89 replies to this topic

[941827]

From Today's NYT:

Manny Being Manny Is Hurting the Red Sox By DAN ROSENHECK

***

All of today's best P.B.P. systems agree that Ramírez is the worst defensive left fielder in baseball, and by a comfortable margin. This holds true even after accounting for the effect of the Green Monster wall in left field.

"Manny is at the far end of the as-bad-as-you-can-get-in-the-field spectrum," said Mitchel Lichtman, who designed one highly regarded P.B.P. defensive statistic called ultimate zone rating, and who consulted for the St. Louis Cardinals from 2004 to 2006.

***

At one extreme, the most conservative estimates suggest that Ramírez's outfield play last season, when compared with an average left fielder's, cost the Red Sox 13 runs. That's not good, but not bad enough to make him meaningfully worse than elite corner outfielders like the Angels' Vladimir Guerrero, the Yankees' Bobby Abreu and the Pirates' Jason Bay.

By contrast, Lichtman's system says Ramírez was 32 runs below average last year, which would make him one of the game's most overrated and overpaid players. If he was actually that awful, he was no more valuable than the Mariners' Raúl Ibáñez, the Blue Jays' Reed Johnson or the Angels' Juan Rivera.

The truth most likely lies in the middle: the average of the P.B.P. systems' results for Ramírez is 15 to 20 runs below average. That's enough to take a significant bite out of his value. If his hitting begins to deteriorate this season (he turns 35 in May), and his base running remains poor (subtracting another two runs or so), he is likely to be worth about as much in 2007 as Oakland's up-and-coming Nick Swisher. In other words, an All-Star, but not a franchise player, and certainly not worth anywhere near his $20 million salary.

The article goes on to suggest that the Sox would be wise to move Ortiz to first, Manny to DH and Youks to either LF or third (requiring a Lowell trade).

I remain unconvinced that any statistical system can accurately measure defensive prowess, particularly (despite the article's assertion to the contrary) things like ability to deal with the Monster. Manny's not a great defensive LF'er, but to say he costs the Sox 15-30 runs a season with his defense is laughable.

Edit: There is also an ongoing thread about this in "The Sandbox."

Edited by 941827, 18 March 2007 - 10:21 PM.

[philly sox fan]

There's a lenghty thread over at BTF about this article with some dissenting views on the validity of UZR in Fenway.

Link

STATS, which provides the data that is fed into UZR, apparently counts balls of the wall as in play for whatever reason. It seems possible - and perhaps likely - that a more discerning look at the input data could knock his rating into the -10 to 15 range.

[bellyofthebeast]

From Today's NYT:
The article goes on to suggest that the Sox would be wise to move Ortiz to first, Manny to DH and Youks to either LF or third (requiring a Lowell trade).

I remain unconvinced that any statistical system can accurately measure defensive prowess, particularly (despite the article's assertion to the contrary) things like ability to deal with the Monster. Manny's not a great defensive LF'er, but to say he costs the Sox 15-30 runs a season with his defense is laughable.

There's a similar thread in The Sandbox where Lurker Blessyouboys84 posted a pretty interesting link.

http://www.insidethe...com/manny.shtml

Edited by bellyofthebeast, 18 March 2007 - 11:15 AM.

[Todd Benzinger]

To find the juicy stuff from Philly's link, you have to click around and scroll down.

Go toBTF SOx Therapy Defense

and scroll down to post 27 in the discussion area.

Chris Dial's post states:

To sum up - Manny is every bit as awesome as his offense indicates. Maybe not 4 WAR, but 3 WAR, rather than, as MGL stated last fall, "about average due to his defense".

In my Gold Glove article last season, I stated that "as I look at more data, I think Manny might be average", it is getting closer to that.

He also proposes a research project for members of RSN: reviewing the video of last season and doing PBP analysis of Manny's Wall balls.

[PedroKsBambino]

There's a similar thread in The Sandbox where Lurker Blessyouboys84 posted a pretty interesting link.

http://www.insidethe...com/manny.shtml

That analysis by mgl is what is being commented on, basically, in the Sox Therapy thread and the BTF thread.

Suffice to say that there's a lot of people who find mgl's conclusions in that link to be marginal at best, and I say that as someone who thinks mgl is a terrific analyst generally.

[The_Powa_of_Seiji_Ozawa]

That analysis by mgl is what is being commented on, basically, in the Sox Therapy thread and the BTF thread.

Suffice to say that there's a lot of people who find mgl's conclusions in that link to be marginal at best, and I say that as someone who thinks mgl is a terrific analyst generally.

I agree. It's analysis like that which gives statheads a bad name. However, if it means getting a better bat in the lineup, I do think there is some value to playing Ortiz at 1B at least occasionally outside of interleague play, when the affected personnel are slumping. It might mean a few more at-bats for Wily Mo over the course of the season, since I am willing to bet that under the current conditions he will probably not come close to the 400 ABs that some are predicting.

[mangotree101]

Lichtman's article was discussed here extensively when it was posted last year, but I can't find the thread. This one goes over a lot of the same ground, though...

Edited by mangotree101, 18 March 2007 - 01:51 PM.

[OCD SS]

There's a similar thread in The Sandbox where Lurker Blessyouboys84 posted a pretty interesting link.

http://www.insidethe...com/manny.shtml

That article lead to the What is Manny Really Worth? thread. My concern with Manny's defense has always been the short balls that fall right in front of him. There are also the balls that hit low on the monster that he could've caught if he'd been able to get back to them instead of just playing the carrom every time (not that that isn't the correct strategy, but it is a strategy that comes out of Manny's defensive skill, IMO). But my accounting for these has always been pretty subjective, and I wouldn't be surprised to find that the - 10-15 number is more accurate.

[glennhoffmania]

I remain unconvinced that any statistical system can accurately measure defensive prowess, particularly (despite the article\'s assertion to the contrary) things like ability to deal with the Monster. Manny\'s not a great defensive LF\'er, but to say he costs the Sox 15-30 runs a season with his defense is laughable.

