Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Primate Studies > Discussion
Primate Studies
— Where BTF's Members Investigate the Grand Old Game

Wednesday, November 02, 2005

Tweaking Zone Rating

Sean outlines a simple method for turning a player’s zone rating into runs prevented.

Evaluating Defense

For most of baseball history, it simply was not possible to use statistics to evaluate defense with any accuracy. Much improvement has been made in the last 2 decades.  Range factor, fielding runs, zone rating, win shares, and ultimate zone rating (UZR) are a few.  From time to time a new system will come along and become the new standard, totally changing the ways we think about defense in baseball.

This is not such a system.  I cannot hope to match something like UZR in accuracy.  Instead, what I’ve done is combine data from several sources, all freely available, at least for this year.  Zone rating is the base of the system.  Also included are assists and errors for outfielders, and double plays for middle infielders.  Catchers are rated on stolen bases, caught stealing, passed balls, errors, and wild pitches.


For the infield:


Start with zone rating, which can be found for all players on ESPN’s site.  You’ll need to estimate how many plays were made and how many plays were available; this information is not there.  For plays made, assists are a reasonable proxy for 2nd base, 3rd base, and shortstop.  For 1st basemen, I use (assists/2 + putouts/6).  Divide plays made by zone rating to get plays available.

Compare the player to league average.  Take plays made and subtract (plays available * position average zone rating).  This tells you how many plays were made over an average fielder at the position, multiply by 0.75 to convert into runs.

Double plays:

This is done at the team level first. Here are the steps:

1. Estimate the number of runners the pitchers had on first base: (1B + BB + HBP – balks – SB – CS – PB – WP – pickoffs)
2. Get the number of groundouts.  This total is available in the MLB.com stat section.
3. Estimate expected double plays = (league dp/league runners on 1b) * (groundouts/ average team groundouts)
4. Subtract expected double plays from actual double plays

To apply to individual fielders, give half the credit to the second basemen and half to shortstops.  The run value is 0.75 above a normal out.  Double play runs are allocated to the second basemen and shortstops based on their percentage of innings played, though you could argue for allocating based on actual double plays turned instead.

1st and 3rd basemen as of now have no double play rating.  I have no measure for a 1B ability to prevent errors.


For outfielders:

Use putouts as a measure of plays made; otherwise the steps are the same as infielders for measuring plays save.  I use 0.85 runs per play for the outfield, as more of the hits saved would be expected to go for extra bases.

Assist and errors: Get expected assists and errors by player inning total * league assist rate (or error rate) Compare actual to expected.  Run value used is 1 for assists and 0.5 for errors

.
Catchers:

The statistics used are stolen bases allowed, caught stealing, errors, passed balls, and wild pitches.  If you can find wild pitches by catcher, by all means use it.  Otherwise, use team wild pitches and prorate based on the catcher’s innings played.

Run value:

SB = - 0.19

CS = + 0.44

PB and WP (over or under league average) = + 0.28

Error (over or under league average) = + 0.48

SB attempts (over or under league average) = + 0.086

The last measure compares stolen base attempts to the average stolen base attempts that an average catcher allows in the same number of innings.  This gives some credit to catchers whose reputation keeps runners nailed to first (such as Yadier Molina)  The run value represents the expected double plays from keeping the runner on first (0.1144 in 2005) times the run value of the double play (0.75)


The Results:

Gold Gloves:


C: Yadier Molina + 12, Ivan “Bones” Rodriguez + 11, Mike Matheny +10, Joe Mauer +10


No surprises here


1B: Todd Helton +11, Darin Erstad +10, Mark Teixiera +9


On saving errors, I think Erstad saved a ton from Figgins and the other 3b, few from Kennedy, who rarely makes a bad play anyway.  Orlando Cabrera was saved a few errors as well, but the official scorers are more responsible than Erstad.


2B: Mark Ellis +23, Mark Grudzielanek + 23, 3 tied at +15


3B:  Scott Rolen +12 (only 56 games!), Eric Chavez +9, Freddy Sanchez +7


SS: Jack Wilson +28, Juan Uribe +15, Adam Everett +14


Great SS defense was on display in the world series.  Uribe’s diving grab for the next to last out of the series:  Was that play legal?  Isn’t it foul if the player winds up in the stands?  Still a great play.


LF: Cliff Floyd +16, Kelly Johnson +12, Carl Crawford +10


Every defensive system has at least one result that makes you say WHAT?  Floyd is mine.


CF: Jeremy Reed +20, Vernon Wells +15, Aaron Rowand +14


RF: Alexis Rios +11, Jeff Francoeur +11, Austin Kearns +10


And the trailers:

C: JD Closser –11, Fasano, Kendall, Piazza –8


1B: Richie Sexson –15, Jason Giambi –12


2B:  Bret Boone –18, Rickie Weeks –17, Alfonso Soriano –13


3B:  Mark Teahen –21, Jorge Cantu –17, Hank Blalock –15, Alex Rodriguez –15


Cantu did this in part time, he was also –11 at 2B.  In the field he’s Jorge Cannot.


SS:  Michael Young –22, Russ Adams –18, Edgar Renteria –15


Jeter was near average at –3, that’s 2 years in a row.


