How much of the inefficiency from last year was in the OF? I realize Gonzalez was probably over-rated (and seemed to get worse after his injury), and Loretta was definitely over-rated (and just not very good), but Youkilis seemed adequate and Lowell was quite good. So I thought the IF was decent all-around, while no one in the OF seemed above-average, and Manny and Wily Mo and probably Crisp were all below average (right?).

Gonzalez to Lugo is clearly a downgrade. But Pedroia nd Drew, I think, will be serious upgrades.

here's hoping Crisp finally gets the hang of centerfield, though I have my doubts he is ever going to be even adequate there.

Personally, I think Ellsbury is a better player right now, not just 2 years from now.

I was expecting Drew to be a serious upgrade until I saw these numbers, now I'm not so sure. You seem pretty sure, kevin -- what's your thinking there?

And to all, what's the general view on Pedroia's defense?

THT's zone rating appears to be very different from Stats'. These are the Stats numbers-

2006: Nixon .873, Drew .891

Drew's career ~.900 Stats ZR is well above average. His .665 THT ZR from 2006 is also solidly above average - it's just that Nixon had a crazy good year last year by the THT numbers.

I have no idea what to do with any of that, but there it is.

Toby, Fenway is as kind to righfielders as it is cruel to leftfielders. We will see a spike in Drew's zone rating, once he figures out the weird angles out there.

Pedroia is going to kick ass defensively. He's a huge upgrade over Loretta.

Pedroia's zone rating was .857 according to ESPN. If he can keep that up, he'll be a GG candidate.

Courtesy of the
2007 THT Season Preview, here are the projections for Red Sox starting fielders, compared to how the Sox did at that position last year. One note on the -54: That figure is not adjusted for park, and thus should probably be more like -35. Ramirez's projection is also not adjusted for park, and if it was, it would likely be around -10.

Youkilis +9 (0)
Pedroia -5 (-12)
Lugo -2 (+17)
Lowell -6 (+8)
Ramirez -24 (-54)
Crisp +4 (-20)
Drew +6 (+10)

So in total, the Red So project to be around -18 runs this season, versus -51 last year. I'd say the defense is improved, if only because Crisp isn't injured, and has more experience playing center field in Fenway.

Those are ridiculous projections for Pedroia and Lowell. There's no rational basis for such a negative rating, especially for Pedroia.

Likewise, there's no rational basis for such a positive one for Crisp.

THT's zone rating appears to be very different from Stats'. These are the Stats numbers-

***

What we do is show is how many plays a player made on balls in zone, divided by the number of balls in zone, and break out the plays he made on balls out of zone. Stats zone rating is simply play made on balls in zone plus plays made out of zone, divided by balls in zone plus plays made on balls out of zone. Our method is superior, if only because it shows more data (it's also better because STATS zone rating rewards sure-handed fielders while punishing rangy guys). To calculate runs above average, I use plays made divided by balls in zone, which avoids the STATS issue.

The other difference is that BIS uses smaller zones than STATS, which allows for a more precise definition of balls in and out of zone.

The other difference is that BIS uses smaller zones than STATS, which allows for a more precise definition of balls in and out of zone.

Well, doesn't that ALSO punish rangy guys, who have an expanded zone because they can cover more ground?

Those are ridiculous projections for Pedroia and Lowell. There's no rational basis for such a negative rating, especially for Pedroia.

Likewise, there's no rational basis for such a positive one for Crisp.

***

Pedroia, I tend to agree with, because we have just 213 innings worth of information for him. He's probably better than the projections show. Lowell and Crisp, on the other hand, have long histories, and I stand by our projections for those two. Lowell has a great season last year (+12 plays), but here are his plays (not runs) above average in the previous seasons:

2005 -28
2004 -21

I know he has a great reputation, but prior to last season, the numbers have said something very different (according to The Fielding Bible, Lowell was great on bunts, and not so good with everything else). Hey, it wouldn't be the first time that a Gold Glove fielder turned out to be not so good (Jeter is the best example, but Griffey and Alomar work very well too). I guess it's possible that the Boston coaching staff improved Lowell's positioning, but by that much? I doubt it.

Crisp, on the other hand, sucked last season (-15 plays), but was +31 plays in 2005 and +20 in 2004.

And if you believe in observation over numbers, you still have to realize you haven't seen either of these guys much before last season. The numbers confirm that Lowell was good last year and Crisp was not. How would you know how they performed in previous years?

The other difference is that BIS uses smaller zones than STATS, which allows for a more precise definition of balls in and out of zone.

Smaller? Do you mean larger? Or am I missing something really obvious?

