Chris-

This is the best description of Zone Rating that I’ve read. Many surprises and questions.

My main surprise is how much of the field is covered. I had the impression that there were big gaps, but as far as I can tell the gaps seem quite small:

One ground ball slice between first and second, and between short and third.

Two ground ball slices “up-the-middle” between second and short. (How do pitchers figure in that?)

One slice each along the foul lines for each corner outfielder.

One slice in each of the outfield “gaps”, but only for “line drives”.

In my own studies of ZR and UZR, there is usually a pretty good agreement in the infield, but many large differences in the outfield. It seems to me that the differences may be because of the comparatively large _depth_ gaps in the outfield.

UZR keeps track of the rate of outfielder plays made in shallow areas. This is problematical, because so many shallow fly balls can also be caught by infielders. The Andrew Jones ratings under UZR (and other systems, including DRA) are no doubt too high because he has taken many discretionary chances from middle-infielders. The same problem might also apply for weak outfielders, who might play deep with the understanding that infielders on the team will take whatever flyballs they can (perhaps that explains the Bernie Williams UZR ratings and the high number of flyballs that Jeter catches).

It seems to me that the easiest fix would be to include all _line-drives_ hit into the currently included outfield zones _regardless_ of depth. To the extent that Andruw Jones-types genuinely catch short line drives, then he’s saving runs.

The other way ZR could be improved is by counting _all_ ground balls and line drives, regardless of whether they are hit in ‘greater-than-50%-fielded’ zones. A zone in which 40% of batted balls are fielded successfully is probably _precisely_ the type of zone in which good fielders distinguish themselves from mediocre fielders. Are there any key zones that fall just short of the 50% threshold?

Is it possible to buy ZR raw data that includes this “missing” information, so that people can create their own zone ratings? How much would it cost? Such ratings would have a level of detail and accuracy somewhere between plain ZR (with simple run-weights included) and a theoretically perfectly calculated UZR, while being quite simple.

A few other thoughts—

What do you think of the use of park factors in UZR? My sense is that they increase rather than decrease distortions, and should only be applied in very extreme cases, such as Coors Field and Fenway.

How accurately and consistently is the coding of batted balls as line-drives v. flyballs, or as to placement of batted balls? What kind of quality controls does STATS, Inc. apply?

Thanks again for a great article.

Thanks Chris, this is a great article. I don't suppose you (or anyone else) have any plans to write equally excellent articles about the other major defensive metrics?

As MAH said, my main reaction is that I didn't realize how much of the zones are in a fielder's responsibility. That laser right over the bag and into the OF corner?

That's in the 1B or 3B zone.

For infielders at least, groundballs out your zone can't account for too many plays, unless your positioning is different from average.

So I guess infielders who miss ground balls in their "zone" due to their coaches incorrectly shading them a certain direction get unnecessarily penalized.

Is it possible to buy ZR raw data that includes this “missing” information, so that people can create their own zone ratings? How much would it cost?

Upwards of $5k a season. A ton.

What do you think of the use of park factors in UZR? My sense is that they increase rather than decrease distortions, and should only be applied in very extreme cases, such as Coors Field and Fenway.

That is my general feeling, but I don't have the data to really say. Both MGL and Mike Gimbel believe in PFs and there are PFs for strikeouts, so I believe that it may be necessary. I'd be surprised if most PFs weren't 98-102, so except for very extreme cases it won't amtter at all (save for trying for extra precision).

How accurately and consistently is the coding of batted balls as line-drives v. flyballs, or as to placement of batted balls? What kind of quality controls does STATS, Inc. apply?

Very. Multiple scorers - now many with tiVo, and every game on MLB.tv, plus review when scorers disagree. I used to get a report card on my % correct.

my main reaction is that I didn't realize how much of the zones are in a fielder's responsibility.

Hopefully this will help clarify *why* I don't think ZR is so flawed.

So I guess infielders who miss ground balls in their "zone" due to their coaches incorrectly shading them a certain direction get unnecessarily penalized.

that is possible, but how common do you believe it to be? And how much of an impact (how often do they get incorrectly positioned, vs how often they get a ball they wouldn't have because of the re-positioning)?

that is possible, but how common do you believe it to be? And how much of an impact (how often do they get incorrectly positioned, vs how often they get a ball they wouldn't have because of the re-positioning)?