I agree 100% (except for the extra slashes). I'm not a stat head but I'm also not a complete idiot, and I've yet to read or hear anything that convinces me that defensive metrics are reliable, especially when you're talking about Fenway.

But I do like the idea of playing Ortiz at 1B more so that Manny can DH and WMP can get some more ABs. I think resting Manny's knees is just as important a factor in this as his "league-worst defense" though.

[mangotree101]

But I do like the idea of playing Ortiz at 1B more so that Manny can DH and WMP can get some more ABs. I think resting Manny's knees is just as important a factor in this as his "league-worst defense" though.

I agree. The NYT's idea of playing Youks in LF is pretty idiotic, though -- he looked completely lost out there at the end of last season. Getting ABs for WMP has to be one of Tito's priorities this year.

[The_Powa_of_Seiji_Ozawa]

I agree. The NYT's idea of playing Youks in LF is pretty idiotic, though -- he looked completely lost out there at the end of last season. Getting ABs for WMP has to be one of Tito's priorities this year.

Indeed. But I'm afraid WMP is to Tito Francona as Gerald Green is to Doc Rivers. At least we won't hear Tito telling us that WMP has to earn his innings...

[Pumpsie]

That article lead to the What is Manny Really Worth? thread. My concern with Manny's defense has always been the short balls that fall right in front of him. There are also the balls that hit low on the monster that he could've caught if he'd been able to get back to them instead of just playing the carrom every time (not that that isn't the correct strategy, but it is a strategy that comes out of Manny's defensive skill, IMO). But my accounting for these has always been pretty subjective, and I wouldn't be surprised to find that the - 10-15 number is more accurate.

Manny plays the most shallow left field at Fenway imaginable. No one plays more shallow than Manny does. There are very few balls which fall in front of him. (Remember Mike Greenwell? Everything fell in front of Greenie because he played with his back against the Wall) and his defense was considered better than Manny's by these ridiculous stats. The more valid concern would be balls which hit low off the Wall because Manny plays so shallow. But many of these are just singles because they're hit so hard. I think that Manny plays a very smart Fenway left field that limits the damage his limited range can do. That's why I think the stats are bogus. They really make no sense at all.

Where I DO worry about Manny is playing on the road in spacious left fields where his lack of range is most likely to be a problem. But the stats don't really jive with that either.

UZR IS screwed up when it comes to Fenway and someday someone will figure out just how.

Edited by Pumpsie, 18 March 2007 - 04:22 PM.

[OttoC]

Manny plays the most shallow left field at Fenway imaginable. No one plays more shallow than Manny does. There are very few balls which fall in front of him.

Ramirez plays shallow because he doesn't come in well on balls hit in front of him. Even though he plays shallow, there are still balls he does not get to that he should. On the other hand, he is not playing as shallow as he did a couple of years ago because too many non-wall-balls got over his head. I think that his playing shallow just allows him to get to balls in front of him that most left fielders would get to anyway. And they would play deep enough to get to some of the line drives that get over Ramirez's head. This is particularly noticeable at home.

Is Ramirez's lack of fielding prowess more costly in Fenway Park or on the road? There's a lot of discussion about whether Fenway's unique configuration is appropriately accounted for in fielding metrics, especially wall balls. While this may be true, he does play roughly half his games in other parks, so it might be interesting to see how his compare to other fielders with all games at Fenway omitted.

[glennhoffmania]

Is Ramirez's lack of fielding prowess more costly in Fenway Park or on the road? There's a lot of discussion about whether Fenway's unique configuration is appropriately accounted for in fielding metrics, especially wall balls. While this may be true, he does play roughly half his games in other parks, so it might be interesting to see how his compare to other fielders with all games at Fenway omitted.

I'd like to see that comparison as well. As for Manny at home, I think he's at least adequate based solely on watching him. He plays the wall very well. He never goes back to the scoreboard for a ball that ends up 10 feet up on the wall. He plays the caroms very well, and with his quick release and accurate arm he's able to hold guys to singles instead of doubles. He seems to play the corner well. With Coco in CF to help cover the left-center gap, I don't think Manny's lack of range is a huge issue. So overall, I think he's fine in Fenway but he's noticeably worse on the road when he doesn't have the advantage of knowing the park so well. This is why for the past couple of years I've been advocating using him as the DH for 20-30 games on the road while Ortiz plays 1B.

[TomRicardo]

Manny Ramirez is the Moby Dick of Defense Sabremetrics.

With some defnsive ranking systems, it as if they can prove that Manny is so poor defensively that he is not worth the 20 million then that they have validated the use of defensive statistical systems. It is discussed at ad nauseum. Often his defense short comings are exagerrated. He is a poor left fielder although he is no means disastrious.


Manny may be a bad fielder but there is no way bad enough to offset his offense. Playing him in left field is a necessary evil.

[The Gray Eagle]

Ramirez got worse defensively last season as the year went along. By August, he could barely move out there. That's about the time when I noticed some balls falling in front of him, which hadn't happened often in previous years.

His defensive numbers from last year should be worse than the ones from the previous few years, as he seemed to be less mobile out there later in the season, especially in terms of getting a jump on the ball.

[Philip Jeff Frye]

Manny Ramirez is the Moby Dick of Defense Sabremetrics.

And the Sox are the Moby Dick of the New York Times lately, between this and Murray Chass's endless ramblings. Kind of odd considering the economic relationship involved...

[paulftodd]

I just can not buy into the defensive metrics when they have Manny costing this team so many runs. I am not a Manny apologist by any means and if he was costing the team this many runs I would jump at embracing these poor numbers as supporting my intuitive sense. However, the fact is at times I feel Manny is actually an above average LFer at Fenway, and average or slightly below average on the road.