LF: Manny Ramirez –25, Hideki Matsui –9, David Delucci –9


That figure includes a park factor for Manny, the only park factor used in these ratings. He’s at –32 without it.


CF: Ken Griffey –25, Jose Cruz –13,  Mark Kotsay –11


Kotsay had some injuries, as for the others, I hate watching bad centerfielders who were drafted in the first round by the Mariners and whose daddies were good players in the 70’s.


RF:  Jason Lane –15, Gary Sheffield –14, Matt Lawton –14


I wouldn’t have guessed Lane, he looked pretty good in the playoffs.  Maybe its just because he dives all the time.


Many of the run values for game events come from Tango Tiger’s site, http://www.tangotiger.net/

Some, such as the value of an outfield assist, are simply my best guess.

For all players with at least 250 innings at a position for 2005, you can download the results from this excel spreadsheet: 2005 Defense.  For catchers, there is 2005 data only.  For infielders and outfielders, I have included a true talent estimation, based on 4 years of data.

Chone Smith Posted: November 02, 2005 at 01:59 PM | 88 comment(s) Login to Bookmark
  Related News:

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. nycfan Posted: November 02, 2005 at 03:40 PM (#1715727)

Boy does Texas have an awful infield. They really need to trade Soriano in the offseason before people catch on that he’s not very good.

   2. AROM Posted: November 02, 2005 at 03:47 PM (#1715738)

Thanks Chris and Dan for getting this posted.

If I were Texas, I’d try to move Blalock and Soriano for pitching to teams impressed by their hitting stats, play Young at 2nd, consider Teixiera as a 3B (he’s played there before, his range is good, though I don’t know about his arm).

Adrain Gonzalez takes over 1st and Ian Kinsler at short.

   3. Matt Clement of Alexandria Posted: November 02, 2005 at 03:48 PM (#1715739)

CHONE!!

   4. Der Komminsk-sar Posted: November 02, 2005 at 03:55 PM (#1715748)

In the field he’s Jorge Cannot.
The year before he figured out how to hit, I called him Jorge Can’tDo.

I like that you’re not trying to claim the superiority of this metric, but I’d like more reasoning as to why you’d advocate using this over one of the zillion other methods of evaluating defense (say, FR). You may also want to get into more of the detail of the limitations of your measure.

   5. Joshemy Posted: November 02, 2005 at 03:59 PM (#1715752)

In regards to Cliff Floyd…

He looked much improved in LF this year by pure observation (and I think most of the Met fans around here would agree with me) and I would say he was above average this year, though best LF is pushing it a lot.

   6. philly Posted: November 02, 2005 at 04:01 PM (#1715759)

Well you’ve got Kevin Millar at -1.  Most systems seem to think his ability to cut off the 2B (while often not making outs) makes him a good defender at 1B.

On the basis of that, I’ll call it a success.

Although it’s not obvious to me why that would be the case given the inputs are mostly the same, no?

   7. Kyle S Posted: November 02, 2005 at 04:04 PM (#1715761)

Kelly Johnson was a pretty crummy left fielder. As DSG’s defensive ratings loved him too, he must have gotten really lucky and had a lot of balls hit right at him to “game” these systems. I assume the run values are “raw” as opposed to per 150?

Very nice work, “Chone.”

   8. AROM Posted: November 02, 2005 at 04:07 PM (#1715769)

On a team level (using DER converted into runs), this method correlates with an r of .60

I’m not really sure if that’s better or worse than other methods.  I’m not advocating using this over anything else.  I would suggest using different methods to complement each other, as long as they have some degree of accurracy at a team level.

Something like FR (I’m assuming you mean the one from Hidden game of baseball over 20 years ago) doesn’t hold up on a team level.

   9. SG Posted: November 02, 2005 at 04:12 PM (#1715775)

Interesting stuff Chris.  How do you calculate average zone rating?  Are you multiplying Total Chances times ZR for each player and summing that, then dividing by league total chances?

   10. AROM Posted: November 02, 2005 at 04:15 PM (#1715784)

SG, that’s exactly right. 

I’m the author, I chose to go with Chone as a pen name since there is another Sean Smith and Angel fan who has posted here.

   11. Mike Emeigh Posted: November 02, 2005 at 04:19 PM (#1715790)

Was that play legal? Isn’t it foul if the player winds up in the stands?

Defintion of a catch:

A fielder may reach over a fence, railing, rope or other line of demarcation to make a catch. He may jump on top of a railing, or canvas that may be in foul ground.

Rule 7.04(c) states:

If the fielder or catcher, after having made a legal catch, should fall into a stand or among spectators or into the dugout after making a legal catch, or fall while in the dugout after making a legal catch, the ball is dead and runners advance one base without liability to be put out.

Rule 7.04(c) makes no distinction as to whether the ball is fair or foul, and indeed the reference to the dugout in the rule suggests that no distinction was intended.

It seems to me that the interpretation of the rule should be this:

1. If Uribe didn’t make the catch before falling into the stands, it wasn’t a legal catch - the rule says that a fielder may *reach over* the fence to make a catch, not that he can *go into* the stands to make a catch.

2. If Uribe did make the catch before falling into the stands, then once he fell into the stands the ball should have been declared dead and the runners advanced one base.

