -3 | -14 | -6 | -9 |
| Luis Rivas, Min | -10 | -4 | -25 | -19 | -14 |
| Roberto Alomar, CWS/NYM | -10 | 0 | -27 | -26 | -15 |
| Ronnie Belliard, Col | -3 | -9 | -19 | -34 | -16 |
| Todd Walker, Bos | -26 | -9 | -24 | -15 | -19 |
THIRD BASE
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Damian Rolls, TB | 18 | 3 | 27 | 30 | 19 |
| David Bell, Phi | 11 | 5 | 27 | 29 | 18 |
| Morgan Ensberg, Hou | 16 | 2 | 5 | 23 | 12 |
| Eric Chavez, Oak | 21 | 7 | 12 | 0 | 10 |
| Jose Hernandez, ChC/Pit/Col | 15 | 4 | 6 | 13 | 9 |
| Geoff Blum, Hou | 18 | -1 | 5 | 13 | 9 |
| Robin Ventura, NYY/LA | 3 | 9 | 20 | 3 | 9 |
| Adrian Beltre, LA | 5 | 2 | 22 | 5 | 8 |
| Casey Blake, Cle | 2 | -3 | 9 | 23 | 8 |
| Aaron Boone, NYY/Cin | 15 | 8 | 1 | -1 | 6 |
| Chris Stynes, Col | 3 | -4 | 12 | 5 | 4 |
| Tony Batista, Bal | 5 | 1 | 8 | -1 | 3 |
| Scott Rolen, StL | 5 | -1 | 3 | 5 | 3 |
| Sean Burroughs, SD | -2 | 0 | 13 | -1 | 2 |
| Corey Koskie, Min | 6 | 7 | 7 | -11 | 2 |
| Mike Lowell, Fla | 0 | 2 | -1 | 4 | 1 |
| Joe Crede, CWS | 3 | 1 | 1 | 0 | 1 |
| Vinny Castilla, Atl | -6 | -3 | 4 | 10 | 1 |
| Todd Zeile, NYY/Mon | 6 | 4 | -3 | -4 | 1 |
| Joe Randa, KC | -3 | -1 | 4 | -5 | -1 |
| Bill Mueller, Bos | 5 | 3 | -9 | -7 | -2 |
| Edgardo Alfonzo, SF | -13 | -2 | -3 | 5 | -3 |
| Hank Blalock, Tex | -5 | -3 | -8 | 2 | -4 |
| Jeff Cirillo, Sea | -5 | -1 | -19 | 10 | -4 |
| Aramis Ramirez, ChC/Pit | -5 | -1 | -3 | -6 | -4 |
| Wes Helms, Mil | -18 | -6 | -4 | -5 | -8 |
| Troy Glaus, Ana | -11 | -5 | -15 | -7 | -9 |
| Ty Wigginton, NYM | -2 | -3 | -24 | -17 | -11 |
| Eric Hinske, Tor | -23 | -6 | -12 | -17 | -14 |
| Shea Hillenbrand, Bos/Ari | -3 | 0 | -27 | -28 | -15 |
| Eric Munson, Det | -23 | -9 | -22 | -38 | -23 |
FIRST BASE
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Travis Lee, TB | 19 | 1 | 24 | 28 | 18 |
| Todd Helton, Col | 29 | 4 | 24 | 2 | 15 |
| Lyle Overbay, Ari | 32 | 7 | 10 | 8 | 14 |
| Doug Mientkiewicz, Min | 5 | 2 | 20 | 17 | 11 |
| Derrek Lee, Fla | 8 | 5 | 16 | 11 | 10 |
| J.T. Snow, SF | 6 | 2 | 5 | 20 | 8 |
| Kevin Millar, Bos | 11 | 2 | 3 | 12 | 7 |
| Ken Harvey, KC | 16 | 1 | 10 | -4 | 6 |
| Tino Martinez, StL | 10 | 0 | 11 | 2 | 6 |
| Sean Casey, Cin | 0 | -3 | 19 | 2 | 5 |
| Richie Sexson, Mil | 21 | 3 | -8 | -2 | 3 |
| John Olerud, Sea | -2 | 2 | 3 | 8 | 3 |
| Jeff Conine, Bal/Fla | 0 | -1 | 3 | 7 | 2 |
| Mark Teixeira, Tex | -3 | 1 | -8 | 14 | 1 |
| Ryan Klesko, SD | 3 | -1 | 2 | -2 | 1 |
| Carlos Delgado, Tor | 6 | 1 | -4 | -3 | 0 |
| Scott Spiezio, Ana | 0 | -2 | 4 | -2 | 0 |
| Paul Konerko, CWS | -2 | 2 | -7 | 2 | -1 |