Well I was thinking if they were re-positioned for a pull/slap hitter and it did help them get to the ball, some of the time they would be getting to a ball not in their zone and therefore not improving their ZR. This is just a minor point though. With the amount of chances most infielders get each season, they probably lose/gain a negligible amount of ZR points due to unique positioning.

some of the time they would be getting to a ball not in their zone and therefore not improving their ZR.

That does improve a player's ZR. Just not as much as it should.

That does improve a player's ZR. Just not as much as it should.

Oops. I misread that paragraph. Sorry.

Chris, thanks for the nice, clear explanation of UZR (there are a few typos, if you are interested).

I've been thinking a little bit about ZR and UZR, especially with regard to fielder positioning, which is major obstacle in putting together a good PBP fielding metric.

There are too many "defense" threads going on, BTW. At least this one should stay afloat for a while, as opposed to those in "Primer" or whatever "blog" it is called now (is it just me, or is the organization and layout of BTF really confusing?).

Anyway, with regard to the "metits" of ZR, I don't think "flawed" is the right word. It is what it is, and has its limitations. It is a coarse metric as compared to UZR for obvious reasons. Those reasons are mostly that it does not distinguish between various "speed" balls and that it does not distinguish among the location of balls within a zone. A player could have a "false" good ZR if he has a lot of balls hit right at him within a zone, or has a lot of balls hit at the "edges" of his zone. As well, the balls hit outside of his zone are not handled real well.

Speaking of, I agree with Chris that balls fielded outside of a player's zone should probably not be counted in the denominator, but for a different reason. Since a player should probably get "extra credit" for those outs, one way to do that is to not count them as opportunities. Either way though (counting them or not), the "math" is screwed up (it "bastardizes" the definiton of ZR, which is essentially balls fielded divided by opps).

Other obvious drawbacks are that it doesn't account for other contexts (besides the speed of the balls), like runners on base and outs, (which affect the positioning of the fielders), and handedness of the batters (which also affects the positioning and the speed of the batted balls).

Anyway, the nice thing about ZR in terms of its weaknesses (that is a better word than "flaw," as flaw suggests that a "mistake" was made in constructing its methodology) is that most of those wekanesses tend to "even out" as the sample of data gets larger (more innings). That is nice. With other coarse metrics that is not always so. For example, BA is a coarse metric as compared to OBA or OPS, but the weakness in BA does NOT even out in the long run because of player bias.

As far as park factors, I do very little park adjusting other than in unusual parks and zones, like Coors Field and LF at Fenway. The way I do that is to regress the sample park facors quite a bit unless I "know" there is a good reason for an unusual factor (as in the above examples).

So I guess infielders who miss ground balls in their "zone" due to their coaches incorrectly shading them a certain direction get unnecessarily penalized.

Here are some thoughts re: positioning which, as I said, is the major obstacle in any PBP defensive metric. If the BIP distribution is typical, or at least supposed to be typical, then if a player is playing in an atypical position for whatever reason, then if that "nicks" his R or UZR, then he deserves to have it nicked. IOW, ZR or UZR, without adjusting for fielder position, is a combination of "true" fielding performance plus position. If a fielder is "out of position," that's too bad. I don't mean that literally - I mean that in the context of defining UZR or ZR as "positioning plus fielding performance."

Of course what we really want to do is measure actual fielding performance and ultimately (as the sample sixe increases to get rid of the luck) fielding skill, and NOT the result of atypical fielding performance. The problem is that we don't have a record of fielder positioning. If we did, then it would be a snap to get a real picture of fielding performance (using a methodology like UZR). That would not get rid of all the problems in UZR, the biggest one (other than fielder positioning) being discretionary plays.

Anyway, if a fielder positions himself atypically, and for good reason (like his pitchers for some reason allow an unusual BIP distribution - whether that exists on any significant level, I don't know), then his UZR (and I think ZR) is generally going to be overstated. For example, that is what was suggested with reagard to Swisher's UZR. I don't know that there is anything that can be done to account for this. You can look at the BIP distribution and if it is unsual AND a fielder has an unusually good rating, there is some sugestion that he may have been playing in an atypical position, but by no means is this a foregone conclusion. There are going to be many instances where a fielder is playing in a normal position, making some great plays, and there happen to be lots of balls hit in or near those zones in which the great plays are being made.