The fact that these PBP stats did not record how high a ball hits off the wall and treat all balls off the wall as in play make them meaningless for evaluating Fenway LFers. Fenway has the smallest foul territory in LF making catchable foul balls a rarity and costing the LFer, and those cheap HR's into the monster seat that are catchable in other parks also hurt the Lfers rating. Over at the BTF one poster (Misterdirt) recorded numbers from retrosheet data from 2003-2005 showing that Manny caught 64.4% of all FB hit to LF, and 12.1% of LD while the opponents Lfer at Fenway caught 66.9% of FB, and 10.4% of LD. This puts Manny at average at least, at least for 2003-2005. His better performance on LD's supports the fact that he plays a shallow LF in Fenway. In 2006 his knee did bother him to some extent and he may have declined somewhat from the 2003-2005 numbers.

The defensive metrics expressed as runs do not cover the strength or effectiveness of an OFers throwing (they also do not measure the effectiveness of a 1B in catching errant throws, or middle IFers turning the DP, although they may rank these separately).

There was an interesting article over at the THT on OF throwing performance in 2006 that did so. Manny came in 3rd in LF, saving 4.2 runs above average, helped no doubt by Fenways short LF and his quick release,

Slightly off topic but Coco not surprisingly was ranked low at 22 (Damon was dead last at 26). Drew was somewhat of a surprise in RF, ranking near the bottom at 18 of 21 (Trot was 9th) resulting in this comment by the author.

Seeing J.D. Drew so low surprised me, since he is considered to have a strong arm. It may be that his shoulder problems in 2006 were hampering his throwing.

[flymrfreakjar]

Drew was somewhat of a surprise in RF, ranking near the bottom at 18 of 21 (Trot was 9th) resulting in this comment by the author.

His arm looks pretty good here. Fun note: He also homered twice that game.

[bowiac]

I just can not buy into the defensive metrics when they have Manny costing this team so many runs. I am not a Manny apologist by any means and if he was costing the team this many runs I would jump at embracing these poor numbers as supporting my intuitive sense. However, the fact is at times I feel Manny is actually an above average LFer at Fenway, and average or slightly below average on the road.

I'm sorry - are you saying that you're only going to embrace numbers which fit your intuition? I may be misunderstanding you here, but that seems to be a pretty curious standard to hold a stat to.

[Vermonter At Large]

Well ... I still think the numbers are very controversial. I brought this up in the DSG chat, but I don't think the defensive metrics are using the correct out values in their computation. There are essentially two sets of outvalues that can be used with lwts, depending on if you are using a probablisitic (Base Runs or Batting Runs) model or cumulative (Runs Created) model. In my view, they should be using the smaller cumulative value (about -.11 runs per out) because UZR is a cumulative statistic. Why they continue to use the probabalistic (-.30 runs per out) value is beyond me. They also calculate the value of a defensive miscue as -.81 runs, which I think is absurd. This not only includes the higher probabalistic out value, but charges the fielder about .50 runs for runs that the ensuing batters are likely to get from that non-out. I believe that the fielder should be charged about .18 runs for the extra plate appearance, but the rest of what happens should fall on the pitcher.

At any rate, in addition to the assymetrical ballpark zone problems brought out in the chat, I think that the run penalization the MGL uses for UZR is about twice what it should be.

[5belongstoGeorge]

Manny Ramirez is the Moby Dick of Defense Sabremetrics.

If Manny is Mobey Dick then there are more than few Ahab-like Saber-dudes floating around on the baseball sea.

If Manny was -15 runs in the field that would make him even more valuable than the number that was assigned in the What is Manny really worth? thread. If I recall correctly that number was set at -22 runs.

[paulftodd]

I'm sorry - are you saying that you're only going to embrace numbers which fit your intuition? I may be misunderstanding you here, but that seems to be a pretty curious standard to hold a stat to.

Fair enough. When my intuition and the manufactured numbers agree, I generally accept them without a great deal of scrutiny. If they do not agree, I know either my intuition is wrong or the numbers are wrong. So I look at how the numbers are generated, and what assumptions are made, and if I agree with how the numbers are generated, I accept that my intuition is faulty. If the the way the numbers are generated make no sense, and the assumptions are flawed, and they do not measure all they say they are measuring, or if they do not mean what they say or imply that they do, I stick with my intuition. The numbers have convinced me that Jim Rice may not be the slam dunk HOFer I thought, just a marginal one, but the defensive stats have not convinced me that Manny is as bad as they say he is in LF. Until they do a better job of validating the stats, I am a skeptic.

MGL is at least honest about UZR which is not available now, his comments from the BTF thread make me feel a bit better about not accepting the various rankings as gospel.

Using scouting reports and observation (video, etc.) can reduce the uncertainty level of a "conclusion" (best estimate) drawn from sample data, and give us a more accurate number, if done and used properly (the observation and scouting reports) of course. I use a "lazy approach" (to quote some other poster) for my final defensive numbers for several reasons. One, it is not particularly important (to me) how accurate I am. Two, I don't generate defensive numbers for a living. Never did. This relates to one. Three, and most importantly, I have 10,000 other things that I do in my baseball research. I would have to put some or all of them aside if I took the time to use a "non-lazy approach." I have to balance streamlining all of my methodologies with an effort to use what I consider a quality one (methodology).

Finally, this thread is about half interesting and useful and about half assinine. The assinine part is why I rarely post here anymore

Not saying they do not have some value, but they are not so accurate as to not be questioned when they do not match your intuition, and the various metrics available do not even agree with each other, but in Manny's case they all agree he is the worst LFer.

Now Manny may very well be a below average LF er now, but I just do not believe he is the worst in MLB and is costing this team 20 runs per year in the field. Dewans Fielding Bible had a separate ranking for arms in the FB and had Coco rated above Manny among LFers in 2005. How can I believe that stat after watching Coco throw last year? He also ranked Manny last among LFers in 2005, saying he cost the team -31 runs (not sure what his 2006 numbers for Manny were since they are not available now, it must have been worse, because the overall OF defense defense according to his system was -69 runs, worst in the MLB, and 60 runs worse than 2005).

[Worst Trade Evah]

I liked the part in the BTF thread where mgl accused philly of making "too much sense".

I know what he meant, but it read kind of funny. Plus, philly was right, as he always is. -10 to -20 depending on the state of his knee I have no trouble believing; -32 is pretty unlikely (though I realize mgl hedges this in various ways).