Drew's Stats ZR is .890, his THT ZR is .665. There aren't enough balls out of zone in the world to make up for that - THT must be using a much larger zone in right field, as I read it.

Well, doesn't that ALSO punish rangy guys, who have an expanded zone because they can cover more ground?

***

What? No, the zones are the same for everyone. I really have no clue what you're talking about here, but our version does not punish rangy players.

Crisp, on the other hand, sucked last season (-15 plays), but was +31 plays in 2005 and +20 in 2004.

Those are CF numbers for 2006 and LF numbers for 04/05, right? I assume kevin's point is that Crisp had trouble with the new angle on balls and the new responsibilities in CF last year, and that will carry over. I can't say myself, having not seen much of Coco before last year, but it's a reasonable argument.

Those are CF numbers for 2006 and LF numbers for 04/05, right? I assume kevin's point is that Crisp had trouble with the new angle on balls and the new responsibilities in CF last year, and that will carry over. I can't say myself, having not seen much of Coco before last year, but it's a reasonable argument.

***

Actually, Crisp played mostly in CF in 2004, and a little in 2005. In those two years, he was +6 plays as a center fielder, which is like +9 per 150 games. I don't believe that his fielding ratings in left (which are excellent) are any less important than what he's done in center field, but he has shown himself to be a pretty good center fielder in the past as well.

Actually, Crisp played mostly in CF in 2004, and a little in 2005. In those two years, he was +6 plays as a center fielder, which is like +9 per 150 games.I don't believe that his fielding ratings in left (which are excellent) are any less important than what he's done in center field, but he has shown himself to be a pretty good center fielder in the past as well.

Well then, your numbers are different than the ones I saw on major league graphs (or whatever the site is), which had compelling graphics indicating that Crisp has trouble on balls hit straight over his head, which is a real problem for a guy you expect to play centerfield for you.

And we have a major difference of opinion on whether a leftfielder's numbers automaticaly translate to centerfield. IMO, they don't, especially in this case. Crisp was awful in center last year. He takes way too many chances and runs terrible routes.

I don't know why you would think he was going to do any better with that this year.

Well then, your numbers are different than the ones I saw on major league graphs (or whatever the site is), which had compelling graphics indicating that Crisp has trouble on balls hit straight over his head, which is a real problem for a guy you expect to play centerfield for you.

***

David Pinto of Baseball Musings did those graphs, and what Crisp's graph for 2006 shows is that he was slightly (SLIGHTLY) below average on balls hit directly at him, and BETTER THAN AVERAGE on balls hit both to his right and to his left, making him an above average center fielder overall. So the numbers you're citing like him even more than mine do.

David, congrats on a job well done as far as the projections are concerned. The hardcopy of the preseason book arrived at my house yesterday from Lulu and I'm very happy with it - it's a wonderful first effort. My principle gripe is a printing error on the cover that is likely Lulu's fault (it says "HARDBALL TIMES PRESEASON BOOK" in black times roman font, centered near the top of the cover, and somewhat overlaps the great pic of Felix and the cool THT logo in the top right).

David Pinto of Baseball Musings did those graphs, and what Crisp's graph for 2006 shows is that he was slightly (SLIGHTLY) below average on balls hit directly at him, and BETTER THAN AVERAGE on balls hit both to his right and to his left, making him an above average center fielder overall.

That's not what I remember at all. I remember him having below average numbers at CF, and the reason for it was he had really bad numbers on balls hit right at him. The last thing you want in a centerfielder is oen who has problems going straight back or forward on balls, because if he misses it, it's a triple. There is nobody close to help him out.

And sure enough, he had those same problems in 2006. playing mostly CF. I haven't seen any reason to believe he's splved that problem yet. Maybe he has but I to go on what I know rather than what I wish wwere true.

What we do is show is how many plays a player made on balls in zone, divided by the number of balls in zone, and break out the plays he made on balls out of zone. Stats zone rating is simply play made on balls in zone plus plays made out of zone, divided by balls in zone plus plays made on balls out of zone. Our method is superior, if only because it shows more data (it's also better because STATS zone rating rewards sure-handed fielders while punishing rangy guys). To calculate runs above average, I use plays made divided by balls in zone, which avoids the STATS issue.

The other difference is that BIS uses smaller zones than STATS, which allows for a more precise definition of balls in and out of zone.

Ugh. If they use smaller zones (different), then it isn't necessarily superior that their handling of data is better.

If they (BIS) have different zones (smaller according to you), then their "in zone/out of zone" isn't really comparable - thus you cannot (accurately) make a comparative statement as you have.