That's all I have for now...

then his UZR (and I think ZR) is generally going to be overstated

I disagree with this. What you're saying here (I think--correct me if I'm wrong) is that his "true talent" will be overstated. His UZR is, by definition correct. It's not important whether that is a result of positioning or actual defensive ability. All that matters is that a player make more plays than expected, how he does that is unimportant (well, there is some adjustment for run value of the plays, but that is going to be generally minor). These kinds of small details, I think, bog down the greater problem with UZR which is its treatment of discretionary plays. What would be nice would be if STATS recorded which plays more than one player could have made, which Mitchel could delete them before calculating UZR. I don't know why they don't do that.

David, no I don't mean his true talent will be distorted. We're getting into semantics here, which is never a good thing (unless you are debating semantics, I suppose), but I think I clearly (well, maybe not so clearly) state what I meant.

If you want to report how many runs a fielder cost or saved as compared to a league average fielder playing in a typical position on the field, then yes, UZR does that no matter where the fielder is positioned. But, is that an "interesting" result? No, it is not (IMO). The only reason we come up with these metrics, David, and debate them, is to answer some relevant (to some other argument, debate, or question) or interesting question. Sometimes the line is blurred - what is interesting to one person may not be interesting to another. That is fine as long as we are clear in the questions. Most arguments are not really arguments at all - they are "confusion" caused by a failure of all parties involved to be "on the same page" as far as the question is concerned. Usually there is either more then one question or one or more parties are not clear as to what the question is.

That being said, there is no "interesing" reason to ascertain how many runs a player saves or earns as compared to a league average player, positioned "averagealy" on the field. We "care" about two things, and I think I articulated or at least referred to them in my last post. One, his actual fielding "performance" or talent. That is what you are talking about, David. If a fielder is positioned incorrectly for whatever reason (it might actually be that he is positioned "correctly" but the hitters just didn't "know" that, if you know what I mean) then yes, as I said, UZR is a record of his "bad" fielding by virtue of his being out of position as well as how he actually fielded the ball. If he is an average fielder, talent-wise, and is out of position, his UZR or ZR will be worse than average. You say, in that case, that UZR is by definiton correct. By whose definition? If you define UZR as, "Runs saved or cost as compared to a fielder with average skill AND at a typical position," then yes, it is by definition correct. If I define UZR as, "Runs saved by an average fielder in an optimal position," then, no, this player's UZR is not correct, by definition.

Anyway, whatever you want to call it doesn't matter. If a fielder is out of position and we don't know it, then his UZR will not represent his fielding "performance" (unless you want to define fielding performance as including a fielder's position - I think that would be a bad defintion - if someone was watching this fielder and you asked them to rate his performance, they would rate it in terms of how well he got to balls, right?). It is simply not an interesting or useful result. In fact, it is a piss-poor result, excuse the language, to have UZR or any other defensive metric rate an average fielder as above-average or even great because his postioning is optimal but he is really an average fielder. And to pre-empt a comment, I think it is a stretch to say that we should call him a "great fielder" because he is able to figure out that his optimal position is not typical, with the assumption being that most or all other fielders would not be able to figure that out and that the coach or manager (or some scout) is not telling him where to play.

But the bottom line is that we want as much information as possible and then we can answer as many questions as possible. Not knowing where fielders are positioned is a critical gap in the data. Even though you may say that the results are still "fine" becasue it still gives us a "valid" UZR according to the definition (at least "your" definition), it is not a good thing because it does not able us to answer the much more useful and interesting question of how good a fielder a player really is and how many runs he has (and likely will) save as compared to an average player who also plays optimally in the field. The problem of the fielder who plays "out of position" (not typically, but also sub-optimally) is a tricky one. On one hand, we do want to "dock him" for being out of position (or for just preferring, for example, to play shallow or deep in the OF), but on the other hand, we would like to know how good a fielder he really is (had he played, or will play, optimally). Again, if we knew where fielders were positioned (easy enough for the data collectors to do), we could answer all of these questions and we wouldn't be left to guess, especially when we get a UZR or ZR that does not "look right," as with Swisher.

That's all I'm trying to say. I don't think there's really a disagreement here. Going back to two paragraphs ago, if someone wants to give "credit" to a fielder for playing optimally and "overstating" his actual defensive talent (not including his ability to position himself optimally), that's fine by me. Personally, I want to know which fielders actually get to balls better than others and which fielders rate high or low in UZR or ZR because the are playing non-typically but optimally or non-typically and sub-optimally (for whatever reason).

