How Age Affects Infield Defense Using UZR Component Skills
by Jeff Zimmerman (TucsonRoyal) on May 26, 2009 11:00 AM EDT
17 comments
In a previous article I looked at how the individual UZR components (Arm, Range, Errors) change as outfielders age. I have extended that study to include infield defense.
I looked at players who played the same infield position two years in a row and compared their UZR components, Range, Double Play, Errors and Total UZR (no range calculations for infielders) from the first to second year, weighting each pair by the number of innings played. If you want the complete explanation of all the calculations, please read the Complex Explanation in the outfielders articlep>
Results:
Here are the graphs for the four infield positions and a combination of all four. I am not sure how much stake I would put into the combination data since the hardest (shortstop) an easiest (first base) positions on the diamond are both included as infielders, but will still make it available for people examine. I assumed each player had a 0 (average) value for all attributes at age 22 and then plotted how his numbers would change as he aged:
As with the outfielders, the number of complete seasons worth of data are small for the early 20's and late 30's age brackets. Here is a graph of the number of complete match pair seasons for each age:
Observations:
-
Total UZR decreased between 4 to 8 UZR for all positions across the 12 year testperiod and, as with outfielders, a decrease in range was the leading contributor.
-
Age seems to have more of an effect on 2B and SS than 1B and 3B, mainly because Range is more important at SS and 2B.
-
Double Play rates remain constant over the years, not ever varying by more then 1 run.
-
Error rates also remain relatively constant, except for 2B error rates which start declining at age 31, and 3B error rates wich increase steadily over the whole 12 year period.
I hope that gives everyone a little more insight on how aging affects a player's ability to field.
0 recs |
17 comments
|
Email
Print
Comments
3B
That post-age-29 jump in range for third-basemen is crazy. I wonder if it’s partly a result of aging (but still competent) middle infielders moving down the defensive ladder, while the younger set is loaded down with Ryan Brauns and Jim Thomes who aren’t destined to stick at the position, due to 3B’s status as the pinnacle of the “hulking slugger” defensive spectrum.
by psiogen on May 26, 2009 1:07 PM EDT reply actions 0 recs
With my methodology, players moving over should be a problem.
Since each year pair is matched, the numbers of Young Sluggers and 2B/SS all saw a jumps that year.
I think it is also one of the deals with lack of data. Here are the players that really contributed for the jump from 29 to 30.
Wes Helms -16.3 to 0.9
Brandon Inge 8 to 17.4
Scott Rolen 9.5 to 21.5
Tony Batista -18.2 to 2.1
Mike Lowell -9.8 to -0.2
Adrian Betre -3.9 tp 13.4
Troy Glaus -15.4 to -1.6
All of these played quite a few innings at 3rd from 29 to 30 so they weight pretty heavy in final calculations.
by Jeff Zimmerman (TucsonRoyal) on May 26, 2009 1:33 PM EDT up reply actions 0 recs
Instead of basing these on age... (some loose thoughts)
Would it make more sense to develop general trend lines that are somewhat based on the previous season’s UZR?
These age studies are biased toward players who are able to stay at a position longer. My thought here is that the older guys stick at a position longer because their original UZR levels were so high that it took longer for them to become replaceable. The other half of that is that younger, crappier defenders are moved or replaced quite readily.
Basically, a 30 year old SS with a high UZR will likely stick at SS, while a 22 year old with a below average UZR will probably be moved or replaced. His low UZR also has less room to drop.
There’s also a league bias since crappy fielders can be a DH in the AL, but not the NL.
Thoughts?
by NoNameOnCard on May 26, 2009 2:36 PM EDT reply actions 0 recs
Would it make more sense to develop general trend lines that are somewhat based on the previous season’s UZR?
That is exactly what I did. if you go and read the methodology and all I looked at was the difference from one year to the next. This study is exactly what you describe.
There is league bias, but since I matched the players from one year to the next, this should not be a problem.
by Jeff Zimmerman (TucsonRoyal) on May 26, 2009 2:51 PM EDT up reply actions 0 recs
You misunderstood me.
You found a delta based on the player getting a year older and assigned to that player’s specific age. The thing is that not all players age at the same rate. Some players are still playing strong defense at shortstop at 42, while others have significant decline at 28 or sooner.
My basic point, I think (hence “loose thoughts”), is that I’m not sure a 22-year-old who plays the position for two seasons should really have a significant impact on the 22-year-old age group. If he’s moved/replaced before he turns 24, then his defense was clearly not good enough to stay at that spot on the field. He should affect the low end of the UZR scale rather than the 22-24 year old groups.
Do you get what I’m saying?
by NoNameOnCard on May 26, 2009 5:30 PM EDT up reply actions 0 recs
My explanation sucked.
My response to Sky (below) makes a little more sense, but here’s a piece I left out.
My idea is that this study would create a trend line based not on age but still on getting older. For example, a shortstop with a 6 UZR will be worth ?? UZR next season – regardless of how old he is.
Arguably, a 24-year-old with a 6 UZR is just as valuable defensively as a 42-year-old with a 6 UZR for the season in which they produced that 6. Obviously, you’d expect more of a drop off from the 42-year-old going into the following season, but I think this angle should be explored a bit more.
I have no idea how.
by NoNameOnCard on May 26, 2009 5:52 PM EDT up reply actions 0 recs
I don't see ignoring absolute age a good thing.
One, it’s just going to be basic regression.
Two, and more importantly, we might find that we expect 22 year olds to get better until age 28, and then decline gradually, and finally drop off a cliff around age 38. In fact, that’s similar to what happens for hitters overall. And when you break fielding down by component, it gets really interesting, no?