[Tangotiger]

They also calculate the value of a defensive miscue as -.81 runs, which I think is absurd.

The 0.80 figure is correct, and an illustration is provided here in post#11:

http://www.tangotige...0247.shtml#1011

[Rudy Pemberton]

The problem I always have with this data is that the circumstances for all fielders are very different, this is especially true for LF who play half their games at Fenway Park. I would be interested to see Manny compared to LF playing opposite him, but those numbers seem to be tough to get at and would still not be perfect due to pitchers, but it would at least normalize for environment (park, weather, etc.). Fielding is also greatly impacted by pitching and other fielders...now, I know many can suggest that the Red Sox pitchers were unlucky because of poor OF defense, and while that may be true, the pitchers that appeared to be the "unluckiest" (Snyder, Hansen, Delcarmen, Johnson, etc.) make me wonder if crappy pitching can make defenders appear worse as well.

I'm interested in the point made in the NYT article about how not even the worst defensive 1b could do as much damage as Manny, defensively. I find that really, really hard to believe. I don't think Manny is a good defender or anything, but I also don't think he's the worst of all time, nor do I think LF defense is particularly important to the Sox.

[OCD SS]

Well ... I still think the numbers are very controversial. I brought this up in the DSG chat, but I don't think the defensive metrics are using the correct out values in their computation... I believe that the fielder should be charged about .18 runs for the extra plate appearance, but the rest of what happens should fall on the pitcher.

This is something you've mentioned before; why do you think a defensive measure needs to split the credit and debits for each play among all of the participants? Wouldn't that just lead to a broad measure of how well each team pitches and defends (a defensive version WPA)? Isn't that counterproductive when you're trying to figure out how good an individual fielder is (especially with an eye to how good he might be when he switches teams)? The pitchers are essentially the starting point of the data, but other than balls hit back to them they're not what's being measured.

Now Manny may very well be a below average LF er now, but I just do not believe he is the worst in MLB and is costing this team 20 runs per year in the field. Dewans Fielding Bible had a separate ranking for arms in the FB and had Coco rated above Manny among LFers in 2005. How can I believe that stat after watching Coco throw last year? He also ranked Manny last among LFers in 2005, saying he cost the team -31 runs (not sure what his 2006 numbers for Manny were since they are not available now, it must have been worse, because the overall OF defense defense according to his system was -69 runs, worst in the MLB, and 60 runs worse than 2005).

Who do think Manny is better than?

You're using Crisp throwing from CF (with an injured shoulder) to Manny throwing from LF as the basis for your comparison of a previous year? This may point to why your intuition might not align with some of the numbers. Remember that (Dewan's) throwing isn't just a base measure of arm strength, it's really attempting to measure how baserunners treat OF arms. Since Crisp is much faster, he was probably getting to most balls quicker than Manny, and so most runners would see the fielder with the ball and stop. Crisp didn't have much time in LF previously, so opposing coaches didn't have much of a book on him in '05, and might have been a bit cautious. OTOH Manny is pretty slow, and comes with a flaky reputation, so more people were attempting to run on him in '05. His throwing numbers for '03 and '04 show a certain amount of decline, so it looks like team were trying to run on him alot more, and he nailed a few more runners, but teams were still running on him alot. I don't know if there is any ballpark adjustment for his throwing numbers; I can't comment on how Dewan treats a wallball as an opportunity to advance.

[OttoC]

The following tables show the number and percentage of doubles by type--fly ball, ground ball, line drive--that were hit from 2003 through 2006 that were handled by Red Sox and opposing left fielders during games at Fenway Park and in Boston's road games. Additionally, there is a breakdwon for Manny Ramirez for all the categories.

Fenway   FB   GB  LD   Tot
Red Sox  152  42  148  342
Ramirez  121  35  110  266
Oppos.   156  60  162  378
  
Fenway   FB	  GB	  LD  
Red Sox  44.44%  12.28%  43.27%  
Ramirez  45.49%  13.16%  41.35%
Oppos.   41.27%  15.87%  42.86%

	  
Road	 FB  GB  LD  Tot
Red Sox  56  48  123  227
Ramirez  38  39   88  165
Oppos.   46  45  122  213
	  
Road	 FB	  GB	  LD
Red Sox  24.67%  21.15%  54.19%
Ramirez  23.03%  23.64%  53.33%
Oppos.   21.60%  21.13%  57.28%

Because hit location is not available in my data, one cannot determine how many balls were hit of the left field fence in Fenway Park, but the data shows that percentage of fly-ball doubles is nearly twice as high in Fenway Park games than in Boston's road games for both the Red Sox and their opponents. This is obviously a testament to the short left field in Fenway.

We also see that ground-ball doubles occur at nearly half the rate in Fewnway Park as on the road. Again, one can look to the short field in left: the left fielder plays shallower at Fenway Park so batters don't have enough time to stretch the ground-ball singles into doubles. You get a similar effect with line drives.

One also sees that the percentage of fly-ball doubles is higher, both at home and on the road, against Red Sox left fielders than it is against opposing left fielders. There is a small-sample-size caution needed here, but one can be tempted to believe that there is a lack of range for Boston's left fielders. One does wonder why there is a lower percentage of line-drive doubles recorded against Boston's left fielders on the road than by the home team's. In both cases, homw and away, Manny Ramirez has a lower percentage of line-drive doubles than the opposition or other Red Sox left fielders.

It's very tempting to make suppositions, but I'm not sure that the data permits it.

These numbers do not take into account balls that were hit to left but fielded by the center fielder because hit locations were not available in this Retrosheet data. Also, the source that Retrosheet used for play-by-play accounts immediately prior to 2003 do not include batted ball type, so I was limited to the last four seasons for this analysis.

[Rudy Pemberton]

Wow, very interesting. Thank you.

[PedroKsBambino]

The pitchers are essentially the starting point of the data, but other than balls hit back to them they're not what's being measured.

But they affect what's being measured, I think, because they have some control over batted-ball type as well. I think VAL's point is an interesting one (it also came up in the DSG chat if people want to read more there) and I go back and forth on it myself.