Then I can present all the information and the user/reader can decide for himself what he likes and doesn't like and how he wants to use that information. Again, with the case of Swisher, if indeed he is not that great a fielder (or even a good one) and the reason he has such a high UZR is because he positions himself (or he is positioned by a coach) very differently from the aveage RF'er, because his pitchers produce a very atypical distribution of BIP, at least in RF, I don't think there are too many people that want to hear that he is a "great" fielder (according to UZR). I think they want to know how good a fielder he actually is (talent-wise, getting to the ball-wise), and I think they might find it mildly interesting that for whatever reason he is positioned a lot differently than other RF'ers (IOW, I don't think too may people want to give him much "credit" for that atypical positioning). I could be wrong though. I doubt it, considering all the flak I got for reporting that Swisher had the best UZR in the AL...

there are a few typos, if you are interested

I am. I'll try to fix them.

There are too many "defense" threads going on, BTW

Sorry - I agree that is annoying. The problem is - if these three pieces are one article, no one can sit through it...

We should focus the discussion, and make sure we reference the correct parts.

Since a player should probably get "extra credit" for those outs, one way to do that is to not count them as opportunities. Either way though (counting them or not), the "math" is screwed up (it "bastardizes" the definiton of ZR, which is essentially balls fielded divided by opps).

This is a small issue with ZR. I think it is easily fixed (and you have the data to "fix" it).

I could go on, but mgl's post 10 is excellent and covers most of what I think.

These kinds of small details, I think, bog down the greater problem with UZR which is its treatment of discretionary plays

These are eliminated in ZR, I think. DSG, you just answered a HUGE point for me. Albeit inadvertently, but your side comment, read by the proper eyes makes it very useful.

ZR doesn't include pop-ups. Pop-ups are generally described as FBs that don't travel 220 feet. That's where you get most discretionary plays. there are some between the CF-LF/RF, but not as many as up front. And plays that deep for the CF-LF/RF, are only moedrately discretionary. While maybe both *could* catch the FB, one or the other usually has teh better positioning or is closer.

Here is where DSG provided me a breakthrough:
I have studied the relationship between ZRchances and OF Putouts (PO) for a long time.

For some insane reason, a player's ZR chances in the OF were always a handful fewer than his POs. I couldn't figure that out. this is it: Those are the discretionary popups behind the IF. Now I have to compare Andruw Jones ZR chances compared to his putouts to see if I can get a feel for his "vulturing" of discretionary plays. At least see if he has a higher number.

Wow. that's a big breakthrough for calcualting RS from normal regular data, and increases the ability (of me anyway) to determine/approximate OF chances. Well, I always did it this way: PO - 5 (for a full-season) divided by ZR. Now I have an explanation.

Fantastic.

Nice article, it's good to get all this info down in one place.

I found it interesting that 2B are responsible for 6 zones, while SS are only responsible for 5. I guess it makes sense, if you consider the 50% threshold: SS with the longer throw will make fewer plays than the 2B would make in the "mirror-zone" across the infield. Also, the 3B handles 4 zones and the 1B only 3. Makes sense also, since the 1B has to be able to retreat to 1B or hold runners on. The upshot is that each side of the infield has a 1-zone hole, but they are not in the same place (relative to the foul line). The hole on the left side is zone G, the 5th from the line, while the hole on the right side is zone U, the 4th from the 1B foul line.

I'm not sure what all this means for constructing a defensive metric, but I found it kind of interesting.

Chris,

I don't know if I made it clear or not, but when I made that post (and still), I had come around to the opinion that ZR adds to UZR in the outfield, precisely for that reason. I don't know that it's better (batted ball speed not being accounted for, plus run values play a bigger role in the OF), but it certainly can and should be used to check weird UZR ratings.

Mitchel,

I disagree. UZR is measuring performance. Positioning goes into that. If a player knows how to position himself, or even if a coach knows how to position him, we should still give that player credit for what actuall happened on the field, that is that he saved more runs than the average player at his position would have with average positioning. Positioning is part of defense; there's no reason to say that it shouldn't count. That's like saying that a player who doesn't swing at pitches outside of the strike zone shouldn't get credit for it. Sure, Vlad Guererro (or whoever) might have better numbers if he didn't chase pitches out of the zone, but he does. He might be a better hitter than Jason Giambi (or whoever) but that's not what's important, what's important is who contributed more runs to his team. Same goes for fielders.

Oh, and I think never said how good an article this was. It's a very good description of ZR -- nice job, Chris.

If a player knows how to position himself, or even if a coach knows how to position him, we should still give that player credit for what actuall happened on the field, that is that he saved more runs than the average player at his position would have with average positioning.