—MWE

   12. SG Posted: November 02, 2005 at 04:22 PM (#1715801)

Crap, sorry ARoM, this seemed like a Dial type article.  Good work anyway.

   13. AROM Posted: November 02, 2005 at 04:29 PM (#1715820)

No prob, it is a Dial type article, Chris has something along the same lines in the works.

   14. AROM Posted: November 02, 2005 at 04:36 PM (#1715835)

In this system, I use 1.00 as the run value of an outfield assist.  Chris thought it should be 0.75, the same as any other assist.

My opinion is that an average OF assist has to be worth more, as it is more likely to happen at 3rd or home than an infield assist.

I looked at the run value of Guerrero’s 2004 assists using retrosheet and before and after run expectancy, and the average run value of his 13 assists that year was 0.84

That’s a small sample, but I’m not sure how to go about finding the average value of OF assists.  With Ray Kerby’s ASS, I was able to see assists broken down by what runners were on base, but not the number of outs or where the assist went. (With runners on 1st and second and a single is hit, I don’t know if they get the runner at the plate or the batter advancing to second)

Anyone know of a study on OF assists?

   15. Chris Dial Posted: November 02, 2005 at 04:37 PM (#1715838)

I’m not really sure if that’s better or worse than other methods. I’m not advocating using this over anything else.

Your data has an R=0.8 with the data MGL posted.  That’s pretty strong.

I will have a full list for my GG article (you, MGL, Grassko, maybe BPro).

   16. Chris Dial Posted: November 02, 2005 at 04:41 PM (#1715846)

Also, DPs is problematic because they are a “team” play.  A bad SS can completely screw over a 2B.  After a decade of work, Defensive Average, MGL and I came to the conclusion (independently) there isn’t enough difference to indicate much. 

You give half credit, but I think the problem is the number of DPs started by the 3B, where the 2B takes them all.  To sum up - you have far too much weight on DP runs.

One of the GREAT (and there are many graet things) about the Hardball Times book is DPs started.  That’s a great stat that could be investigated more.

   17. Scoriano Flitcraft Posted: November 02, 2005 at 04:55 PM (#1715870)

That figure includes a park factor for Manny, the only park factor used in these ratings. He’s at –32 without it.

I have a WAG that there may be a park factor for Floyd at Shea.

   18. Mike Emeigh Posted: November 02, 2005 at 05:02 PM (#1715883)

My opinion is that an average OF assist has to be worth more, as it is more likely to happen at 3rd or home than an infield assist.

Not that much more likely, actually.

I haven’t looked at recent data, but the last time I looked a couple of years ago, I found that most OF assists actually occur at 2B. The major plays here are (1) runners trying to stretch singles into doubles; (2) runners trying to advance on the back end of plays at 3B or home being thrown out; and (3) the occasional force plays on shallow pops to the OF.

—MWE

   19. Mike Emeigh Posted: November 02, 2005 at 05:17 PM (#1715914)

Also, DPs is problematic because they are a “team” play. A bad SS can completely screw over a 2B.

And vice versa. It’s more likely to be a problem with the 2B because (a) his job as the pivot man is harder and (b) he’s usually in the middle of more of them because there are more GBs hit to the left side.

—MWE

   20. Chris Dial Posted: November 02, 2005 at 05:21 PM (#1715919)

I have a WAG that there may be a park factor for Floyd at Shea.

What would be causing that?

   21. dcsmyth1 Posted: November 02, 2005 at 05:32 PM (#1715938)

I have a silly question about ZR vs UZR. At what point does increasingly granular information start to become too noisy and random, and actually start to make the ratings worse, rather than better? Or is there no such point?

   22. Scoriano Flitcraft Posted: November 02, 2005 at 05:34 PM (#1715947)

What would be causing that?

I’m not a stat guy Chris, which is one reason it is little but a WAG. However, I have always sorta kinda thunk that Shea was a bad LF HR park, <u>i.e.</u>, there are more fly balls to be caught out there. If true, it could be the wind pattern for all I know. I also think the Mets were very tough at home against right handed hitters but I have no idea how to allocate that b/w pitching (e.g., Pedro) and defense.

   23. John M. Perkins Posted: November 02, 2005 at 06:23 PM (#1716050)

I’m very much not a statitician, but ...
several years ago, Sagarin argued that more computer rankings needed to be added to the BCS calculation even if that devalued the Sagarin Ratings.  More systems took care of bumps in the data—Chone’s Floyd for instance.

So what I’d like to see is a BCS type of ranking of defense.  Take every defensive calculation out there, rank everybody 1-30 (maybe separate rankings for the non-primary fielders), remove the best and worst for each player, and add them up. [And maybe a version on the side where each position is scaled rather than rotisserie ranked.] That’s EVERY defensive system all y’all can get your hands on,[Fielding .pct; FR; ZR, UZR; Davenport’s; Chone’s; Strat; DMB; ZiPs; and apologies for those I’ve missed] with encouragement for more statheads to join the play.

Of course, I want someone else to run the rankings.

   24. Old Matt Posted: November 02, 2005 at 06:32 PM (#1716069)

Jeremy’s right. For the second year in a row, Cliff definitely looked above average, but I would have never guessed the best in LF.