[5belongstoGeorge]

It's very tempting to make suppositions, but I'm not sure that the data permits it.

Please, assume away.

A post that contains such interesting information has both earned and demands assumptions!

Your data points to a pretty clear idea, but it would be better if you said it rather than I.

[OCD SS]

But they affect what's being measured, I think, because they have some control over batted-ball type as well. I think VAL's point is an interesting one (it also came up in the DSG chat if people want to read more there) and I go back and forth on it myself.

VAL brought this up in the DSG chat, but IIRC it wasn't really addressed; it's something that's stuck with me and that I've kept thinking about though, so I wanted to get back to it since the opportunity presented itself.

I think that the batted ball type should be adjusted for (along with things like the pitcher's handedness) in the initial data set (which some, if not all, of the PBP metrics do); really all this does is determine how large a data set you have (by removing uncatchable balls, or balls out of the zone, etc). But what Val seems to suggest is to change the run value by giving some fraction of the distributed fielding runs to the pitcher, for example, for giving up a ton of screaming line drives that are hard to catch (if I read him correctly).

Part of the problem stems from the sloppiness with which the term "runs" is thrown around. Part of the issue is that these aren't platonic, runs scored, they're measured as against league average at the position which makes it harder to keep adjustments for the more accepted 3 year standard in mind as the sample set (say AL LFs) changes from year to year (and which could lead to something like UZR+ 100 years down the road when trying to compare different eras/enviroments). (This is part of the reason I'm coming around to Dewan's use "Plays" as opposed to "runs.")

The problem I have is not that Val's argument of shifting some of the credit to pitchers might just lower the expected swing between the extremes (Manny doesn't get docked quite so much and winds up as -10, but someone like Crisp (in Cleveland) would loose a certain amount and find himself at only a +7 *), but that you'd have to adjust for fluky results based on the pitching staff (for example, using the above examples, add Carl Crawford getting an extra boost in LF because he's saving the crappy TB staff a ton of hits). In my mind that's what the PBP metrics are doing, removing and/or normalizing anything that's somewhat fluky and not really germaine to some guy with a glove on his hand running to try and catch the ball.


* These numbers are just WAGs for illustrative purposes.

[Pumpsie]

One assumption you can make about the preceding data is that Manny fields considerably LESS line drive doubles at Fenway than anyone else because he plays the balls off the wall better than anyone else and has a very strong, accurate arm and holds more batters to singles in those situations.

Also there may be fewer line drive doubles against Boston left fielders on the road simply because our pitchers don't give those up against the lineups they face as often as the opposition pitchers do against OUR lineup. It probably has little to nothing to do with fielding.

Defensive stats continue to be the bugaboo of the stathead. How much of the results are due to pitching? How much to fielding? How would it be different if another fielder played the same ball? The evidence is not conclusive, to say the least.

Edited by Pumpsie, 19 March 2007 - 11:38 AM.

[bowiac]

But they affect what's being measured, I think, because they have some control over batted-ball type as well. I think VAL's point is an interesting one (it also came up in the DSG chat if people want to read more there) and I go back and forth on it myself.

DIPS is like Rome - all roads lead there eventually.

Regarding Dial's research, the big question that it seems to me needs to be resolved, is that of the road numbers. MGL has said that per UZR, Manny does most his damage on the road, and that's why even with Fenway park factors (which are significant - it's not like MGL is just taking Fenway numbers at face value), he scores so poorly. When I asked Dial about this, he indicated that Joe Arthur disagreed with that assesment of Manny's defense, but he didn't elaborate. I have yet to get a straight answer for either side however about what the exact home/road numbers are (but maybe they've been posted somewhere where I haven't seen?). Those splits are a pretty big part of analyzing these claims from Dial.

Edited by bowiac, 19 March 2007 - 11:47 AM.

[PedroKsBambino]

DIPS, of course, at this point suggests exactly what I stated---pitchers do have some control over batted ball type and that's why you'd consider them in the analysis, imo. The 'original' DIPS approach that you ignore all BIP has been shown pretty conclusively to be an oversimplification, though certainly a huge analytical step forward overall.

I raised the point about road numbers in the BTF thread previously; no answer as yet.

[twothousandone]

I'm interested in the point made in the NYT article about how not even the worst defensive 1b could do as much damage as Manny, defensively. I find that really, really hard to believe. I don't think Manny is a good defender or anything, but I also don't think he's the worst of all time, nor do I think LF defense is particularly important to the Sox.

If that is true, then the real culprits are Francona/Epstein, right? IF Ramirez is the worst fielder in baseball, he should be DH. Since Ortiz DH's (and has since before Youks and Lowell were on the team) The Sox must have concluded that Ortiz' D will hurt the team more than Ramirez' D.

For those of us who think qualitatively, who is a worse LF (or OF) than Ramirez? Ciff Floyd? Luis Gonzalez? Carlos Lee? Adam Dunn? From the stats it looks like Spezio gets put in the OF a bit, even though he's no good. Is Soriano worse in the OF than Ramirez? I don't think there's that much difference, but I haven't seen those toher guys play all that much. I remember Floyd never seemed to be all that good. I'd gladly take Manny Ramirez over all those guys, but it's maybe close with Soriano (on a 2-year deal).

[PedroKsBambino]

The other issue there is health. It has been oft-speculated that one of the reasons Ortiz has been so much healthier with the Red Sox than he was with the Twins is that he hasn't played the field very often.

One can certainly look at the Minn injury history and wonder if this is truly the case, I think, but the suggestion is out there that the health consideration is another part of it.

[The Gray Eagle]

Maybe the staying healthy equation needs to be tweaked to account for Ramirez's health, not just Ortiz's. Ramirez is the one who missed a month last year and had knee issues all year, and he's the one who's older and plays a more physcally demanding position.

Maybe we'd be more likely to keep them both healthy if Ortiz played more games at first and Ramirez DH-ed more.