| Shea Hillenbrand, Bos/Ari | 16 | 4 | -12 | -13 | -1 |
| Ben Broussard, Cle | -5 | -2 | 5 | -5 | -2 |
| Jeff Bagwell, Hou | -13 | -1 | 7 | -5 | -3 |
| Randall Simon, ChC/Pit | -3 | -4 | 0 | -7 | -3 |
| Robert Fick, Atl | -11 | -4 | 0 | -3 | -5 |
| Jason Giambi, NYY | 3 | 3 | -24 | -5 | -6 |
| Wil Cordero, Mon | 3 | -2 | -20 | -8 | -7 |
| Jason Phillips, NYM | -19 | -4 | -3 | -1 | -7 |
| Jim Thome, Phi | -5 | 0 | -19 | -5 | -7 |
| Carlos Pena, Det | -19 | -5 | -4 | -9 | -9 |
| Eric Karros, ChC | -6 | 0 | -36 | -7 | -12 |
| Scott Hatteberg, Oak | -15 | -1 | -19 | -22 | -14 |
| Fred McGriff, LA | -31 | -1 | -13 | -19 | -16 |
CENTER FIELD
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Mike Cameron, Sea | 28 | 17 | 31 | 15 | 22 |
| Mark Kotsay, SD | 16 | 3 | 41 | 19 | 20 |
| Andruw Jones, Atl | 15 | 9 | 20 | 24 | 17 |
| Dave Roberts, LA | 8 | 5 | 32 | 7 | 13 |
| Carlos Beltran, KC | 6 | 15 | 8 | 20 | 12 |
| Darin Erstad, Ana | -5 | 10 | 25 | 12 | 11 |
| Milton Bradley, Cle | -2 | 5 | 18 | 16 | 9 |
| Jim Edmunds, StL | 18 | 5 | 7 | 6 | 9 |
| Endy Chavez, Mon | 2 | 9 | 22 | -1 | 8 |
| Torii Hunter, Min | 2 | 10 | 8 | 12 | 8 |
| Johnny Damon, Bos | 3 | 7 | 14 | 15 | 7 |
| Alex Sanchez, Det/Mil | -2 | 0 | 8 | 9 | 4 |
| Kenny Lofton, ChC/Pit | 5 | 6 | 5 | -1 | 4 |
| Gary Matthews Jr., Bal/SD | 0 | -4 | 7 | 7 | 3 |
| Luis Matos, Bal | 5 | 1 | 4 | -1 | 2 |
| Vernon Wells, Tor | -5 | 3 | 8 | -3 | 1 |
| Juan Pierre, Fla | -2 | 1 | -1 | -1 | -1 |
| Marlon Byrd, Phi | 0 | 4 | 0 | -7 | -1 |
| Corey Patterson, ChC | -3 | 0 | 3 | -4 | -1 |
| Scott Podsednik, Mil | -8 | -1 | -6 | 7 | -2 |
| Doug Glanville, Tex/ChC | 2 | 2 | -14 | -3 | -3 |
| Chris Singleton, Oak | -8 | 4 | -12 | 2 | -4 |
| Rocco Baldelli, TB | 6 | 7 | -26 | -9 | -5 |
| Marquis Grissom, SF | -6 | 5 | -11 | -10 | -5 |
| Preston Wilson, Col | -16 | -1 | -5 | 0 | -5 |
| Eric Byrnes, Oak | -3 | 5 | -23 | -2 | -6 |
| Carl Everett, CWS/Tex | -3 | -1 | -18 | -6 | -7 |
| Craig Biggio, Hou | -10 | 8 | -23 | -3 | -7 |
| Steve Finley, Ari | -6 | 8 | -53 | -13 | -16 |
| Bernie Williams, NYY | -6 | 1 | -48 | -24 | -19 |
RIGHT FIELD
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Jose Cruz, SF | 26 | 6 | 22 | 16 | 18 |
| Richard Hidalgo, Hou | 28 | 8 | 18 | 8 | 15 |
| Ichiro Suzuki, Sea | 16 | 2 | 16 | 8 | 11 |
| Magglio Ordonez, CWS | 13 | -1 | 9 | 9 | 8 |
| Dustan Mohr, Min | 3 | 1 | 0 | 23 | 7 |
| Terrence Long, Oak | 10 | 1 | 16 | 0 | 7 |
| Aaron Guiel, KC | 13 | 4 | 24 | -15 | 7 |
| Juan Encarnacion, Fla | 0 | -2 | 19 | 4 | 5 |
| Jose Guillen, Oak/Cin | -3 | -4 | 16 | 6 | 4 |
| Jody Gerut, Cle | 8 | -1 | 0 | 3 | 3 |