DSG,
however, most players are positioned "typically". So assuming a typical positioning is the correct evaluation. As I noted elsewhere, it is possible (depends on why UZR has Swisher so high) that *any* OAK RF would have been positioned that way.

Heck, maybe someone can check out a few games at MLB.tv and see if he was somewhere else (shallow or something).

John (and SMK and MAH and MGL),
thanks for the compliments.

What you note is interesting. STATS actually *added* that extra 2B zone (and the fourth 3B zone) because they found 2B's made that play.

Chris,SMK,Anaheim,DSG,John and MGL,

Thanks for a great thread. I've learned a tremendous amount. I know it sounds a little wishy-washy to say this, but it's good to have UZR as well as "DC-ZR" ("Dial-Chone ZR") and Glassko's system based on GB/FB data.

however, most players are positioned "typically". So assuming a typical positioning is the correct evaluation. As I noted elsewhere, it is possible (depends on why UZR has Swisher so high) that *any* OAK RF would have been positioned that way.

I wasn't referring to that specific case. Still, I do believe that if it's a matter of positioning, Swisher should get the credit. It's like saying that we should add .5 points to Maddux's career ERA because of Mazzone. However, it is possible that Oakland does a good job of positioning its fielders since it is one of the major league teams that has a sophisticated defensive system, I think. More likely with Swisher, however, is that there is some error in UZR resulting in his rating. Either a programming bug or some strange run value thing. Perhaps it does have to do with discretionary plays. That could explain Kotsay's low UZR (though he rates poorly in all systems). Mitchel, do you have the number of soft fly balls that went to zones between CF and RF (looks like zones 89, 89D, and 89S) for Oakland and what % were caught by Swisher vs. league average? Also perhaps the same for zone 34D?

Michael,

No "l". It's Gassko. Not Glassko, Grassko, or any other variation I've seen on Primer (though they have been very creative). :)

Something that I have grappled with for a while, but never took the time ro resolve it:

UZR basically compares each fielder's percentage of outs made in each PS zone (STATS converted to PS zones) to the average fielder (at that position of course). This is regardless of the percentage of outs made by ALL fielders (league-wise). That was probably a confusing description, so I'll give an example. Let's say the average CF'er makes 10% of all plays in zone X and the RF'er makes the other 90% (all fly balls in that zone are caught). This suggests of course that it is an "easy" zone, since all balls are caught in that zone. Since this is an easy zone, it also suggests that there are probably many discretionary plays in that zone - balls that either the CF or RF can catch. Now let's say that a particular CF catches 20% of the balls in that zone and the RF catches 80%. The CF is going to have a nice UZR number for that zone and the RF is not. However, it is likely that the extra 10% that the CF caught, as compared to an average CF was discretionary - either the CF or RF could have caught it. I have thought about several things to avoid this problem. One, to not count "against" a fielder any ball that is caught by another fielder (UZR does NOT do that now). In this case, the CF would indeed get extra credit for catching those extra 10% of balls, but the RF would not be "docked" for only catching 80% rather than the league-average 90%. Right now, UZR credits the CF with the extra balls AND docks the RF.

I have also thought about only doing that in zones where most of the balls are caught, since that implies that, like I said, it is an "easy" zone. I have thought about doing some combination, like only docking a player a little bit when the ball is caught by another fielder and docking him the full amount if the ball is not caught by another fielder (the assumption being that if a ball is caught by a certain fielder, it may have been able to be caught by another fielder. This last method is probably the correct one, mathematically, but I am not sure.

I have also toyed around with the idea of not using at all zones in which almost all of the fly balls are caught, since again, these are probably easy fly balls anyway, and as Tango states (and correctly so), all easy balls are just "noise" anyway (other than errors).

Anyway, given the limitations on the data we have, there still has to be a better way of handling possible discretionary plays than the way UZR handles it now. Any suggestions?

If you "tweak" UZR for discretionary zones, you should simply not count those zones at all, I think. Otherwise, you're still punishing players for "strange" positioning. Let's take your example. Now let's say that the CF knows that the RF can catch all balls in that zone, so he shades a little towards left field. Now, instead, the RF makes 95% of all plays in that zone, and the CF makes 5%. You're still docking the CF, and it makes little/no sense to.

On the other hand, what if the RF continues to make 90% of the plays in that zone, but the CF now only makes 5%? Now, instead, only 95% of all balls in that zone are being caught. Who do you dock for this? It probably makes the most sense to split the "credit", but that benefits the CFers UZR, and hurts the RF, even though the RF is doing what's expected of him, and the CF is not. The CF is going to get credit for the extra balls he catches in left-center, but only lose half of that in the balls he is not catching in right-center. Even though, on a team level, there is no benefit. In other words, half of the blame that should be going to the CF is being transferred to the RF. Again, it seems to make more sense to just not use that zone.