[OCD SS]

One assumption you can make about the preceding data is that Manny fields considerably LESS line drive doubles at Fenway than anyone else because he plays the balls off the wall better than anyone else and has a very strong, accurate arm and holds more batters to singles in those situations.

This is partly covered by things like the arm ratings in Dewan I referenced above (I think mgl has a component for throwing in UZR as well). FWIW I don't think your analysis quite holds up given those numbers; Manny doesn't have a "very strong, accurate arm", he has an average arm with a quick release (that is pretty accurate) that's being measured in throws from a shallow OF. Most of this idea that Manny has a great arm and takes away a lot of doubles comes from his high number of baserunner kills in '05, a number/season that so far looks pretty anomolous in his career. It's also subject to the "web gem effect", where we remember him doing that alot without really double checking the numbers, which I don't think bear out this assumption.

I think there's probably a lot more balls that hit low off the base of the monster (and were catchable by a better fielder) that turn into doubles because Manny is playing the carrom than there are line drives that he keeps to a single that another fielder also couldn't keep to a single.

Also there may be fewer line drive doubles against Boston left fielders on the road simply because our pitchers don't give those up against the lineups they face as often as the opposition pitchers do against OUR lineup. It probably has little to nothing to do with fielding.

Philly listed numbers in the BTF thread that show how the Sox staff ERA+ shifted from really good (~130 ERA+) in '05 to below average in '06 (~98 ERA+). This is one of the reason to normalise for pitching staffs when measuring defense, which is something the PBP systems attempt to do. The numbers aren't based simply on "total chances", and it doesn't mean you get to throw out the data because you don't like it.

[wade boggs chicken dinner]

DIPS is like Rome - all roads lead there eventually.

Regarding Dial's research, the big question that it seems to me needs to be resolved, is that of the road numbers. MGL has said that per UZR, Manny does most his damage on the road, and that's why even with Fenway park factors (which are significant - it's not like MGL is just taking Fenway numbers at face value), he scores so poorly. When I asked Dial about this, he indicated that Joe Arthur disagreed with that assesment of Manny's defense, but he didn't elaborate. I have yet to get a straight answer for either side however about what the exact home/road numbers are (but maybe they've been posted somewhere where I haven't seen?). Those splits are a pretty big part of analyzing these claims from Dial.

I posted a link to what Joe Arthur had found in the "What's Manny Really Worth" thread; the link appears to be broken. But assuming I was lucid at that point, here's how I summarized it:

To summarize:

ZR for opposing teams: .865 away / .645 Fenway
Manny's ZR (pre-All Star Break): .865 away / .680 Fenway
Manny's ZR (post-ASB): .722 away / .543 Fenway
Manny's ZR (year): .837 away / .609 Fenway

Joe Arthur speculates (and I agree with him) that this may be evidence that Manny was indeed injured.

One other interesting note: "Manny had significantly more opportunities at Fenway. Not only 550 innings at Fenway vs 481 on the road, but also more strikingly 2.55 BIZ per 9 innings at Fenway vs 1.72 away. The combination meant that he had 63% of his fielding opportunities (as defined by ZR) at home, rather than the roughly 50/50 split we might normally expect."

I'll see if I can't find the original thread again.

[Pumpsie]

On just about any other team in baseball, Manny would be the DH and his worth would be automatically 15-22 runs greater.

So, Big Papi is the "weakest link" when it comes to Sox defense. Manny is supporting Papi's numbers, not only by batting behind him in the lineup but by playing the field and allowing Papi not to run himself down or get injured. And, conversely, Manny makes himself more susceptible to his own knee and hamstring issues by playing the field day in and day out.

So, when we calculate Manny's "worth" to the Sox, that has to be taken into consideration, I think. The front office has concluded that Papi is more fragile than Manny and/or a worse defender.

[OCD SS]

Something I brought up at the tail end of the BTF thread kind of dovetails with this; namely that the DH is one of the last places that offers an ineffeciency that is not being exploited by teams when the construct their offense.

Most teams use their DH spot to rest guys and still get them some ABs while not subjecting them to the rigors of playing the field. OTOH a few teams have found an incredible hitter who do not appear to suffer from the "DH penalty" common to most hitters (from The Book, IIRC). Papi and Hafner are the most notable examples, but other teams are starting to catch on: the A's with Thomas, and trying again with Piazza, the Jays by signing Thomas, and the ChiSox by aquiring Thome. The question is what is worth more, finding a player who can hit at DH without penalty (turning it into a legitimate position in my mind) or keeping the roster flexibility that lets a team move a slightly tired injured rotation through the DH slot, but otherwise gives most of those ABs to the Phil Nevins of the world.

As it applies to the Sox (and all the talk that Ortiz should play 1B and Manny should DH more), Manny does suffer the DH penalty (to the tune of about 40 points of OPS over both his career as a LF and in the last 3 years), but Ortiz is less conclusive. Currently his 3 year numbers as a 1Bman and DH are exactly the same, but when I looked into this last year (using 3 year numbers) the split was closer to Manny loosing 100 points as a DH and Ortiz loosing 100 points as a 1Bman. Clearly this stems from a SSS with regards to the number of PAs each gets, and the fact that their career trajectories are very different. The fact that the Sox use of each player is pretty set also contributes to this.

[5belongstoGeorge]

The question is what is worth more, finding a player who can hit at DH without penalty (turning it into a legitimate position in my mind) or keeping the roster flexibility that lets a team move a slightly tired injured rotation through the DH slot, but otherwise gives most of those ABs to the Phil Nevins of the world.

I've always believed the teams that rotate various guys into DH typically downgrade their offense more often than not. If they are not downgrading then the team must have had a poor DH to start with. The Angels throw any old crap into DH to see if it sticks except when Vlad is hurt. Los Halos are not exactly a run scoring machine.

I have a serious question/request:
Could someone either post here or PM me links to the various What is Manny Really Worth ? threads? I'd like to get caught up. There sees to be a fair amount of discussion that took place since the original was posted by the St. Louis Cardinal's guy. I missed the follow up stuff and it evidently won't go away.