| Vladimir Guerrero, Mon | 0 | 2 | -1 | 4 | 1 |
| Raul Mondesi, NYY/Ari | -6 | -1 | -2 | 11 | 0 |
| Bobby Higginson, Det | 2 | -7 | 6 | -1 | 0 |
| Jermaine Dye, Oak | 3 | 2 | -3 | -5 | -1 |
| Sammy Sosa, ChC | -6 | -4 | 3 | 2 | -1 |
| Trot Nixon, Bos | -5 | -2 | 4 | -4 | -2 |
| Reggie Sanders, Pit | 5 | -1 | -12 | 0 | -2 |
| Bobby Abreau, Phi | -3 | -3 | -5 | 0 | -3 |
| Bobby Kielty, Min/Tor | -8 | -3 | -6 | 2 | -4 |
| Roger Cedeno, NYM | -8 | -3 | -6 | -4 | -5 |
| Xavier Nady, SD | -5 | -5 | -7 | -5 | -6 |
| Gary Sheffield, Atl | 3 | -1 | -10 | -16 | -6 |
| John Vander Wal, Mil | -8 | -4 | -21 | -2 | -9 |
| Larry Walker, Col | -10 | -4 | -16 | -7 | -9 |
| Shawn Green, LA | 6 | -1 | -20 | -23 | -10 |
| Juan Gonzalez, Tex | 13 | -3 | -26 | -25 | -10 |
| Aubrey Huff, TB | -16 | -4 | -16 | -12 | -12 |
| Jay Gibbons, Bal | -10 | -6 | -17 | -19 | -13 |
| Reed Johnson, Tor | -19 | -4 | -19 | -9 | -13 |
| Tim Salmon, Ana | -15 | -3 | -43 | -24 | -21 |
LEFT FIELD
| Player, Team(s) | Rate2 | Win Shares | UZR | Pinto | Cedeno Number |
| Brad Wilkerson, Mon | 18 | 3 | 15 | 12 | 12 |
| Carl Crawford, TB | 10 | 0 | 10 | 13 | 8 |
| Jay Payton, Col | 13 | -2 | 15 | 4 | 7 |
| Garret Anderson, Ana | 23 | 2 | -3 | 5 | 7 |
| Luis Gonzalez, Ari | 10 | 2 | 11 | 2 | 6 |
| Randy Winn, Sea | -2 | 2 | 18 | 2 | 5 |
| Lance Berkman, Hou | 0 | 0 | 12 | 4 | 4 |
| Geoff Jenkins, Mil | 0 | -5 | 14 | 3 | 3 |
| Barry Bonds, SF | 8 | 2 | -8 | 8 | 3 |
| Rondell White, KC/SD | 2 | -3 | 12 | -1 | 2 |
| Carlos Lee, CWS | 0 | -2 | 13 | -1 | 2 |
| Albert Pujols, StL | -3 | -4 | 6 | 7 | 2 |
| Raul Ibanez, KC | -10 | 2 | 13 | 1 | 1 |
| Craig Monroe, Det | 0 | -7 | 12 | -2 | 1 |
| Larry Bigbie, Bal | 5 | -5 | 9 | -11 | 0 |
| Jacque Jones, Min | -8 | 0 | 4 | 2 | -1 |
| Shannon Stewart, Min/Tor | 6 | 1 | -11 | -6 | -2 |
| Brian Giles, Pit/SD | 3 | -1 | -7 | -7 | -3 |
| Moises Alou, ChC | -8 | -5 | 1 | 0 | -3 |
| Cliff Floyd, NYM | 2 | -4 | -16 | 4 | -4 |
| Jeromy Burnitz, LA/NYM | -8 | -4 | -4 | -4 | -5 |
| Adam Dunn, Cin | -6 | -6 | -2 | -9 | -6 |
| Hideki Matsui, NYY | -11 | -4 | -6 | -3 | -6 |
| Pat Burrell, Phi | 2 | -3 | -15 | -7 | -6 |
| Todd Hollandsworth, Fla | 6 | -3 | -21 | -12 | -7 |
| Chipper Jones, Atl | -13 | -5 | -12 | -4 | -8 |
| Dmitri Young, Det | 0 | -6 | -13 | -22 | -10 |
| Matt Lawton, Cle | 0 | -3 | -17 | -27 | -12 |
| Manny Ramirez, Bos | -10 | -2 | -28 | -12 | -13 |
| Terrence Long, Oak | 0 | 1 | -38 | -19 | -14 |
I am indebted to Dave Studemund, David Pinto, and Tangotiger for providing information I requested, as well as timely feedback. Thank you gentlemen.