When healthy, the Mets’ defense in the outfield was pretty slick in 2005.

   25. DSG Posted: November 02, 2005 at 06:34 PM (#1716072)

Chris,

It’s GASSKO. No “r”.

Anyways, nice article Sean. A few comments:

1. For double plays, you must include an estimate of runners on second. Otherwise a team with a lot of doubles or to a lesser extent, steals, allowed is going to look worse.

2. Is a DP really worth .75 runs more than a regular out? I would think that a DP is worth about the same as a CS over a regular out, roughly .2 runs.

3. You shouldn’t be using outfield assists. They’re worthless (they correlate negatively with actual arm rating done using PBP data).

I have some more but have to go right now.

Nice job.

   26. Chris Dial Posted: November 02, 2005 at 06:39 PM (#1716086)

If you look at the data in Chone’s spreadsheet, you can see that Floyd was actually rated as 8 runs for FBs and 8 runs for Arm.  He had 16 assists, and we’ve already noted that Chone may have the OF assists counting too much.

FLyd was “above average”.  However, I think Floyd’s 8 RS *is* the best in the NL (although Kelly Johnson does score well).

what htis helps to demonstrate is that BIP distribution is very important. 

Floyd did look better. He had fewer leg issues and he is pretty fast.

The numbers define how many FBs he caught (that applies to K. Johnson too), not how he looked . 

Say it with me - we’re aren’t selling jeans here.

   27. Chris Dial Posted: November 02, 2005 at 06:40 PM (#1716089)

Chris,

It’s GASSKO. No “r”.

My apologies.  I should stick with DSG.

   28. Chris Dial Posted: November 02, 2005 at 06:43 PM (#1716096)

You shouldn’t be using outfield assists. They’re worthless (they correlate negatively with actual arm rating done using PBP data).

I’ll suggest that baserunner holds (which is what I assume you are referring to with “arm rating from pbp data) is far more altered by the type of BIP in a given situation and the baserunner than the OF.  I like using OF assists, because there is some difference between OF and they are outs on base.

   29. AROM Posted: November 02, 2005 at 06:46 PM (#1716102)

David,

1. Steals (and CS, WP, PB) are part of the DP calculation.

2. You’re probably right, CS is a good comparison.  I expect I’ll change a few weights before I do this for 2006 data.  This remains a work in progress.

3. Something to look into

   30. AROM Posted: November 02, 2005 at 06:49 PM (#1716112)

I think assists have some value, because throwing a baserunner out is much more valuable than holding him one base.

I remember Bill James writing years ago about a weak armed OF with a lot of assists (Tim Raines?), saying that even if everyone’s running on him, there just is not enough plays available for those extra bases to hurt more than the 20 assists help.

   31. Cabbage Posted: November 02, 2005 at 06:54 PM (#1716128)

Both the Rolen and Grudzelanik ratings lead me to suspect that the system may still have some trouble with groundballish pitching staffs.

I’ve always wondered what defensive rating should do about something like that.  Certainly, we wouldn’t want to call an average defender who recieves many fielding chances “great”.  However, let’s take two hypothetical 2B of equal skill.  If they convert (random number for the sake of the example) 75% of balls hit into their area into outs, but one player recieves an extra 100 chances, shouldn’t he recieve some credit if playing at an above average level for more time?  Just as its easier to hit .300 in 200 ABs vs. 450ABs, shouldn’t players who play above average defense get some extra credit for doing so in spite of more chances to screw up?

Perhaps fielding stat Nirvana would be similar to a batting line.  Instead of .310/.420/.604 in 650PAs, we would get +15 FR/500 (fielding runs per 500 chances) in 320 chances.

This would also be somewhat useful when evaluating a players role.  High K, flyball staffs do not rely upon infield defense, and a great defensive 3B will be of less value than to a team like the Cardinals.

   32. Mike Emeigh Posted: November 02, 2005 at 06:58 PM (#1716142)

You shouldn’t be using outfield assists. They’re worthless (they correlate negatively with actual arm rating done using PBP data).

For LF and CF, this is definitely true; guys with high numbers of assists also normally give up a lot of baserunner advances. For RF, it’s not (or it didn’t used to be, anyway - I haven’t looked at this in a couple of years); RFs who have good assist totals typically are very good at also preventing the extra base. There were a couple of exceptions (Mondesi was one, Dave Parker during his years in Pittsburgh was another) but not many.

I count triples as a baserunner advancement; there are no parks left like Forbes Field that are triple-friendly, and almost all triples are the result of a runner (even though they are mostly fast runners) taking a liberty with an outfielder. Some fraction of doubles should also be counted as a baserunner advancement, too.

—MWE

   33. Mike Emeigh Posted: November 02, 2005 at 07:03 PM (#1716159)

Both the Rolen and Grudzelanik ratings lead me to suspect that the system may still have some trouble with groundballish pitching staffs.

Zone ratings in the infield do correlate (to a small degree) with staff groundball tendencies, as do zone ratings in the OF to flyball tendencies (to an even smaller degree). I think the Rolen/Grudz ratios are driven to some extent by the DP factor; the Cardinals turned a tremendous number of DPs for the runners they had on base.