Edited by 5belongstoGeorge, 19 March 2007 - 05:32 PM.

[TomRicardo]

If that is true, then the real culprits are Francona/Epstein, right? IF Ramirez is the worst fielder in baseball, he should be DH. Since Ortiz DH's (and has since before Youks and Lowell were on the team) The Sox must have concluded that Ortiz' D will hurt the team more than Ramirez' D.

No, that is not a logical assumption. The decision to have Ortiz DH does not have to have anything to do with Ortiz's D. I think it is more about the fact that Ortiz cannot play 150 games a year on the field. Manny might be poor defensively but he can go out and play 140 games a year on the field.

[cjust]

I’m not sure if the following has any merit (or has been done before), but I was curious what Manny would look like if you did kind of a hybrid of Defensive Efficiency and Range Factor. Something like

(Put-outs + Assists) / (Opposing AB – Opposing K – Opposing HR).

I was thinking that every time the ball is put in play, there is an opportunity for a fielder to create an out. This way, I could see how many times Manny creates an out on defense per ‘opportunity’ as opposed to other left fielders that the Sox faced.

All of the below data is from Retrosheet.

For Manny, I looked at all events where an out was recorded and Manny was playing left field. Last year, that happened 101 times at Fenway and 81 times on the road. During the times Manny was in LF, there were 2149 At-bats, 411 K’s, and 61 HR at Fenway. There were also 1881 At-bats, 371 K’s, and 89 HR on the road.

Manny’s defense at Fenway
101/(2149-411-61) = .0602 outs per ball in play

Manny’s defense on the road
81(1881-371-89) = .057 outs per ball in play

For comparison, the numbers for opposing left fielders…
Opposing LF defense at Fenway
156/(2754-493-83) = .0716 outs per ball in play

Opposing LF defense on the road
182(2865-563-109) = .083 outs per ball in play

Based on these numbers, if Manny was an average LF, he would have made 120 outs at Fenway in his chances and 118 on the road. That is -56 outs in 123 games or -74 outs per 162.

Depending on how you value the opportunity cost of a missed out, that could be a lot or relative little. At .3 per runs per out, that is 22 runs every 162 games.

[Vermonter At Large]

The 0.80 figure is correct, and an illustration is provided here in post#11:

http://www.tangotige...0247.shtml#1011

Tango, I've read this example before and it makes sense at a certain level, but there are a few problems with it. Let's focus on my main one: why do you assign the .30 out value? I think I understand your using that value for your probabilistic work (using league average as the mid point), but in the cumulative model - which your example surely falls into, shouldn't the outval be .11 or thereabouts?

[Vermonter At Large]

The problem I have is not that Val's argument of shifting some of the credit to pitchers might just lower the expected swing between the extremes (Manny doesn't get docked quite so much and winds up as -10, but someone like Crisp (in Cleveland) would loose a certain amount and find himself at only a +7 *), but that you'd have to adjust for fluky results based on the pitching staff (for example, using the above examples, add Carl Crawford getting an extra boost in LF because he's saving the crappy TB staff a ton of hits). In my mind that's what the PBP metrics are doing, removing and/or normalizing anything that's somewhat fluky and not really germaine to some guy with a glove on his hand running to try and catch the ball.
* These numbers are just WAGs for illustrative purposes.

My problem stems from trying to fit everything into a holistic model that works on the individual player level, the team level and the league level, for both runs allowed and runs scored. In my work it is essential that the defensive runs allowed at a team level fit into the overall model both offensively and defensively. From the offensive model, I have thus far only used fielding errors as a measure for defensive runs, and of course I have to do that by assigning average league values (since there is no real statistic available in primary data bases to tell how many errors offensive teams have benefitted from) to the teams in order to generate runs (normally about 50 runs per team in the current era) and that brings theoretical runs scored to within +/- 20 runs at the team level, which is close, but not close enough.

On the defensive side, I am still compiling data (having to compile some categories manually), and there is more promise in that model. Clearly there is more texture to defensive runs than just errors, or at least more than the standard lwts value for an error (.478 runs) allows for, but it's not very much (plus or minus perhaps 15 runs per team). The question of how much of the value of the hits that a team allows can be charged to the defense and not the pitcher may never be ascertained from this level of modelling, but it should be possible to take a look at the deviation and get a rough idea of how much defensive swing there is from team to team. Another thing I will look at is to see how strong the correlation is between Defensive Runs Allowed (through errors) and non-HR batting runs, since it's logical that teams that make a lot of errors would also give up a higher number of non-HR batting runs. These things might bring us into a position where we can at least provide a sanity check for run-based defensive metrics and know whether or not we're in the ballpark. I'm just not there yet.

[paulftodd]

Who do think Manny is better than?

You're using Crisp throwing from CF (with an injured shoulder) to Manny throwing from LF as the basis for your comparison of a previous year? This may point to why your intuition might not align with some of the numbers. Remember that (Dewan's) throwing isn't just a base measure of arm strength, it's really attempting to measure how baserunners treat OF arms. Since Crisp is much faster, he was probably getting to most balls quicker than Manny, and so most runners would see the fielder with the ball and stop. Crisp didn't have much time in LF previously, so opposing coaches didn't have much of a book on him in '05, and might have been a bit cautious. OTOH Manny is pretty slow, and comes with a flaky reputation, so more people were attempting to run on him in '05. His throwing numbers for '03 and '04 show a certain amount of decline, so it looks like team were trying to run on him alot more, and he nailed a few more runners, but teams were still running on him alot. I don't know if there is any ballpark adjustment for his throwing numbers; I can't comment on how Dewan treats a wallball as an opportunity to advance.

As I am just 1 voice I will defer you to Tangos 2006 Scouting Report for who others rank below Manny.