UPDATE (11:45 am): Tangotiger noticed the outfield win shares were off the charts, and sure enough, I used the wrong per 162 number in my spreadsheet. I've corrected the formula and the charts above now have the corrected numbers. Like I said, indebted.
UPDATE (12:40 pm): Updated the corner infielders as well. Sorry for the slopiness, but I think it's all straight now.
Posted by Avkash at February 17, 2004 09:29 AMNice stuff. I've never felt Cedeno is horrendous in right field -- just, as you suggest, bad. It's center field he just cannot handle at all, and it was Howe and Phillips's fault he was asked to last season.
Posted by: Sam M at February 17, 2004 10:25 AMAgree with Sam on Cedeno, but one question?
Should average WS rate be for all OF. Shouldn't there be a different rate for CF... it seems to me its harder than RF or LF.
Posted by: Jeremy Heit at February 17, 2004 10:53 AMGreat analysis but you forgot the most important of defensive metrics: "BBoH"
I have a well thought out and impressive critique on my site.
Posted by: Norm at February 17, 2004 11:09 AMNice job, Avkash. To think, the Mariners could have had both Cirillo and Cedeño:
http://noslenblog.blogspot.com/2004_01_04_noslenblog_archive.html#107376909849390182
Posted by: Steve at February 17, 2004 11:14 AMI thought so too, Jeremy. Dave at Baseball Graphs lumps them all together as OF, but I don't think the solution is as easy as splitting them up by LF/CF/RF, because you've got guys (take Hideki Matsui, for just one example) that played a significant amount at more than one position. I don't think you can just chop up the fielding win shares they received weighted by games or innings they played at each position. Hit-deki got 2.49 WS overall, but who can say how much he earned as a CF and how much as a LF? If you just weight them by time spent on the field, you're not accounting for the respective difficulty of the positions.
Then again, I'm an idiot and should never open my mouth about these things. Very cool stuff, Avkash. I love the name.
Posted by: Mike at February 17, 2004 11:23 AMI'm making no judgements about win shares per se, but clearly it is not appropriate to average win shares in with these other metrics, especially in the OF. The scale and baseline are completely different for win shares, even though the author of this article has done an admirable job in converting it to something that looks like runs per 162 games.
Also, you want to be very careful in averaging metrics and taking the results seriously (as being better than any of the individual metrics). In general, in order to do that, you have to have reason to believe that one metric complements another and/or offers something that another one does not. And even then, you want to make sure that they are all on roughly the same qualititative level. If they are not, then even if there is some complementing, the inferior ones will bring down the average to a level worse than that of any of the individual superior individual ones. So you want to make sure that there is a high degree of independence among the individual metrics and that they all use about the same granularity of data.
In this case, besides the fact that I am biased towards UZR, as I said, win shares does not belong in this list, I am not familiar with rate2, and although Pinto's "system" appears to be good and very close to UZR, it is a black box as far as I know.
If I were in charge of the world, I would not be averaging UZR or Pinto's system with anything, any more than I would average lwts with OPS or BA with OBA. IMNSHO, UZR is the gold standard, principally becasue it relies on PBP (hit type and location) data (so does ZR and defensive average, but on a very "gross" level), whereas all the other metrics can only "estimate" such data...