In additiopn the skepticism on their scoring system on FBs and LDs is very well founded, furthering the warines with which one should view BIS data.

That's not what I remember at all.

***

Are you kidding me? Why not actually, you know, LOOK at the numbers you're referencing?

Here's the chart: http://www.baseballmusings.com/cgi-bin/DisplayCharts.py?PlayerID=1572&fpos=8&year=2006

Here's his PMR rating: http://www.baseballmusings.com/archives/018503.php

Here's how it translates to runs: http://anaheimangelsblog.blogspot.com/2006/11/pmr-centerfielders-here-is-david.html

The system that you are referencing says Crisp was 10 runs ABOVE average in 2006! Your memory is shoddy, Kevin.

If they (BIS) have different zones (smaller according to you), then their "in zone/out of zone" isn't really comparable - thus you cannot (accurately) make a comparative statement as you have.

***

What are you talking about? Yes you can. The BIS zones being smaller means that balls are more accurately labeled as in-zone or out-of-zone. To take it to the theoretical extreme, if you had just one zone, all balls would be labeled as in-zone, which obviously wouldn't be very good.

For a person who has long advocated that it would great to have BIZ and OOZ split up in the data, you seem awful combative when someone else ends up getting their hands on it first.

What are you talking about? Yes you can. The BIS zones being smaller means that balls are more accurately labeled as in-zone or out-of-zone. To take it to the theoretical extreme, if you had just one zone, all balls would be labeled as in-zone, which obviously wouldn't be very good.

For a person who has long advocated that it would great to have BIZ and OOZ split up in the data, you seem awful combative when someone else ends up getting their hands on it first.

Heh. If the zones are not defined the same, then it isn't the same "BIZ", "OOZ". Do you not understand that?

I appreciate your knuckleheaded example, but it is knuckleheaded. Perhaps you'd be better served by defining the zones (as I have done with ZR) and illustrate where teh "superiority" exists.

TIA, DB.

Er, DG.

Want to talk defense? I have been doing considerable work with Manny and the green Monster the last few days, and it appears Manny may be *average*. At worst, just 10-15 runs below average *AT WORST*. the park effect is as great as people suspected.

I did this by going through every pbp for the last two seasons (2006, 2005). Ye, that's not fun, but we really need that rather than some "averaged" PF by looking at innings. I mean, we have access to the video, let's do it right.

Want to talk defense? I have been doing considerable work with Manny and the green Monster the last few days, and it appears Manny may be *average*. At worst, just 10-15 runs below average *AT WORST*. the park effect is as great as people suspected.

I did this by going through every pbp for the last two seasons (2006, 2005). Ye, that's not fun, but we really need that rather than some "averaged" PF by looking at innings. I mean, we have access to the video, let's do it right.

What baseline are you comparing him to? Is it Manny vs. the LFers who he plays against in Fenway?

I say pbp, but I mean, going through every play - and removing balls that were hit off the wall. I need to do 2004, but really, there are 40-50 BIP off the wall per season that have been counted against Manny (or other Boston LF) that should simply be subtracted from his total -
that would be 250-30= 220 - 172/220 = 0.800 or -12 (or so).

Now there is still some work to do wrt removing LDs and FBs, but I am working on it. Joe Arthur deserves an IMMENSE amount of credit for solving this. Okay, he really just amplified what appeared to be the problem, and Anaheim Rally Monkey's matched innings (which is still heavily flawed wrt sample) supported, meant the effect was far more dramatic than I had suspected. I mean 30 runs is a lot, but it is real -

there is *roughly* 0.5-0.6 BIP off the wall per game!!

that needs to be gone over through the video on mlb.tv, but the data is there - we have to mine it - preferably, some SOB from RSN, rather than me.

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.

And the credit really goes to Joe Arthur and teh efforts he made by tracking Manny's home and road performance - thanks, Joe, for putting forth the effort and sharing it widely so we could all change/adapt our focus in research.

Moving forward, we need to look at BAL RF, etc.

BTW, yes, AROM's matched innings and other sysems that made park adjustments based on opponents' efforts is great, but simply counting "Balls of the wall" is FAR more accurate.

Thanks to those contributing to the collaboration, which includes those just saying "there's no way a LF is -50 runs in a season". . Every dissenting voice helps - that's why full disclosure in research like this is so critical. thanks again.

I have to say, I'm really surprised that this team grades out so poorly. I would have thought Drew would be an upgrade, Lugo would be a breakeven, and Pedroia an upgrade.

What baseline are you comparing him to? Is it Manny vs. the LFers who he plays against in Fenway?

huh? I am comparing him to every other LF in the AL. However, I am removing balls that he couldn't make a play on. So his chances drop dramatically.