—MWE

   34. AROM Posted: November 02, 2005 at 07:11 PM (#1716183)

The Cardinals had everything going for them in turning DPs:

1) extreme groundball staff
2) excellent work by the middle infielders
3) a catcher who kept lead runners nailed to first

   35. Chris Dial Posted: November 02, 2005 at 07:12 PM (#1716190)

I think the Rolen/Grudz ratios are driven to some extent by the DP factor

DPs aren’t in ZR anymore.

   36. Chris Dial Posted: November 02, 2005 at 07:13 PM (#1716193)

we would get +15 FR/500 (fielding runs per 500 chances) in 320 chances.

This is what my rating does…

   37. Mike Emeigh Posted: November 02, 2005 at 07:20 PM (#1716217)

DPs aren’t in ZR anymore.

No, but they are in Chone’s ratings, which is what we were discussing WRT Rolen and Grudz.

—MWE

   38. AROM Posted: November 02, 2005 at 07:33 PM (#1716257)

Grudz is +17 just by zone, with DP’s pushing him up.

Rolen is all zone, the DP rating I use only covers the middle infielders.

   39. Chris Dial Posted: November 02, 2005 at 07:34 PM (#1716263)

Sorry, Mike, I had already known that Rolen had no DPs (they are not in his 3B ratings at all), so I assumed you meant ZR.

   40. Mister High Standards Posted: November 02, 2005 at 07:46 PM (#1716287)

Boy does Texas have an awful infield.

Or their is something inherently weird about Texas’s ball In play distribution.

The way I would normally look at it to compare the starters to the backups and see if the results on a per inning basis were similar.  That approach won’t work for the Rangers since the last few years the Rangers have been getting 12-1300 defensive innings from the starters. 

Considering that Young and Blalock have fairly good reputations defensivly, I think its reasonable to assume until we KNOW differently that their is something unsual about the Rangers BIP distrubution, and the results for this specific team should be taken with a grain of salt.

   41. Chris Dial Posted: November 02, 2005 at 07:50 PM (#1716294)

AFAIK, Young doesn’t have a good defensive rep.

   42. AROM Posted: November 02, 2005 at 08:02 PM (#1716320)

I would have a problem with the Texas ratings if in spite of these ratings, their pitching was good.

Since its not, I can buy that their defense is made up of butchers.

   43. AROM Posted: November 02, 2005 at 08:06 PM (#1716325)

There could also be a ballpark effect in the Texas rating.  The only park factor I use is the Green monster factor.

I really have no idea how much a park affects ground balls.

   44. Mike Emeigh Posted: November 02, 2005 at 08:23 PM (#1716346)

Or their is something inherently weird about Texas’s ball In play distribution.

I doubt it.

Texas does have a slight groundball staff (reflected in the GO/AO rate and the HR allowed, which was unusually low for a Ranger staff). They gave up a huge number of non-HR hits, and allowed the most singles in the league by a good margin. That’s usually reflective of a poor defensive infield.

Oakland allowed the fewest singles in the league (close to 200 fewer than Texas), with a GO/AO rate and a HR allowed total not much different than Texas’s. The ranking for Ellis and Chavez are good, of course, and Crosby was +12, just missing the top group. I think those are fair assessments of the defensive performance of the two infields.

—MWE

   45. Mike Emeigh Posted: November 02, 2005 at 08:25 PM (#1716351)

I really have no idea how much a park affects ground balls.

Not much, any more, with high-tech groundskeeping methods becoming standard and teams resorting to less trickery along the lines of what Pittsburgh and the Giants used to do.

—MWE

   46. Mister High Standards Posted: November 02, 2005 at 08:32 PM (#1716367)

Dial - You must not watch many Ranger games.  Both home and road announcers rave over Young glove.  Jamie Newberg also thinks highly of his glove.

I would have a problem with the Texas ratings if in spite of these ratings, their pitching was good.

I understand what you’re saying, but I don’t think it NECASARILLY follows.  I mean it might follow it might not.  I think the Rangers were 11th in park adjusted run prevention in 2005.  Its possible that the defense and pitching were equally poor.  Its possible that the pitching was mediocre and the defense was relatively worse than the 11th or the opposite could also be true.  I don’t think we know which of the scenarios are true, because we are so foggy on separating pitching from defense.

   47. Mister High Standards Posted: November 02, 2005 at 08:33 PM (#1716369)

Mike - good points.

   48. Chris Dial Posted: November 02, 2005 at 09:02 PM (#1716435)

You must not watch many Ranger games. Both home and road announcers rave over Young glove.

I watch “some”.  I was never impressed.  And to be clear - not about his glove, about his ability to turn GBs into outs (low range left/right, not surehandedness).

   49. Cabbage Posted: November 02, 2005 at 09:52 PM (#1716549)

This is what my rating does…

sweet.

I think assists have some value, because throwing a baserunner out is much more valuable than holding him one base.

In a Franoceur thread from august, someone suggested that the most valueable arm is the one no one knows about.  Kind of like how Pudge was so good early in his career that his value suffered because no one tried to run on him.

BTW is that true about Pudge? I wasn’t following baseball closely at the time.

