Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Baseball Newsstand > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Saturday, September 30, 2006

Defensive projections for 2007

I thought I’d get these out here while I have the chance.  You won’t find the 2006 ratings here, but I’m sure Chris Dial will publish something once the season ends.  For the projections, 2 more games isn’t going to change more than a fraction.

Rally Posted: September 30, 2006 at 02:43 PM | 24 comment(s) Login to Bookmark
  Tags: projections, sabermetrics

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. The Keith Law Blog Blah Blah (battlekow) Posted: September 30, 2006 at 05:39 PM (#2192400)
No Prince Fielder?
   2. zfan Posted: September 30, 2006 at 05:42 PM (#2192405)
Great! Pierre +3, Murton +2, Jones -1. Very interesting.

Will you augment this later with other players? I was looking for Ryan Theriot for example; probably not enough AB for your cutoff, but maybe the most AB at 2b for the Cubs this year.
   3. SG Posted: September 30, 2006 at 06:37 PM (#2192481)
Good stuff ARoM. Are you including an aging factor in your projections, or just the weighted average?
   4. Rally Posted: September 30, 2006 at 06:40 PM (#2192486)
Prince Fielder: ZR of .802, run value of -8 for 2006.

He had 217 chances, below my cutoff level of 250 for projections. Because of the regression component, any player who doesn't make the cutoff will have a projection pretty close to zero. That doesn't mean they are average fielders, just means I don't have enough data.

That goes double for Theriot. He's had 57 chances at 2B for a ZR of .842 and Rv of +1.

In his case there's not much stats can do to judge him. You're much better off checking out Tango's scouting report for the fans.
   5. Rally Posted: September 30, 2006 at 06:48 PM (#2192501)
I did factor in an aging component just by subtracting one for everybody. I haven't done the math yet but I've seen MGL post a few times on aging factors and it seems players lose about a run per year on average. I probably should only apply that to players 25 or older or something, but almost everyone here is older than that so I didn't take the time to add in ages.

Will you augment this later with other players?

When the season's over somebody will post ZR for all players, 2006. I don't know if it will be me, SG, or the Dial Lama. I don't plan on doing projections for players under the cutoff though - I just don't see much value in it because of the regression.

For the regression I used 500 chances at league average. Any experts want to chime in? Does that seem the right amount of regression? I used the same figure fro all positions, so a shortstop with one full year's worth of chances will be regressed about 50% to the mean, while a 1B will need about 2 years to be regressed only 50%.
   6. DSG Posted: September 30, 2006 at 08:33 PM (#2192646)
If you send me a spreadsheet of ZR ratings for players with at least 250 chances, I'd be happy to calculate the correct regression amount for you.

It's daviddsg@gmail.com.
   7. Jack Sommers Posted: September 30, 2006 at 09:24 PM (#2192695)
Why Tony Clark, but no Conor Jackson??



Thanks.
   8. Jack Sommers Posted: September 30, 2006 at 09:26 PM (#2192701)
Never mind...I just re read the playting time criteria.
   9. zfan Posted: September 30, 2006 at 11:13 PM (#2192820)
Thanks, ARM. Your cutoff makes sense. Tango doesn't include a lot of players who get limited playing time, too--like Theriot.
   10. mgl Posted: September 30, 2006 at 11:54 PM (#2192848)
Just perusing the list, it looks very similar to UZR projections which basically use the same methodology. 4 years weighted, regerssed and age adjusted. I did my latest research on aging curves yesterday. It does look like players lose around 1+ (per 150) per year (at the same position of course), although it is not clear at what age the decline starts and it probably depends on the position. The decline could start very slowly at age 23 or 24 (just like all of the speed-based metrics), or it could start at around age 28. There actually could be a slight increase prior to age 28 as well. Again, it probably depends on the position. When I try to look at the aging curves position by position, I really get too much noise. About the only position that looks like there is a "learning curve" at a young age, maybe up until age 30 or so, is 1B. Even then though I am not sure. Too much noise, not enough data, even in 7 years of UZR data, which is what I used. Since on the averge, players do lose at least 1 run a year, even including the early ages, Chone would be better off subtracting 2 runs since he is using 4 years of data so some of that data is 3 and 4 years old. Even if there is little decline in the early ages, not too many major league players are that young anyway. If he wanted to refine it a little, without getting too complicated, deducting .5 runs per season used (2 runs for players with at least 4 prior seasons) would be a good idea. Perhaps even no runs for players less than 28, .5 run per season used for players from 28 to 33, and a run a season after that. Nah, on second thought, just leave it the way it is!

UZR needs around 350 "chances" (a "chance" is merely the number of balls an average fielder fields) to regress 50%, so 500 for ZR sounds good to me.