Darren,
I'll look at team data later and we'll see. Any projection of Pedroia is folly.

I really wonder what kevin thinks of Pedroia's defense/offense/future/hall of fame speech.

that would be 250-30= 220 - 172/220 = 0.800 or -12 (or so).


I'm not following these numbers at all. Is it that Manny had made 202 plays out of a possible 250, then you subtracted out 30, making it 172/220, an .800 ZR, which would mean a -12 R/150? Is that right?

If so, why did you subtract 30 instead of 40-50? And -12/150 is not average, is it?

Are you kidding me? Why not actually, you know, LOOK at the numbers you're referencing?

You're referencing 2006 numbers. I'm talking about the 2005 numbers that Pinto put up. I can't seem to find them now. I can find 2004 and 2006 but not 2005.

Sorry I wasn't clear about that. We had this discussion a year ago, before Crisp was to appear full-time in CF and we were speculating on how well he would do. Matt alluded to that thread.

Nothing I saw of Crisp, including the 2006 graphs, sway me from that initial impression made by those graphs. In 2006, he again had problems at balls hit straightaway.

This doesn't surprise me at all.

Why would it? Nothing could surprize you. You flatline every brain activity test ever invented.

What? No, the zones are the same for everyone. I really have no clue what you're talking about here, but our version does not punish rangy players.

It's quite simple. If you shrink zones, then you are rewarding guys who have little range but are surehanded with the balls they can get to.

You're referencing 2006 numbers. I'm talking about the 2005 numbers that Pinto put up. I can't seem to find them now. I can find 2004 and 2006 but not 2005.

***

The ones that show him to be +18 runs above average?

It's quite simple. If you shrink zones, then you are rewarding guys who have little range but are surehanded with the balls they can get to.

***

You should read more carefully. First, I count both plays in-zone and plays out-of-zone in the numerator, so you're wrong. What the smaller (and more numerous) zones mean is that we have a more accurate number for the number of CHANCES in-zone that the player had.

that would be 250-30= 220 - 172/220 = 0.800 or -12 (or so).


I'm not following these numbers at all. Is it that Manny had made 202 plays out of a possible 250, then you subtracted out 30, making it 172/220, an .800 ZR, which would mean a -12 R/150? Is that right?

If so, why did you subtract 30 instead of 40-50? And -12/150 is not average, is it?

Sorry, Darren, I kind of short-handed it. I am sorry for my sloppiness.

Yes, I suspect the PF for the Green Monster is going to be on the order of 40-50. Last year Manny only played 1031 innings, and I specifically counted 29 FB off the wall (there were more that others were in LF for - Youk and Pena, etc).

No, -12/150 isn't average, but I suspect there is still some flaw in teh LD vs FB measures, and on the order of 13 runs worth, that make it up. As I went through every ball in play to LF against hte Red Sox for 2005 nd 2006, I noticed a significant disparity in FB vs LD based on the BIP. for 2006, there were 50+ LDs off the wall, and 35 FBs off the wall. those numbers were almost exactly reversed in 2005, indicating a shift in scoring paradigm. And some LDs in 2006 were scored, by STATS, as FBs that counted against Manny - probably about 15-17, rendering him "average".

Like I said last October,

This offseason, I am going to try to solve the Fenway park factor. We've seen good new research from other people, and I hope to establish a PF for Manny. The more I read, the more I think Manny may be *average* on defense. I know, I need to give some demonstration for that.

Well, Dan Rosenheck got me off my duff, and I went to the video. thanks, Dan.

At worst, just 10-15 runs below average *AT WORST*. the park effect is as great as people suspected.

That looks about right, Chris. Kudos.

How did you correct the park factor?

First, I count both plays in-zone and plays out-of-zone in the numerator, so you're wrong. What the smaller (and more numerous) zones mean is that we have a more accurate number for the number of CHANCES in-zone that the player had.

Depends on the zone descriptions. It's not necessarily more accurate.

What the smaller (and more numerous) zones mean is that we have a more accurate number for the number of CHANCES in-zone that the player had.

No you don't because you're making assumptions about zones that may not be valid.

Or what Chris said.

How did you correct the park factor?

the best possible way for Manny - I went to the video, vcounted balls he could not catch and subtracted those from his chances.

That indicated he had 30-50 fewer chances than ZR (AND UZR!!!) figured he had. RallyMonkey's "adjusted innings" did a pretty good job of assessing it.

Actually counting is obviously the best route. this is a VERY VERY simple scoring chnge for both STATS and BIS - I can't imagine why they wouldn't have their scorers simply mark a ball hit off the wall. It's poor system design.

the best possible way for Manny - I went to the video, counted balls he could not catch and subtracted those from his chances.