   50. Spivey Posted: November 02, 2005 at 10:10 PM (#1716591)

Rauseo: The Rangers announcers are some of the worst there are. Incredible homers and little understanding of baseball. Michael Young tends to short arm throws and also doesn’t have particularly good range. Blalock also doesn’t have good range. Michael Young was well regarded by scouts as a defensive 2B coming up and Blalock was well regarded defensively too. I think it’s just likely those are wrong. I wonder how much road announcers end up just assuming that home announcers will be right about things like defense and agree with them. The Rangers announcers talk up Soriano’s defense and I know if you asked them they’d say he’s an above average defender who just has a *little* problem with his hands.

I agree of course with Sean Smith that Soriano and especially Blalock need to be traded. Kinsler isn’t supposed to be good defensively though, and his offense at AAA this last year wasn’t anything to write home about.

Watching the Rangers play a lot, and then watching the playoffs, the difference is really night and day. Plays that would spectacular plays for the Rangers fielders are made regularly by good defensive infielders.

   51. Spivey Posted: November 02, 2005 at 10:12 PM (#1716594)

Sean Smith: Can you post what Laynce Nix’s numbers were? Thanks.

   52. AROM Posted: November 02, 2005 at 10:23 PM (#1716617)

Spivey, Nix was +5 ZR, average for assists/errors.

There is a link to an excel file at the end of the article.  All players with 250 or more innings at a position are included.

   53. Spivey Posted: November 02, 2005 at 10:37 PM (#1716651)

Ok. Thanks.

   54. villageidiom Posted: November 03, 2005 at 12:01 AM (#1716857)

LF: Manny Ramirez –25, Hideki Matsui –9, David Delucci –9

That figure includes a park factor for Manny, the only park factor used in these ratings. He’s at –32 without it.

Manny is bad, but I have trouble believing that in 2005 he was almost as bad by himself as an entire outfield of Matsuis. Having seen a lot of both of them in 2005, this seems absurd to me.

   55. AROM Posted: November 03, 2005 at 12:22 AM (#1716898)

Try UZR, then.  That system has Manny rated as -47, park adjusted.

MGL didn’t post NL numbers, but tells us Griffey is even worse.

   56. greenback calls it soccer Posted: November 03, 2005 at 12:37 AM (#1716928)

3) a catcher who kept lead runners nailed to first

What made Yadier Molina even more valuable were all the pickoffs, which I assume aren’t counted above.

   57. Rough Carrigan Posted: November 03, 2005 at 12:51 AM (#1716947)

Is there any chance that the Texas infield bakes in the course of the summer and becomes harder than most infields thereby playing faster and making every player’s range seem poorer than it is?

   58. AROM Posted: November 03, 2005 at 01:01 AM (#1716964)

Maybe, but if so Teixiera was unaffected.

What made Yadier Molina even more valuable were all the pickoffs

How many did he have?  If I can find a source showing pickoffs by catcher I’d include it in the ratings.

   59. Spivey Posted: November 03, 2005 at 01:08 AM (#1716970)

Is there any chance that the Texas infield bakes in the course of the summer and becomes harder than most infields thereby playing faster and making every player’s range seem poorer than it is?

DSG uses park factors in his numbers which also show the Texas infield to be terrible. I think UZR uses park factors as well, which shows the same thing.

   60. greenback calls it soccer Posted: November 03, 2005 at 01:34 AM (#1717004)

How many did he have?

Eight or nine, I think. What makes it more valuable (and weirder) is that he mainly picked off trail runners with two outs.

If I can find a source showing pickoffs by catcher I’d include it in the ratings.

Retrosheet will have it, but I suppose that’s a way’s off.

   61. greenback calls it soccer Posted: November 03, 2005 at 01:35 AM (#1717007)

Oh, and that’s a nice spreadsheet.

Your use of “Chone” as a pen name may get Furtado’s lawyers a little antsy though.

   62. DSG Posted: November 03, 2005 at 02:46 AM (#1717102)

DSG uses park factors in his numbers which also show the Texas infield to be terrible.

Actually, I don’t. I’m with MAH here. Park factors are only required for a few situations.

MGL didn’t post NL numbers, but tells us Griffey is even worse.

Griffey was -59 in UZR/150. You read that correctly. -59. (And I hope MGL doesn’t mind me posting this. I’ve tried to avoid posting other ratings shared with me, but I assume Mitchel was going to post this rating anyways).

I’ll suggest that baserunner holds (which is what I assume you are referring to with “arm rating from pbp data) is far more altered by the type of BIP in a given situation and the baserunner than the OF.

UZR (and also what Protrade did maybe a month ago) adjusts for type of BIP, Base situation, etc. If done correctly, IMO, this is much better than assists. Mike E. is probably right about assists for right fielders. MAH has also found that errors are statistically significant for RF as well.

Steals (and CS, WP, PB) are part of the DP calculation.

Right, but what you’re doing is subtracting them from the number of men on first. Also remember that a steal means a man on second. You need to double count in this case. I’ve been doing a little research with this stuff; it’s actually very tough to get good DP ratings without having situation dependent data (which MGL uses for UZR). BTW, a good way to split credit for double plays (Chris, I think, was talking about how this is tough to do) is to base it on A/(A+E) vs. the average at a position. That generally gives you a good idea of a player’s “arm”, which I think would be the key to being good or bad at turning DPs.