All in all, looks like a sound methodology and great work. If any of around 25 to 28 teams I would guess would just go to this web site, they could save themselves millions of dollars each. Easily. That is pitiful, isn't it?

Chone, you should have held out for the big bucks!
   11. mgl Posted: October 01, 2006 at 12:00 AM (#2192858)
The regression amount is simply 1-r (when you regress one sample of ZR on another) for any number of chances (in this case, samples of players with a certain min number of chances).

Since your are using a range of chances the "r" for the average number of chances in your sample will be a little too low (not too, high, right?), so your regression will be a little lower than the 1-r that you get from the regrssion.
   12. mgl Posted: October 01, 2006 at 12:11 AM (#2192867)
Here are some comparisons with UZR. Remember that the database is the same (STATS Inc.). You would expect UZR to have a larger spread in values (more granular data), but as the sample sizes increase, they would converge, although UZR will always be a little larger in spread.

Erstad +2.5
Mient +6.5
Sexson -8.4
Giambi -8.2

A. Hill 12.6
Ellis 10.8
J. Castillo -12.6
Cantu -17.6

Inge +3.8
Beltre +8.0
Wigginton -14.0
Teahan -12.3

Everett +22.8 (Jeter -12.6) (Try convincing anyone IN baseball that they are equal value players, which they are!)
Crosby 1.0
F. Lopez -9.1
H. Ramirez -5.2

I am a believer! Geez, I toil through 15 years of hard work, and this guy just fires these off - and for free! As I used to tell my kids, it is not how hard you work, but how smart you work!
   13. Cowboy Popup Posted: October 01, 2006 at 12:19 AM (#2192876)
"(Try convincing anyone IN baseball that they are equal value players, which they are!)"

Jeter's RARP: 67.8
Everett's RARP: 3.9

Jeter's BRAA: 39
Everett's BRAA: -22

Jeter's RAP: 49.5
Everett's RAP: -13.6

No they aren't, Everett would need to be 60 runs better on defense, which even your numbers don't say he is. And Everett is a hell of alot less likely to repeat his defensive performance this year then Jeter is to repeat his offensive performance this year.
   14. 1k5v3L Posted: October 01, 2006 at 12:23 AM (#2192879)
And Everett is a hell of alot less likely to repeat his defensive performance this year then Jeter is to repeat his offensive performance this year.


How do you figure?

MGL, shouldn't you be stating that Chone's numbers are completely off, even if they are good?
   15. Kyle S Posted: October 01, 2006 at 12:41 AM (#2192895)
I thought MGL has Jeter at like -15, Everett at +44. That would make them almost exactly even.
   16. mgl Posted: October 01, 2006 at 01:41 AM (#2192944)
And Everett is a hell of alot less likely to repeat his defensive performance this year then Jeter is to repeat his offensive performance this year.

There is nothing to "repeat." These are projections. If you don't "believe" the (defensive) projections, fine. I have no problem with that.

But you are right, I did not realize how bad a hitter Everett was until I looked at his offensive lwts for the last few years. He is downright awful. He HAS to have an amazing glove just to play in the majors.

There is probably a 50-60 run difference in their offensive projections, so even a 35 runs difference in defense does not not nearly make them close overall.

So I take back what I said.

MGL, shouldn't you be stating that Chone's numbers are completely off, even if they are good?

What do you mean? His numbers are excellent. Are you being sarcastic? A sample of my numbers (his two best and worst at the IF positions) are listed above. Look at them next to his:


Erstad +2.5, +4
Mient +6.5, +3
Sexson -8.4, -5
Giambi -8.2, -6

A. Hill 12.6, +11
Ellis 10.8, +12
J. Castillo -12.6, -12
Cantu -17.6, -15

Inge +3.8, +8
Beltre +8.0, +8
Wigginton -14.0, -12
Teahan -12.3, -15

Everett +22.8, +13
Crosby 1.0, +8
F. Lopez -9.1, -9
H. Ramirez -5.2, -9
   17. Greg Maddux School of Reflexive Profanity Posted: October 01, 2006 at 01:45 AM (#2192945)
MGL, shouldn't you be stating that Chone's numbers are completely off, even if they are good?

What do you mean? His numbers are excellent. Are you being sarcastic?



He's saying you should run down competing metrics rather than praise them. It's like you're BMW saying, "Oh, yeah, Mercedes makes a terrific automobile."
   18. 1k5v3L Posted: October 01, 2006 at 01:47 AM (#2192946)
I meant that, since you're selling your numbers to MLB teams, while Chone has them out there for free, you should engage in hard core disinformation. After all, as you just pointed out, "If any of around 25 to 28 teams I would guess would just go to this web site, they could save themselves millions of dollars each. Easily." Some of those millions should go to you, no?