Way to go, man. I can't think of a better way of doing it.

Do you plan to do the same this year? I'd love to get a track record of the zone penalties Fenway imposes.

Crisp graded as just a bit above average in CF in 2004 per PMR, compared to the other CF for whom we have PMR data. Against a multi-year average, he was just a bit below average.

His PMR figures in LF in 2005 were excellent using any threshold.

PMR was also a fan of his performance in CF in 2006.

If I were going just by PMR, which I wouldn't really recommend doing, I would guess that Crisp projects to be in the +5-+10 range defensively in CF for 2007.

This has been mentioned in at least one other thread, but I find the revelation that STATS was counting uncatchable balls off The Wall in the denominator for zone rating is fundamentally insane.

No, -12/150 isn't average, but I suspect there is still some flaw in teh LD vs FB measures, and on the order of 13 runs worth, that make it up. As I went through every ball in play to LF against hte Red Sox for 2005 nd 2006, I noticed a significant disparity in FB vs LD based on the BIP. for 2006, there were 50+ LDs off the wall, and 35 FBs off the wall. those numbers were almost exactly reversed in 2005, indicating a shift in scoring paradigm. And some LDs in 2006 were scored, by STATS, as FBs that counted against Manny - probably about 15-17, rendering him "average".


I'm not sure that I'd make this leap. Just because there were 35 LDs/50 FBs in 05, doesn't mean there would be the same ratio in 06. Different pitchers, different hitters, pure randomness, etc could affect that ratio. I assume that your further review of the data might shed more light on this.

Some other things I would want to look at:

--other American OF play 3-10 games in Fenway as well, so they are going to be very slightly affected by Fenway (very slightly)
--is it possible that other OF are affected by the LD/FB issue?
--would other parks, if examined this closely, have similar--if less pronounced--quirks?

Wow, that grammar was a nightmare. My meaning should be clear, though.

Anyway, we don't have any PMR data for Crisp from 2005.

Darren,
oh, I plan to do the physical count of '04 and '03.

this isn't my first time <wink>.

Oh, and one more thing -- the PMR method has actually changed a bit each season. What you see for Crisp in CF in 2006 is not the same method as in 2004.

Also, wrt to visitors, that's one reason I focus on "Wall balls per inning", and we are looking at 0.5-o.6, so we can easily adjust a AL East visitor appropriately.

there are specific parks that are short - BAL, and a couple pthers. However, what I did *not* find was a large number of BIP affecting BOS CF. less than 20, and fewer in '05. So in cases that are smaller, I would lean toward a minimal PF *to keep the calcualtions simple*!!!!. I would prefer to do PFs just like Offense, but we aren't there yet.

We're MUCH MUCH closer thanks to Joe Arthur's work (and everyone else), and obviously MLB.tv.

I say Joe, because he set off the lightbulb in my head. LAWBH and AROM also contributed heavily to my thought process.

And thanks, kevin - I assure you, this will be continued for '07.

Chris - do you happen to know Manny's road numbers? One of the big things that lent credibility to MGL's claims that the Fenway park factor was being done properly was him explaining that Manny's road UZR is almost identical, if not slightly worse than his park adjusted home UZR.

There seems to be a major disconnect between that, and your findings here, unless your system doesn't grade Manny out at being in the -20 range on the road.

LAWBH-
That's more important than we can stress. Correcting past data is going to be very important, and a good reason to enlist Sean Forman and/or Lahman in control of this type of data.

His PMR figures in LF in 2005 were excellent using any threshold.

But his PMR in center was bad, which is the one I was most concerned about going into 2006.

The ones that show him to be +18 runs above average?

Christ, DSG, those are LF numbers. Those don't count. I want to see the damned CF data from that fielding graph that was discussed so heavily last year. I can't find them. I can't even find the gthread.

But his PMR in center was bad, which is the one I was most concerned about going into 2006.

His 2004 PMR in center was essentially average. His 2005 PMR in center was unpublished. His 2006 PMR in center was very good.

Here are Crisp's graphs for 2004 in CF and LF.

Here's 2006, which is all CF.

Chris, remind me -- wasn't there also some discussion of a similar effect in Minute Maid Park, due to the Crawford Boxes -- or am I confusing it with something else?

One of the big things that lent credibility to MGL's claims that the Fenway park factor was being done properly was him explaining that Manny's road UZR is almost identical, if not slightly worse than his park adjusted home UZR.

There seems to be a major disconnect between that, and your findings here, unless your system doesn't grade Manny out at being in the -20 range on the road.