On the other hand, the team is catching only 95% of all balls in that zone even though the league average is 100% (I know, mathematically impossible, but the point bares out). So, overall, the team is doing poorly in that zone, but better in the left-center zone. Those benefits should cancel out, but instead, if you ignore that zone, a team-level UZR would reveal that the team is actually better than it is because that zone is ignored since the CF is getting extra credit for catching more balls in left-center, and not losing any credit for the balls that are dropping in the right-center zone.

In short, I don't really know what to do. The best solution is probably to see how taking out BIP in that zone and the zone opposite (in my example, left-center) would impact team level UZR, and then assign the credit to the player who is resulting in the change. So, for example, in my example, since more balls in the zone opposite are being caught, and less are being caught in the right-center zone, effectively cancelling out, no runs would be added or charged to the CF. However, if the CF was catching less balls in the zone opposite than the team was losing in right-center, the blame for that would go to him.

That might not make sense, so if it doesn't, I'll try to explain it again.

I'll have to digest this for a while...

If you "tweak" UZR for discretionary zones, you should simply not count those zones at all, I think

Uh, no. taking out zones isn't a good idea at all. Gap zones arn't "discretionary". *Occassionally* either guy could make a catch, but they aren't discretionary. Discretionary is a popup behind second base.

And CFs don't play that way. They play where the scouting report says the ball is most likely to be hit. "Shading" due to another good OF involves a step or two, nothing close to a "zone". the zones are 20 feet wide at that point.

In general the OF is too big to play anywhere other than where you expect the ball to be hit - in terms of direction. You can cheat up or back absed on your own speed, but the likelihood of a ball being hit in a direction, being there to adjust for depth of ball in play.

Chris,

You missed my conclusion. I was just putting my thoughts down as they came to me. The conclusion was that UZR should look at the shared zones and determine if the balls caught in them combined on a team level are equal to league average. If so, it should not count those zones on an individual level. Otherwise, it needs to break up the credit or demerit between players responsible for those zones. I have an idea of how to do that, but have to run so I'll post it later.

The conclusion was that UZR should look at the shared zones and determine if the balls caught in them combined on a team level are equal to league average. If so, it should not count those zones on an individual level.

I disagree.

Otherwise, it needs to break up the credit or demerit between players responsible for those zones.

Since zones in ZR do not overlap, there is no need to do this. The OF is covered. You simply credit a player for the play he made.

Right, but if one zone is split 60/20 between CF and RF, it makes no sense to not include the RFers numbers. You're losing a large part of his range.

And please provide a more detailed critique of my idea. There's a good enough chance that it's wrong, but "I disagree" doesn't tell me anything. I want to know why it's wrong as well. Don't hog the knowledge.

I want to know why it's wrong as well. Don't hog the knowledge.

It's wrong because there is no need to give *any* partial credit. Just do UZR like this:

BIZ = X
OIZ = Y
OOZ = Z

ZR = Y+Z/X = A

Convert A to Runs above average. Possibly tweak with some ratio of run value based on OOZ.

then you are giving the fielder *full credit* for all his outs. You don't need to make a partial for zones of non-responsibility. It doesn't genrate more accuracy, AFAICT.

And as I said, the OF doesn't really have any discretionary sections. Only pop-ups.

I think over the last few articles, I've done everything *but* hog the knowledge.

Wait until the next piece. There's a huge knowledge dump. Just HUGE.

I am beginning to wonder about MAH's data, and wheterh or not I can slip his in there.

MAH,
are you listening? If I send you a list of ~100 players, can you share your 2005 ratings?

handedness of the batters (which also affects the positioning and the speed of the batted ball

mgl: How big of a factor is handedness of batter on the speed of the batted ball? How much do you think this changes ZR from what it "should" be? By several plays over 150 games?

My 2 cents on UZR:

1. I would not charge a negative to any player when someone else catches the ball. If its my team, I'm not concerned so much who catches the ball, just that its caught.

2. In zones where more than one player can make a play, I would look at the percent of balls that are caught, without worrying who normally makes the catch. If the CF makes a catch 70% of the time, and the RF 20%, then its +.10 regardless of who fields the ball.