And yes, I'm kidding.
   19. mgl Posted: October 01, 2006 at 06:09 AM (#2193145)
Oh, O.K. I don't sell my numbers to any teams. I consulted with the Cards for 2 years. It was fun. They paid me some decent money and had access to all my data. I am not a commercial person by any means. All of the profits from my book (The Book) I donated to retrosheet. Etc. I would not consult for another team unless they had a complete sabermetric team in place, from top to bottom (well at least from the manager to the GM), sort of like OAK and BOS, I think.
   20. Rally Posted: October 01, 2006 at 01:37 PM (#2193241)
Thanks for the compliments, MGL.

I don't think I could justify charging, though. I spent a few hours crunching numbers, but the real work is by the STATS people who complile it, and I'm just glad its available for free on ESPN and CNNSI.
   21. mgl Posted: October 01, 2006 at 05:42 PM (#2193411)
IK, I was just kidding about that.
   22. 1k5v3L Posted: October 01, 2006 at 05:46 PM (#2193418)
I would not consult for another team unless they had a complete sabermetric team in place, from top to bottom (well at least from the manager to the GM), sort of like OAK and BOS, I think.


Just curious, MGL, but would you add the Josh Byrnes-led Dbacks to this list?
   23. Шĥy Posted: October 01, 2006 at 06:19 PM (#2193477)
Since you are no longer consulting, will you be posting all UZR and SLWTS results now?
   24. Rally Posted: October 03, 2006 at 12:25 AM (#2195014)
I've just updated the page with regression based on 300 chances instead of 500. Thanks to David Gassko for crunching the numbers and suggesting that.

Everett's now at +15, so is Aaron Hill.

Devil Rays have some awful defense with Cantu and Wigginton (-18 each).

No wonder they can't develop a pitcher unless he can strike out 9+ per game like Kazmir. BJ Upton isn't looking like he'll be any better, though perhaps they could try him at 2B since most of his trouble seems to be his throwing.

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Dynasty League Baseball

Support BBTF

donate

Thanks to
Sheer Tim Foli
for his generous support.

You must be logged in to view your Bookmarks.

Hot Topics

NewsblogMLB Teams Spend Record $2.16 Billion in Month Before Lockout
(1 - 11:11pm, Dec 03)
Last: DL from MN

NewsblogOT - November* 2021 College Football thread
(430 - 11:05pm, Dec 03)
Last: frannyzoo

NewsblogNBA 2021-2022 Season Thread
(1261 - 10:52pm, Dec 03)
Last: spivey 2

Sox TherapyLocked Out and Semi-Loaded
(16 - 9:17pm, Dec 03)
Last: The Duke

Newsblog'Chicago!' Stroman says he's joining Cubs
(50 - 9:16pm, Dec 03)
Last: The Duke

NewsblogJackie Bradley, Jr. back to Boston in deal with Crew
(20 - 8:53pm, Dec 03)
Last: NattyBoh

NewsblogFormer Cy Young Winner LaMarr Hoyt Reportedly Dies At 66
(69 - 8:41pm, Dec 03)
Last: Jay Z

NewsblogMLB, union stopped blood testing for HGH due to pandemic
(27 - 8:33pm, Dec 03)
Last: sunday silence (again)

NewsblogMasahiro Tanaka to remain with NPB's Tohoku Rakuten Golden Eagles for 2022 season
(2 - 8:32pm, Dec 03)
Last: vortex of dissipation

NewsblogMcCaffery: Jimmy Rollins, Ryan Howard passed Hall of Fame eye test
(79 - 8:30pm, Dec 03)
Last: cardsfanboy

NewsblogA’s reportedly eyeing Tropicana site for possible Strip ballpark
(26 - 8:23pm, Dec 03)
Last: Walt Davis

Hall of MeritMock Hall of Fame Ballot 2022
(16 - 8:01pm, Dec 03)
Last: LargeBill

NewsblogMarcell Ozuna was choking wife as cops burst in, police video shows
(39 - 5:30pm, Dec 03)
Last: Never Give an Inge (Dave)

NewsblogBaseball Hall of Fame ballot 2022: Alex Rodriguez, David Ortiz join; Bonds, Clemens, Schilling in final year
(97 - 4:51pm, Dec 03)
Last: taxandbeerguy

Hall of MeritMost Meritorious Player: 2021 Results
(6 - 4:24pm, Dec 03)
Last: DL from MN

Page rendered in 0.2181 seconds
48 querie(s) executed