Three, I do, however, think you have a point with players at the extremes, especially the low end. There’s some sampling issues (a -10 UZR guy probably isn’t really -10, more like -5 given proper regression) but yes, a bad fielder is likely to age differently than a good fielder. And you could involve things like speed and strength, which are somewhat measurable via other stats, like 3B/2B ratio or ISO. Great study idea, and complex.
Beyond the Boxscore // Calling BJ Upton lazy is lazy.
by Sky Kalkman on May 26, 2009 6:16 PM EDT up reply actions 0 recs
Yeah.
As I typed that last sentence, I realized that age mattered.
Perhaps some kind of sliding scale based on prior UZR and age. For example, two 26-year-old SS might have pretty different UZRs. Say one has an 8 and the other has a 4. At 27, assuming both are still at SS, they would probably project to have different changes to their UZR.
Putting something together than handles all of that would be like building a defensive projection system similar to the offensive and pitching models of PECOTA or Bill James.
by NoNameOnCard on May 26, 2009 7:44 PM EDT up reply actions 0 recs
I agree with you about the selective sampling style issue.
Players likely to see large decreases in ability are removed from the study by retiring or moving positions.
I don’t think it’s a large issue, maybe hiding a more extreme decline at old old and maybe the entire slope of the difficult positions. Any ideas how to correct for it?
Beyond the Boxscore // Calling BJ Upton lazy is lazy.
by Sky Kalkman on May 26, 2009 3:56 PM EDT up reply actions 0 recs
I'm not sure.
The sample size would be small, but looking at players like Ken Griffey Jr., Ichiro, Omar Vizquel, Ivan Rodriguez, Derek Jeter, and Scott Rolen would be a place to start – guys who’ve played the same position for a large number of years. The list goes on: Andruw Jones, Rafael Furcal, Chipper Jones, Todd Helton, Derrek Lee, Aramis Ramirez, etc..
Throwing in the data for guys like Jose Reyes at 2B or Albert Pujols in the OF just mucks up the point of the study, in my opinion.
by NoNameOnCard on May 26, 2009 5:35 PM EDT up reply actions 0 recs
Comment to both
I could see the younger ages of the skill positions have fast agers bring the numbers down.
There is just not enough defensive data out there to do the second half, but I might see what I can do. I have all the players in a SQL DB so it would be just running the query to get the number of players. What should be the cut off for number of Innings/ games for the linked seasons. Also, how many seasons should be linked because of right now there are only 7 available seasons.
I really think there is not going to be any real good data until there are ~100 to 200 (wild guess on the exact #) players begin a 12+ year career and retire.
by Jeff Zimmerman (TucsonRoyal) on May 26, 2009 6:14 PM EDT up reply actions 0 recs
Sample size is really the biggest problem.
You could start with an 81 game (a half-season seems to make a lot of sense) cut off and see what it does to the number of pairings. The sample size will shrink, but it might result in a smoother trend line.
by NoNameOnCard on May 26, 2009 7:38 PM EDT up reply actions 0 recs
Just running SS over 2002 to 2008 here are the 5 candiates that played 81 at SS:
“name”
“Edgar Renteria”
“Orlando Cabrera”
“Derek Jeter”
“Jimmy Rollins”
“Miguel Tejada”
If I decrease it to 40 games I get a total of 9:
“name”
“David Eckstein”
“Edgar Renteria”
“Orlando Cabrera”
“Derek Jeter”
“Julio Lugo”
“Jimmy Rollins”
“Jack Wilson”
“Omar Vizquel”
“Miguel Tejada”
I see my Chart of the Week somewhere in here
by Jeff Zimmerman (TucsonRoyal) on May 26, 2009 8:23 PM EDT up reply actions 0 recs
That's 81 games in each of the 7 seasons, right?
What if it only required 5-year stints at a position? Michael Young is an interesting example. He played 3 seasons at 2B, 5 seasons at SS, and is now a 3B. There probably aren’t too many players like him, but I think his 5-year period as an SS would be enough to add to the sample group.
You’d have to run that as 3 different queries, I think – 1 for 2002-2006, 1 for 2003-2007, and 1 for 2004-2008.
by NoNameOnCard on May 26, 2009 9:07 PM EDT up reply actions 0 recs
Increasing it to 5 seasons in the row picks up 7 more players:
Name Seasons
Edgar Renteria 7
Derek Jeter 7
Jimmy Rollins 7
Miguel Tejada 7
Orlando Cabrera 7
David Eckstein 6
Jack Wilson 6
Rafael Furcal 6
Khalil Greene 5
Michael Young 5
Bobby Crosby 5
Royce Clayton 5
by Jeff Zimmerman (TucsonRoyal) on May 27, 2009 12:18 AM EDT up reply actions 0 recs
That's a good number of matched seasons. 64 or so?
I’d be interested in seeing how that looks.
Where’s Omar Vizquel?
by NoNameOnCard on May 27, 2009 1:05 AM EDT up reply actions 0 recs
I used games started and
he had 76 last year and 46 in 2003.
by Jeff Zimmerman (TucsonRoyal) on May 27, 2009 11:43 AM EDT up reply actions 0 recs
Comments For This Post Are Closed
Secondary Sidebar
User Tools
Related FanPosts on Beyond the Box Score
FanPosts
Community blog posts and discussion.
Recommended FanPosts
Recent FanPosts
FanShots
Quick hits of video, photos, quotes, chats, links and lists that you find around the web.
Recent FanShots
BtB on Facebook
Certain photos copyright © 2009 by Associated Press or Getty Images. Any commercial use or distribution without the express written consent of Associated Press and Getty Images is strictly prohibited.