If the ball in that situation is not caught, then the blame has to be shared, -.7 for CF and -.20 for RF.

This would, IMO, solve any problem of discretionary plays.

If the ball in that situation is not caught, then the blame has to be shared, -.7 for CF and -.20 for RF.

IMO, this is unnecessary. Based on the OF coverage, that ball is in one or the others zone. It's the CFs responsibility solely. No negative should be docked on the RF.

I just think that's just the wrong way to analyze defense.

And those aren't *discretionary plays*.

Re: speed of batted balls.
Bals that are pulled are hit harder than balls hit the other way. If a team faces more RHBs, then, in theory, that 3B and SS have their ZR lowered by having more to face more difficult chances than others in the league.

On flyballs, the entire outfield is covered, there are no gaps between outfield zones.

Certainly some of those plays are discretionary. We've all seen two outfielders camped under a high flyball until one gives way.

Since the CF has the authoritie to call off everyone else, charging only the CF would work just fine. Either solution, IMO, is an improvement of charging a negative to a fielder when the ball is caught anyway.

Especially if the fielders themselves become aware of the methodology. Cameron and Beltran spring to mind.

If a team faces more RHBs, then, in theory, that 3B and SS have their ZR lowered by having more to face more difficult chances than others in the league.

Since Pittsburgh had the most innings by lefties, it makes Jack Wilson's top ZR look even better.

Certainly some of those plays are discretionary. We've all seen two outfielders camped under a high flyball until one gives way.

That's *some plays*, not discretionary *zones*. There are no zones where two players are routinely there to make the catch.

Since Pittsburgh had the most innings by lefties, it makes Jack Wilson's top ZR look even better.

Uh, that's the theory, not something I ascribe to just yet.

I think over the last few articles, I've done everything *but* hog the knowledge.

I was being completely tounge-in-cheek. You've really done a great job here.

But your points about how to do UZR (and I'll wait for MGL to chime in) seem to underscore the difference in opinions about how to correctly do defensive ratings. You clearly subscribe to the ZR methodology, with your only qualm, seemingly, being its treatment of BOZ (which is a very legitimate and important point). I tend to subscribe to the UZR methodology more, with some worry about its treatment of discretionary plays. To each his own, I guess.

I was being completely tounge-in-cheek.

I did figure that. So was I.

You clearly subscribe to the ZR methodology, with your only qualm, seemingly, being its treatment of BOZ

I don't know that it is my only qualm, but my biggest.

To each his own, I guess.

This part is true. I think zone analysis is the correct way to do it, because the assignment of "partial credit", to me, isn't right, and the only real way to do it with a UZR methodology.

assignment of "partial credit", to me, isn't right

You see, to me that (and the adjustments made in UZR) is the greatest advantage UZR had--I see it as a positive.

I meant has

Since Pittsburgh had the most innings by lefties, it makes Jack Wilson's top ZR look even better.

Well, Pitt had more innings thrown by LHP last year than they did this year (IIRC) and his ZR was higher this year. I would think if this applied to anyone it would be 1b and 3b because those are the positions where hard hit balls would matter most because of reflexes.

I'm still not sure it would actually make a difference or a big one, as Dial said.

One reason I really think it might not apply to SS is because one of their hardest plays is a soft-hit chopper. I could be mistaken. I often am.

I know the left-right thing makes a difference, because MGL uses it in his UZR adjustments. I'm assuming he studied it, and wouldn't use it if there was no effect.

What I don't know is the magnitude of the effect on the infield positions.

Chris,

Two regrets. First, I generally can't send comments during the work day. Second, I haven't done DRA for 2005 because, among other reasons, my spreadsheets work off of Retrosheet data, which isn't out yet.

DRA ratings are available in my February 2005 article at The Hardball Times, but each player's rating is the average of his DRA ratings for each season in 2001-03 of at least 130 games at the relevant position, not a single-season rating. As I mentioned in the article, I'm a big believer in using at least two seasons of data, since Tango (and, I believe, MGL) consider two-year _UZR_ ratings much more accurate than one-year ratings.

I'm going to do a close read of the comments in this thread to see if I can make any helpful suggestions for UZR.

The new systems that you and Chone have developed enhance zone ratings with non-PBP data. Have you figured out a way to generate a "pure" ZR runs-saved rating system? I made an attempt in February when I introduced the DRAZR linear combination of DRA and ZR, but I recall that we both recognized there had to be some "noise" left in the ZR part.