Joe Arthur's road numbers disagreed with that evaluation of Manny. My evaluation is relatively simple - Manny had 250 chances according to STATS. Of those, I can categorically eliminate 30 chances. So Manny had 220 chances. Manny, based on a .695 ZR, made 172 plays (based on STATS 250 chances). Since I know he made 0 of 30 of the zone chances off the Green Monster, I can simply subtract 30 from 250 and still simply divide by 172 to get his "true" (relative to the rest of the AL) ZR, and thus his Runs Saved (RS). Yes, tweaking the entire league is preferred, but we are subsequently looking at 3 runs for an opposing player. No big deal.

In the end, Manny is not some big disaster area out there and worth 5+ wins over average - I think - and close to being worth his salary.

MUCH MUCH closer than previous evaluations have indicated based on UZR.

No, Srul, that is one of the parks we need to review.

LAWBH-
That's more important than we can stress. Correcting past data is going to be very important, and a good reason to enlist Sean Forman and/or Lahman in control of this type of data.

I concur.

There are, by my count, three different variants of PMR.

PMR-1 is the original iteration.

PMR-2 is a "smoothed visitor" method, which first started appearing in 2005. This only uses BIP against visiting fielders to set the "park model". Read about it here.

PMR-3 takes the smoothed visitor method and replaces velocity of batted balls with distance (or at least mixes it, I'm not sure, it's described here. This was introduced in 2006.

I'm not sure David ever went back and gave us PMR-3 for many of the positions; if he did, I missed it.

His 2005 PMR in center was unpublished.

No it wasn't. We discussed it in a thread last year.

Ah hell, maybe it was the 2004 numbers that we discussed.

Anyway, Blackhawk, I can't see how you can say Crisp was average in 2004 when nearly the entire red line is below the 0.0 threshold.

Oh hell, it looks like David eventually did post what I'm calling PMR-3 figures here for some of the other positions.

Kevin, I wasn't around for the discussion you reference, but I know for a fact that Crisp's PMR figures in CF for 2005 were not released with everyone else's. I don't know if David mentioned Crisp's figures in some other post.

Well, he only payed 10 games in center in 2005 so I don't think the numbers are all that meaningful.

wait, how do you deem a ball uncatchable?
Any ball that hits x feet up on the monster and above?
Can you confidently say that a fielder couldn't have gotten to a ball at catchable height, if it was say 2 feet from his outstretched glove? Might be instinct issues?

Anyway, Blackhawk, I can't see how you can say Crisp was average in 2004 when nearly the entire red line is below the 0.0 threshold.

Compared to a multi-year average, Crisp made 5 less plays than average, which would have come out to around -10 plays over a full season.

However, as a group the CF for which we have data for 2004 come out as below average. If we re-center the average to reflect that, Crisp comes out as slightly above-average (by approximately a run).

I don't believe the multi-year average makes any real sense as a baseline (using a multi-year average, there were only two semi-regular or regular 3B in 2005 who were below average, for instance), and the re-centering I did was an attempt to correct for that. It may have been faulty.

At any rate, I think his performance was likely around average, maybe just a bit below, for that season, as reflected by PMR.

wait, how do you deem a ball uncatchable?
Any ball that hits x feet up on the monster and above?
Can you confidently say that a fielder couldn't have gotten to a ball at catchable height, if it was say 2 feet from his outstretched glove? Might be instinct issues?

These balls hit more than 8 feet up the wall. That's a pretty good estimation.

Another thing re:DSG

If I understand right, you are separating plays as InZone and OutOfZone. I guess you will have more credit for plays made OOZ and less negative credit for plays not made OOZ and vice versa for IZ.
If this is your methodology, does the size of zone matter if you normalise your data? Every fielder would have made X plays above/below average per 100 chances in IZ, and Y plays above/below average per 100 chances OOZ..

These balls hit more than 8 feet up the wall. That's a pretty good estimation.

agreed, was just clarifying.

IIRC, Baltimore, Boston, Minnesota, Colorado, Florida, Houston, and I think maybe San Fran. and Pittsburgh all have at least one part of the OF that appear to need a PF or the same treatment Fenway is getting.

Btw, the threshold thing is another problem we have with looking at PMR over several years.

For reasons mentioned in my previous post, I don't think using a multi-year average makes a lot of sense, and 2005 3B is really the poster boy for that. There are 39 3B listed there, and 37 of them performed better than the multi-year average, and the 39 add up to being 772 plays better than the multi-year average. I don't think there's any way in hell that all the other 3B who played in 2005 added up to being 772 plays below average, so I can only assume that 3B as a whole were well above-average. This, of course, makes no sense.