Posted by: MGL at February 17, 2004 11:49 AMgreat stuff, Avkash, but I agree with MGL. In this case, would it be a better idea to use ranks instead of values?
rich
Posted by: rich at February 17, 2004 11:57 AMThanks to everyone on the feedback.
I think the player who inspired the number tells you just how serious to take the results of adding and averaging. Just something I threw in there.
My goal is to present some great data out there in a way that would make it easier for someone who doesn't want to read the finer details at Primate Studies to digest.
Your point about Win Shares is right on, and Tango sent me an email outlining some of his concerns. There was a mistake in one of my sheets, and I've corrected the error and updated the charts.
My favorite stat among the big four is UZR, though I usually refer to Rate2 because it's available in season, as well as through history.
Cool. When I understand it all, I'll get back to you. Still, cool.
Posted by: Bryan at February 17, 2004 12:16 PMVery nice work.
The only comment I would have is that the "averaging" seems to be weighted more to the UZR/Pinto side of things, since leaders in those categories tend to have higher number of runs than in Win Shares. Maybe a MVP style voting or weighted measure for ranking?
And how does Jose Valentine always rate so high in these defensive measurements? Is he really that good at SS?
Posted by: Endymion Keats at February 17, 2004 12:23 PMI'm still curious about why everyone has issues with WS for fielding. I can see the point about players moving around, but assuming that a player has a reasonable number of chances at a given position, what's wrong with using WS/1000IP to prorate to 162 games, just as Avkash did? There may be a problem with it, but I don't see what it is.
Posted by: Ali Nagib at February 17, 2004 04:48 PMWin Shares isn't bad for this analysis. It just isn't as good as those systems based on play by play (pbp) data.
The issue is that pbp data will never be available for all historic ballplayers, so UZR and the Pinto method can't be applied. That's what Win Shares is for.
Since we don't have historic data for outfield positions split out by lf, cf and rf, Win Shares doesn't account for that. It is a flaw, but it's a historic data flaw.
My problem with Rate5 is that I have no idea how it was constructed, and I don't think Davenport is telling. OTOH, Win Shares is completely transparent.
That's one of the reasons it gets abused, because people know what went into it, and they can see its flaws (and there are several).
Posted by: studes at February 17, 2004 05:57 PMThere's quite a bit of estimating that takes place in WS, so I believe it should be discounted. I would suggest a similar table, but instead of averaging, use a weighted average. Personally, I would weight them as UZR - .35, Pinto - .3, Rate2 - .25, WS - .1
There's not a lot of variance in doing this, as most only move +/- 3 runs. However there are a few that move 5 or more. This seemed significant in that it was more than half way win/loss:
Polanco +5
R.Alomar -5
Baldelli -5
Finley -7
B.Williams -6
Salmon -5
Long -5
I agree with MGL's statements above, but thought this might be helpful...
Posted by: Brian P. at February 18, 2004 12:29 PMWell the normalizing factor is that no matter how you slice it Jeter doesn't look good. I can't calim to be an expert on any of these defensive metrics but I am most convincinly swayed by MGL's UZR methodology.
Posted by: steve at February 20, 2004 06:00 PMThe fact that we can argue over four separate defensive metrics is still unbelivable to me. I may disagree with some of the methodology, but this is how we find new methods and metrics, by just playing around and tinkering with what is already out there. Very well done, an admirable job.
Posted by: Matt at February 20, 2004 08:33 PMgreat system, but where are the catchers?
Posted by: Joe at February 29, 2004 04:51 PMHi, I came across this via Google. I am looking to figure out what are the upper lower bounds of Rate2 (or other fielding metrics) for players at each position.
I.e.: At each position, what are the best and worst Rate2 figures which have been recorded by players playing a full season (or something close to it).
I am too lazy to just do a zillion player queries on the BP site and am wondering if the data is out there somewhere in a sortable database or something fun like that.
Cheers
Mike Waltz
PS I should add that I am looking for the best and worst seasons since 1900 at each position.
Posted by: Mike Waltz at April 17, 2004 06:39 PMThe results are totally meaningless. No matter how you slice it and/or dice it, if you use contrived meaningless stats for each factor used and combine then the results will be a conglomeration of contrived meaningless stats. Defensive stats are always ill conceived when man made.
Posted by: joek at May 8, 2004 09:04 AM