Best,

Michael

I know the left-right thing makes a difference, because MGL uses it in his UZR adjustments. I'm assuming he studied it, and wouldn't use it if there was no effect.

Yeah me too, that's why I asked up thread. Hopefully he can and will answer. I figure he might be able to request the data broken down by handedness of pitcher but didn't actually do a study to see if it mattered.

He (Mitchel) did, and it does. You can read more in part two of the original UZR article.

It makes a difference in UZR - I don't know how much of a difference it makes in ZR.

Would you mind posting a link DSG? I have absolutely no clue how to navigate around BTF anymore.

I'm kind of late to this party, but I would like to congratulate Chris on an excellent article. Please, keep it up.

Thanks, TFTIO, and I expect more tomorrow (er, today).

Hopefully, I'll change the subject later this week.

Where could a noob find the UZR ratings of his favorite players? And / or the zone ratings? Perhaps from the 2005 season? thanks in advance....

Okay - found zone rating on ESPN - how about UZR?

UZR is proprietary and can only be found when mgl decides to post it. If you search the NL GG thread (Maddux is in the title) you can see the UZR's of GG winners and the best and worst in th eleague.

I willbe posting all my defensive ratings with a side-by-side comparison of all teh methods you've seen recently in the next day or so (maybe today!).

I'll be happy to post UZR for anyone who wants it. I'll just need about $5000 upfront for the data and about a year for me to figure out how to process it. No guarantee that my results would even resemble MGL's.

Rally,
your present method approximates UZR just fine. Not quite as good as mine, but really good.

What is the difference in our methods? Is it just the way chances are estimated?

Mostly chances. Probably run value a bit.

Our correlation is almost 1.

As MGL points out, the flaws in ZR do tend to even out over a period of years for a particular player. Three year zone ratings are more useful than one year. In some cases, the flaws in ZR result in distortions over a period of years. For instance, a third base/shortstop combination in which both routinely play closer to the line than usual will result in a higher ZR for the third baseman and a lower ZR for the shortstop.

Team defence markers are, in my view, more accurate than individual ones. A good litmus test for any system is the agreement between the sums of individual defence ratings and team ratings.

For instance, a third base/shortstop combination in which both routinely play closer to the line than usual will result in a higher ZR for the third baseman and a lower ZR for the shortstop.

Do you haev any evidence for this?

Team defence markers are, in my view, more accurate than individual ones. A good litmus test for any system is the agreement between the sums of individual defence ratings and team ratings

I don't think this is necessarily the case, but as I mentioned, I haven't really investigated it.

Rally did with his measure and got some strong agreement with ERA/RA, and my methodology (based off ZR) correlates with Rally's very highly.

There are too many "defense" threads going on, BTW.

Hear, hear. I try to follow this stuff pretty closely, and just noticed this thread.

How deep are the infielders' zones? I assume it partially depends on how hard a ball is hit.

I also wonder what the different percentages are for each zone ... okay, so an average CF will catch > 50% of the fly balls hit to the zones you describe. But I would wager that fly balls hit 330 feet to Zone M are caught essentially 100% of the time, but flyballs hit 380 feet to Zone Q are caught far less frequently. The CF should get more credit for making the catch in the latter case.

Of course, that's what STATS' original "UZR" tried to account for, as well as David Pinto's PMR, and that issue underlies MGL's UZR. And my own look at ZR demonstrates that you can guess the runs prevented pretty closely, regardless of such issues. But it would still be interesting to see that sort of thing presented ...

How deep are the infielders' zones? I assume it partially depends on how hard a ball is hit.


10 foot segments (or they were). MGL is suggesting they are at 5 feet now, but I doubt it for teh OF.

Hear, hear. I try to follow this stuff pretty closely, and just noticed this thread.

Bugger off. Follow *this* section (Dialed In) of BTF closer. This is where all teh good writing is!

Chris, please stop writing "the" as "teh". You have been doing it for God knows how long...

Ten foot segments ... how deep is the dirt on a typical infield? The diagram makes it look like 30 feet or so.

Are "swinging bunts" up the third-base line generally out of the 3B's zone?

please stop writing "the" as "teh".

It is unintentional, I assure you.

Ten foot segments ... how deep is the dirt on a typical infield? The diagram makes it look like 30 feet or so.

Are "swinging bunts" up the third-base line generally out of the 3B's zone?

The dirt is closer to 40 feet deep, I think - 10 feet or so in front of the baseline and 30 feet behind it.

Yes, swinging bunts up the 3B line are the 3Bs zone.