But I worry that my re-centering is insufficient. In 2006, Pinto ditched the multi-year average, so I didn't do any re-centering. But the sample we have for each position (generally around 40 players) still doesn't add up to zero (though I don't recall any position being as out-of-whack as the 2005 3B situation). We shouldn't really expect it, too, but that makes me question the accuracy of the re-centering I did in 2004 and 2005. I don't really know what the solution is, given the data we have.

I think Petco RF is another position with some goofy things going on.

IIRC, Baltimore, Boston, Minnesota, Colorado, Florida, Houston, and I think maybe San Fran. and Pittsburgh all have at least one part of the OF that appear to need a PF or the same treatment Fenway is getting.

I think Dial is arguing that the PF for the Monster is more significant because of the # of chances. 0.5+ chances per game is pretty sizable IMO

Baltimore, Boston, Minnesota, Colorado, Florida, Houston, and I think maybe San Fran. and Pittsburgh all have at least one part of the OF that appear to need a PF or the same treatment Fenway is getting.

Kyle, agreed. However, I think the consensus is that the GM (Green Monster) is by far the largest problem. And it's ~45 plays. I would suggest it is *twice* the problem other parks face. Possibly the Crawford boxes presnt a similar problem, so we need to start counting.

I think we can establish a PF for them with, say, 4 years of "BIP that are not catchable", and use that, on a per inning basis, to adjust all players.

It will result in a factor that is close enough to adjust over history, and still be within a run or two of an exact seasonal evaluation, rendering an "every season measurement" meedless.

But we have to do it. Who wants to take 2003?

Kyle, agreed. However, I think the consensus is that the GM (Green Monster) is by far the largest problem. And it's ~45 plays. I would suggest it is *twice* the problem other parks face. Possibly the Crawford boxes presnt a similar problem, so we need to start counting.

Absolutely. I think it'd be interesting to count for all those parks if we could, of course I wish I just had the money so I could pay STATS to calculate ZR as I specify...

But you can do it yourself, Kyle. I mean, go to mlb.com, go to every home game of these teams and count the balls that are drawn as being on the wall. check to see if they are LDs or FBs. You can do this. then do it for 05 and 04. Each game takes about 40 seconds to open and evaluate on average, so it only takes an hour or hour and a half for each team season.

Honest moment: one reason I reveal all my techniques and formulas is because I am lazy. I want someone else to do teh work after I figure out how it can be done. SG in ATL and AROM pretty much shoed me down in doing all my calculations already, so I'm "teaching you to fish" rather than giving you a fish.

Hey, speaking of SG, did he EVER publish the ZR stats that you guys were accumulating?

Hey, speaking of SG, did he EVER publish the ZR stats that you guys were accumulating?

If he hasn't, I know he's working on it because he had me send him a couple years of data about a month ago.

Chris, I'll have to do that once I get a new job, which should be within a week or two. The schedule I'm on now drains me of all my energy.

I am working with Dan to publish them all

Nothing I saw of Crisp, including the 2006 graphs, sway me from that initial impression made by those graphs. In 2006, he again had problems at balls hit straightaway.

Never have I seen an outfielder play so many singles into triples.

If I understand right, you are separating plays as InZone and OutOfZone. I guess you will have more credit for plays made OOZ and less negative credit for plays not made OOZ and vice versa for IZ.
If this is your methodology, does the size of zone matter if you normalise your data? Every fielder would have made X plays above/below average per 100 chances in IZ, and Y plays above/below average per 100 chances OOZ..

***

Sorry, I don't know what you're asking...

Forgive me for being an ignoramus on this point, but why is the Manny/Fenway thing so hard to figure? Can't we figure out how good a fielder Manny is in the abstract by looking solely at how he does on the road? And can't we figure out how much of an effect the Monster has in the abstract by looking at how all visiting LFs do at Fenway?

That means looking at only half the sample, sure, but that's still useful, isn't it?

Wow - really interesting discussion, I am happy I wandered in here.

Is it possible that if one adjusts ZR for the Monster that they will start to overrate Manny? I am sure this has been brought up before, but does the LF zone at Fenway start the same distance from homeplate as it does in other stadiums (I assume so)? If so, is that likely helping the LFer's rating as he can play closer to HP and catch more OOZ balls then he could in another park and (after this new adjustment) not be penalized for in zone wall balls.

I admit this over-correction will likely be much smaller then the previous Fenway LF problem, but it is something to think about.

Good point TH. Something to add into the mix for sure.

Forgive me for being an ignoramus on this point, but why is the Manny/Fenway thing so hard to figure? Can't we figure out how good a fielder Manny is in the abstract by looking solely at how he does on the road?