Also, Chris, what use do you foresee for the DP started data? It seems to me that DP turned might actually have more value.

Also, Chris, I can provide you with data on other players if need be for your article.

Finally, post #44 is spectacular. A must-read to understanding defense on a conceptual level.

   63. Mike Emeigh Posted: November 03, 2005 at 02:55 AM (#1717110)

How many did he have?

Eight. I should have all of them by the end of the year.

—MWE

   64. Chris Dial Posted: November 03, 2005 at 03:25 AM (#1717141)

DSG,
is your data based on 162 games or actual innings played?

   65. DSG Posted: November 03, 2005 at 03:30 AM (#1717142)

Actual. I can convert it esily if you’d like.

   66. Chris Dial Posted: November 03, 2005 at 04:18 AM (#1717168)

That could be driving down your correlation.  Could you generate it vs. 150 DG (1350 IP)?  I’ll send you an email.

   67. DSG Posted: November 03, 2005 at 04:48 AM (#1717183)

Chris, just replied. And forwarded it to Aaron + Studes.

   68. I Love LA (OFF) Posted: November 03, 2005 at 09:15 AM (#1717348)

Re Floyd: I have to agree with Jeremy and Matt. Cliff looked spectacular in LF this year. He was throwing out runners left and right, running full speed, and even doing summersaults to catch balls.

BTW, if you don’t mind me asking, what were Reyes/Wright’s?

   69. AROM Posted: November 03, 2005 at 03:13 PM (#1717424)

OFF, the last paragraph of the article has a link to a spreadsheet.

You can get the ratings for any player with 250 or more innings at a position.

   70. Slinger Francisco Barrios (Dr. Memory) Posted: November 03, 2005 at 03:22 PM (#1717435)

I remember Bill James writing years ago about a weak armed OF with a lot of assists (Tim Raines?), saying that even if everyone’s running on him, there just is not enough plays available for those extra bases to hurt more than the 20 assists help.

Probably that was Vince Coleman.  If not, Vince Coleman is my personal poster child for this phenomenon.

   71. Slinger Francisco Barrios (Dr. Memory) Posted: November 03, 2005 at 03:58 PM (#1717469)

ARoM, do you have overall team rankings?  I.e., did you run everyone, or just the 250+ guys?

   72. AROM Posted: November 03, 2005 at 04:00 PM (#1717472)

I could put together team rankings.

Possibly by this weekend.

   73. Slinger Francisco Barrios (Dr. Memory) Posted: November 03, 2005 at 04:19 PM (#1717498)

The thing is, when I add up the runs, I get +112.  I figger it should add up to zero, or pretty close to it, unless the scrubs were really atrocious all around.  So I mistrust the team totals one arrives at by adding up the 250+ guys.

   74. I Love LA (OFF) Posted: November 03, 2005 at 04:59 PM (#1717562)

Oh cool. Somehow I missed that.

   75. AROM Posted: November 03, 2005 at 06:00 PM (#1717647)

For infield and outfield, everything should sum to zero, even the dp, assist, and errors stuff.  If it doesn’t when I add in the under 250 inning players, then I’ve done something wrong.

The catchers will not sum to zero.  The pb, e, wp, and sb attempts (the one used for keeping dp in order) will sum to zero, but the SB and CS are linear weight runs.

If all catchers threw out 50% or more of the runners, then catching as a group would have a positive run value, and you could blame it on stupid runners.  In reality, it comes out pretty close to zero.

   76. AROM Posted: November 03, 2005 at 06:00 PM (#1717649)

My plan is to put together team totals, by position, and also compare to a team estimate based on DER.

   77. KJOK Posted: November 03, 2005 at 08:15 PM (#1717914)

I remember Bill James writing years ago about a weak armed OF with a lot of assists (Tim Raines?), saying that even if everyone’s running on him, there just is not enough plays available for those extra bases to hurt more than the 20 assists help.

Probably that was Vince Coleman. If not, Vince Coleman is my personal poster child for this phenomenon.

I was thinking it was Lonnie Smith….

   78. Slinger Francisco Barrios (Dr. Memory) Posted: November 03, 2005 at 08:37 PM (#1717951)

Sounds good, thanks for listening, ARoM.

   79. DSG Posted: November 03, 2005 at 09:38 PM (#1718039)

Sean,

Don’t compare to DER. ZR is much better. You should be using DIPS, and adding the two up to compare to runs allowed.

   80. AROM Posted: November 03, 2005 at 11:50 PM (#1718272)

I want to compare apples to apples.  There are two ways to do this:

Compare DER to ZR plays saved (before converting to runs, a minor but important detail since I use a different run value for IF and OF)

Compare FIP + ZR runs (including dp/e/ast/catcher) to actual ERA. 

I might as well do both.

   81. Chris Dial Posted: November 04, 2005 at 12:11 AM (#1718318)

For infield and outfield, everything should sum to zero, even the dp, assist, and errors stuff. If it doesn’t when I add in the under 250 inning players, then I’ve done something wrong.

Sort of - the fractional run values make mine off by a run or two in various places.

I need someone to help me with the runs from FIP or DER.  I did that and it stunk.

   82. DSG Posted: November 04, 2005 at 12:18 AM (#1718338)

Chris, what do you mean?