I do not know exactly when Coco injured his shoulder, but his throwing was crap all year from what I could see. He couldn't hit the cut-off guy most times which is what any weak armed CFer must do. Manny has a far superior arm, he does not have to throw from CF or for Coco to throw from LF for me to compare his arm to Cocos. While not a cannon he gets rid of the ball quick and it is accurate. Coco was moved from CF in 05 and his reputation for a poor arm was well known, despite that he was ranked quite highly in LF . Unfortunately Coco has a LFers arm with a middle IFers bat playing CF. Recently I saw someone, maybe Adam Stern, tag up on a ball hit to CF that Coco caught without having to go back on, in other words, not very deep. Coco's throw was so weak, it missed the cut-off and must have bounced 40 ft in front of 3B, and rolled weakly the rest of the way. Mirabelli could have tagged up and made it standing up. Hopefully his shoulder has recovered and his throwing improves.

[paulftodd]

Well ... I still think the numbers are very controversial. I brought this up in the DSG chat, but I don't think the defensive metrics are using the correct out values in their computation. There are essentially two sets of outvalues that can be used with lwts, depending on if you are using a probablisitic (Base Runs or Batting Runs) model or cumulative (Runs Created) model. In my view, they should be using the smaller cumulative value (about -.11 runs per out) because UZR is a cumulative statistic. Why they continue to use the probabalistic (-.30 runs per out) value is beyond me. They also calculate the value of a defensive miscue as -.81 runs, which I think is absurd. This not only includes the higher probabalistic out value, but charges the fielder about .50 runs for runs that the ensuing batters are likely to get from that non-out. I believe that the fielder should be charged about .18 runs for the extra plate appearance, but the rest of what happens should fall on the pitcher.

At any rate, in addition to the assymetrical ballpark zone problems brought out in the chat, I think that the run penalization the MGL uses for UZR is about twice what it should be.

The cost of an error, or not being able to make a catch that the average defender would make, is not just the cost of not getting the out, which is 0.3 runs, but also the effect of the the error or non play, which is that you will have an additional ROB, plus any ROB before the play will advance, so if a single is worth 0.5 runs, then the difference between making the play and getting the out and not getting the out and allowing a man to get on base, is 0.8 runs.

The value or cost of an out will obviously change depending on the base out state. With 2 outs and no one on base, the value of the out is 0.116. With bases loaded and 2 out, the value of the out is 0.798 runs. The average cost of the out looking at the cumulative of all the various base outs states in all games played from 1999-2002 was 0.3 runs, so this is what is used.

This is my understanding of what is being done at least, which may or may not be right

On the defensive side, I am still compiling data (having to compile some categories manually), and there is more promise in that model. Clearly there is more texture to defensive runs than just errors, or at least more than the standard lwts value for an error (.478 runs) allows for, but it's not very much (plus or minus perhaps 15 runs per team). The question of how much of the value of the hits that a team allows can be charged to the defense and not the pitcher may never be ascertained from this level of modelling, but it should be possible to take a look at the deviation and get a rough idea of how much defensive swing there is from team to team. Another thing I will look at is to see how strong the correlation is between Defensive Runs Allowed (through errors) and non-HR batting runs, since it's logical that teams that make a lot of errors would also give up a higher number of non-HR batting runs. These things might bring us into a position where we can at least provide a sanity check for run-based defensive metrics and know whether or not we're in the ballpark. I'm just not there yet.

I think you are right about there being a run allocation issue (must be that intuition of yours ). Defense includes pitching plus fielding. You can not charge the fielder and pitcher the same run, or the same amount for the change in the run expectancy for a given play, that is double counting. The fact that the pitcher allowed the BIP should be counted against the pitcher to some extent, and the type of ball that is put in play, LD, GB, FB should determine how much to charge the pitcher, in decending order. LD's are much more likely to be hits than GB or FB and the fielder catching the LD is based more on luck or positioning than skill, while GB and FB are more skill dependent, where a players speed, jumps and routes to the ball determine his ability to get to and make more plays than the average fielder. So maybe the fielder is 25% responsible for the LD, 50% for the GB and Fliner, and 75% for the FB (no science here, just as an example). This may be very well why the defensive runs for a supposedly terrible fielder or team like Manny (-31) and the Red Sox 2006 (-72) seem so out of touch with "my" reality.

But no matter what adjustments are used, until the defensive gurus can address the Fenway Factor in a more effective and transparent way, I am not taking them too seriously for Red Sox LFers.

[Vermonter At Large]

The cost of an error, or not being able to make a catch that the average defender would make, is not just the cost of not getting the out, which is 0.3 runs, but also the effect of the the error or non play, which is that you will have an additional ROB, plus any ROB before the play will advance, so if a single is worth 0.5 runs, then the difference between making the play and getting the out and not getting the out and allowing a man to get on base, is 0.8 runs.

The value or cost of an out will obviously change depending on the base out state. With 2 outs and no one on base, the value of the out is 0.116. With bases loaded and 2 out, the value of the out is 0.798 runs. The average cost of the out looking at the cumulative of all the various base outs states in all games played from 1999-2002 was 0.3 runs, so this is what is used.

This is my understanding of what is being done at least, which may or may not be right

This is true in the present tense (probalistic model), but in the cumulative model, we just throw those totals into a big bucket and come out with an average run value for an error (.478 or .461 runs) which are just the average values for every error made in the past tense, independent of the situational things. Remember that recorded errors include mostly one-base throwing errors (roughly .30 runs per base), but that average also includes decreasing amounts of catching errors, the worst of which would be a ball bouncing off a fielder's glove into the stands for a HR (roughly 1.4 runs). So in the cumulative model, that presumed out isn't part of the overall run value for the error. Outs are independent (negative) entities in the cumulative model.

This is far from ideal, of course, but I have no way of knowing what the actual breakdown of each individual error was in 1926. One way of making it more accurate would be to calculate the actual average value of an infield error (probably somewhere in the vicinity of .35-.40 runs) vs an outfielder error (probably in the vicinity of .50-.55 runs) and look at team individual errors for actual totals, but this is probably as close as you can get historically. Ideally, given accurate and detailed PBP data, you could accurately assess the cost of each fielding error in terms of runs, and I hope this is being done. While doing that, though, I would hope that the PBP compilers in the present tense could actually break down the other events and assign all events some proportional responsibility on a sliding lwts value scale.