The problem is that fielding splits aren't readily available, IIRC. I think JoeArthur's work attempted to do just what you're saying but it was only one year of data.

We know batters do better at home than on the road in general (though, that effect may vary by batter). Therefore, a batter is not an "x" hitter b/c he hit "x" on the road. I can't think why the same wouldn't be true for fielders (perhaps even magnified). Additionally, throwing out 1/2 the dataset is never a good idea. But, here it is even worse b/c of the variability int he numbers with a large dataset (home + away).

Since I know he made 0 of 30 of the zone chances off the Green Monster, I can simply subtract 30 from 250 and still simply divide by 172 to get his "true" (relative to the rest of the AL) ZR, and thus his Runs Saved (RS). Yes, tweaking the entire league is preferred, but we are subsequently looking at 3 runs for an opposing player. No big deal.

But, Chris, aren't you taking the bad away from Manny b/c of the wall, but leaving in credit for the good? So, not only are we just removing the impossible zones from his ZR, but we are leaving in credit for the easier zones in which the wall helps (e.g., shallow hit balls where he plays more shallow b/c of the wall)? I don't know if that would help him or not, but for a layman, it seems to be a problem, no?

Additionally, throwing out 1/2 the dataset is never a good idea.

If the questions are "How good is Manny outside Fenway" and "How good are left fielders other than Manny at Fenway", the data I'm asking about is the entire dataset.

Well, Toby, the other leftfielders play half their games somewhere else and Manny plays half his games in Fenway. I would think ALL leftfielders play better at home. So that should be included in the data. If Manny is being burned more by his home park, in that his homefield advantage is being upended by the peculiarities of the Wall, then I think the best thing to do is to try to account for it, rather than just throwing all the data away.

#84-But, on the other hand, wouldn't some more batted balls shoot by on the left and the right of the Fenway leftfielder, because he's closer to the hitter, and has less time to move laterally before the ball gets past his outstretched glove hand?

But, Chris, aren't you taking the bad away from Manny b/c of the wall, but leaving in credit for the good? So, not only are we just removing the impossible zones from his ZR, but we are leaving in credit for the easier zones in which the wall helps (e.g., shallow hit balls where he plays more shallow b/c of the wall)? I don't know if that would help him or not, but for a layman, it seems to be a problem, no?

As noted, Manny presently doesn't appear to see significant improvement from that. I don't think this would overrate Manny to a significant degree (if at all). While the possiblity exists, I would surprised to see it as significant - FBs that get caught there are likely to be caught anyway, and I don't see Manny as catching as many LDs that are "short".

Toby,
I believe that is how mgl adjusts for the Monster.

As noted, Manny presently doesn't appear to see significant improvement from that. I don't think this would overrate Manny to a significant degree (if at all). While the possiblity exists, I would surprised to see it as significant - FBs that get caught there are likely to be caught anyway, and I don't see Manny as catching as many LDs that are "short".

But maybe an average fielder would?

The problem I have, Darren, is that I'm not sure they are positioned closer to home by any significant amount, *and* the reason LDs aren't a huge portion of ZR is they have to be hit close to yu. I don't think an average LF would catch "a lot" more.

Maybe they would, but I don't think there's any evidence at this time to support that position.

I doubt it would be "a lot" as well.

I doubt it would be "a lot" as well.

Right, but the current problem with the Monster is that it colors the defensive eval a lot.

I see that. Did I sound like I was arguing that the positioning issue would completely negate the balls off the wall?

Darren, not at all, but three of you were worrying about it (you, TH and Josh).

FWIW, there is a parallel discussion going on at S.o.S.H. here: http://sonsofsamhorn.net/index.php?showtopic=16462
In addition to the zone discussion you guys started here, we've been talking about the run values MGL uses for UZR.

I don't see anything new there. that 0.11 argument is the same one Tango presented to me at Fanhome in 2000, and I explained the 0.3 to him.

I admit this over-correction will likely be much smaller then the previous Fenway LF problem, but it is something to think about.

Just from a smell test standpoint, I have to think it has already happened. -10 to -15 runs I can see as that puts him right in line with similar body types at similar ages, but Dial's claim that there is another 13 runs of error out there and that Manny is average just doesn't make sense to me. Think of it this way...how many 34 year old slugging outfielders suffering through a knee injury are average defensively? Was Manny, at his defensive peak in RF, so good that you could even remotely expect this to be true?

Dial's claim that there is another 13 runs of error out there and that Manny is average just doesn't make sense to me.

That's not my claim. I don't claim that.