   83. AROM Posted: November 04, 2005 at 12:40 AM (#1718358)

Quicker than I thought.

R values:

Der plays saved to zone plays saved: .58

FIP + (total my stuff) to ERA: .93

FIP to ERA: .85

So my ratings explain about half the difference between FIP and ERA.

Chris, spreadsheet is headed your way.

   84. Chris Dial Posted: November 04, 2005 at 01:04 AM (#1718372)

Rally,
my ratings and your ratings correlate at 0.97, for the guys we’ve seen from DSG and MGL (about 100 guys).

Oh, I removed DPs and arms from yours and tweaked to 150 DG, so mine, yours, MGL’s and DSG’s would all be on the same scale.  they’ll all be side by side shortly.

Technically, DSG’s has DPs and arms, but I think that won’t matter much.

My spreadsheet does every player.

   85. DSG Posted: November 04, 2005 at 04:06 AM (#1718488)

Chris,

I don’t do DP or arm, so actually they’re all on the same scale.

   86. Chris Dial Posted: November 04, 2005 at 03:40 PM (#1718748)

Fantastic!

Thanks, David.

   87. MAH Posted: November 05, 2005 at 08:49 PM (#1720054)

This is great.  At last we’re going to have multiple, independently
calculated, publicly posted fielder runs-saved ratings featuring
intelligently used zone and non-zone data.  Maybe we can combine the three into one system-the Chone-Dial-Gassko, or Composite Defensive Grade (“CDG”), system.  Unless Chone would like to use his real name, in which case we could call it the Grand unified Defensive System (“GuDS”).  ;-)

Some thoughts-

First, the ratings look very good.  The few times I’ve seen Cliff Floyd this past year he’s impressed me as a hustling player-a +8 runs saved excluding assists sounds reasonable.  Also, the rating for Manny Ramirez seems dead-on.  So often he seems to have no idea where a ball is heading.  I’m not surprised that the Red Sox management hasn’t been keen to keep him.

It’s difficult to allocate DP credit between second basemen and shortstops. I agree with Chris that it’s generally not worth the trouble, at least for contemporary players.  Perhaps the reason is that turning a double play well or poorly is a defensive skill that can be reasonably well assessed from observation (whereas range often isn’t), and players who can’t do it well just aren’t allowed to play at second or short.  If double plays are included, their run weight should be slightly higher than a walk allowed, or about .45, based on Tangotiger’s markov-type calculations and regressions I ran while developing DRA.  When you think about it, completing the double play prevents a runner from reaching first whose at-bat has minimal or negative runner-advancement value, even apart from the effect of the out at second, so it should be worth less than preventing a single.

Outfield assists are tricky.  In my historical DRA studies, assists do
matter a great deal in right field.  Jesse Barfield was the prime example. Also, the fact that the only errors that have statistical significance in contemporary DRA regressions are in right field probably reflects that such errors are predominantly throwing errors, in which the stakes are high.

MWE’s observation that most outfield assists are at second is also consistent with DRA.  The run-weight for centerfield assists is out .75,
whereas the run-weights at the corners are more like 1.25.  That probably reflects the fact that it is very difficult for a centerfielder, who plays deeper than corner outfielders, to _directly_ prevent a runner from scoring from third base.  I would imagine that most centerfield assists occur at second base.  Also as MWE observed, the difference between a double and a triple is largely determined by the rightfielder, no longer by park effects. Obviously, the DRA run-weights are high because regression weights “carry” non-measured factors that correlate with outfield assists and run scoring-such as, possibly, preventing runner advancement when no throw is made, as well as the run-expectation value of having a runner on third base before he scores.

The correlation results also make sense.

The DRAZR ratings I introduced in February at The Hardball Times, which are just a linear combination of DRA and my best attempt at the time to convert ZR into runs-saved, also had a .8 correlation with UZR _without_ any editing of the data.

The correlation between runs allowed and (FIP + your system) is also close to the DRA model accuracy, which is also about the same as the accuracy of models of team _offense_.

Finally, the correlation between your system and DER sounds about right, as flyouts caught by infielders (Infield Fly Outs, or “IFO”) are excluded, and DRA suggests that about 40% of the team-by-team variation in BIP run-prevention is explained by flyouts caught by infielders.  Naturally park effects come into play.  Since DRA allocates IFO outs to pitchers, that gives pitchers a meaningful impact that represents a modification of DIPS.

Does your system account for pitcher fielding plays?  If not, you might get an even better correlation result.

I’ve made a post to Chris Dial’s article about Zone Rating with some
suggestions for improving ZR on its own.  I think it still has some
significant limitations in the outfield, but they might be easily remedied using the suggestions I offer there.  Also, it would be interesting to use David Pinto’s outfielder ratings _without park effect adjustments_ and his infielder ratings with _only_ groundballs included.

Thanks again.  We’re making progress.

   88. Bring Me the Head of Alfredo Griffin (Vlad) Posted: November 06, 2005 at 04:35 PM (#1720610)

Watching Lonnie Smith play defense was a blast.

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Support BBTF

donate

Thanks to
BDC
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Syndicate

Page rendered in 0.9870 seconds
47 querie(s) executed