Defensive values

[Gom]

Interesting point I want to bring up.

 

Many years ago, I used to play a game called APBA. It was a baseball simulation. Back in the days before the internet, and MLB.tv, it was the first I learned about players from the other teams outside of when they played the Yankees.

 

They used to rate defensive players. The max values a player could have were as follows:

 

C: 9

1B: 5

2B: 9

SS:10

3B: 6

OF: 3

 

Catchers also had arm ratings and passed ball ratings. OFs had arm ratings. I won't get into either, as they aren't relative to my point.

 

I'm mentioning this for a reason. I believe that this is an extremely accurate representation of the maximum impact a player has on defense. Now, these numbers are relative. In other words, a great defensive catcher is worth more than a great defensive firstbaseman, etc. The impact that these players have on the game is mainly dependent on the position they play.

 

This is why when you discuss the difference between say...Cameron and Bay...it's not as big as you might think...but when discussing Scutaro and Gonzalez, it might be bigger than you realize.

 

I have other points, but I want to see what you guys think of this concept first.

[yeszir]

On this somewhat arbitrary scale, I would argue that 1B should be a 6-7. They're involved in most infield plays and a bad one can lead to a lot of extra baserunners.

[Cocos Disciples]

You mentioned arm and PB ratings. Are there more specific ratings for other positions (range, arm for infielders, etc)? Without knowing a lot about the system it seems like it would be a solid basic rating. How did they come up with the numbers?

[Gom]

You mentioned arm and PB ratings. Are there more specific ratings for other positions (range' date=' arm for infielders, etc)? Without knowing a lot about the system it seems like it would be a solid basic rating. How did they come up with the numbers?[/quote']

I don't know how they came up with the numbers. I played this game for six years or so, then I went to college and realized that girls and booze were a lot more fun. However, they didn't rate infielders different, they were just given a number. OFers and catchers were given an arm rating to calculate for advancing a base or basestealing.

 

OF arms ranged from 40 [Roberto Clemente type of arm] to 25 [Johnny Damon]

 

Catchers ranged from -4[worst throwing catcher you can think of] to +6[Johnny Bench]

 

I'll list the ranges, low to high.

 

C: 6-9

1B: 2-5

2B: 5-9

SS: 6-10

3B: 3-6

OF: 1-3

 

Now I don't know how they rated these players, but it was his overall defense. Now, remember, I played this game in the 80s, but it was my first impression of defensive values. For OFs, I would put the best arm in RF, the best "fielder" in center, and the worst OF in left.

 

The concept I'm trying to say is that the players had impact defensively, but it was relative. This is the problem with some of the more advanced statistical models. By reading it, they would have you believe that a Gold Glove fielding OFer would have more of an effect over an average fielding OFer than the same comparison at say...first base.

 

What are your thoughts?

[ORS]

Gom, impact is dependent on opportunity. Opportunity can be accounted for. What don't you understand about this?

[Cocos Disciples]

I don't know how they came up with the numbers. I played this game for six years or so, then I went to college and realized that girls and booze were a lot more fun. However, they didn't rate infielders different, they were just given a number. OFers and catchers were given an arm rating to calculate for advancing a base or basestealing.

 

OF arms ranged from 40 [Roberto Clemente type of arm] to 25 [Johnny Damon]

 

Catchers ranged from -4[worst throwing catcher you can think of] to +6[Johnny Bench]

 

I'll list the ranges, low to high.

 

C: 6-9

1B: 2-5

2B: 5-9

SS: 6-10

3B: 3-6

OF: 1-3

 

Now I don't know how they rated these players, but it was his overall defense. Now, remember, I played this game in the 80s, but it was my first impression of defensive values. For OFs, I would put the best arm in RF, the best "fielder" in center, and the worst OF in left.

 

The concept I'm trying to say is that the players had impact defensively, but it was relative. This is the problem with some of the more advanced statistical models. By reading it, they would have you believe that a Gold Glove fielding OFer would have more of an effect over an average fielding OFer than the same comparison at say...first base.

 

What are your thoughts?

 

I like the idea of relativity among positions, but I don't like that a terrible defensive catcher has a higher rating than a fantastic defensive 1B, for example (if I understand the range of ratings, which I'm not positive I do). And while the number would give one a general sense of a fielder's worth, I like the more specific ratings, such as range and arm, and I think they are even more important in determining value because they tell you strengths and weaknesses. It's more precise.

[Dipre]

Interesting point I want to bring up.

 

Many years ago, I used to play a game called APBA. It was a baseball simulation. Back in the days before the internet, and MLB.tv, it was the first I learned about players from the other teams outside of when they played the Yankees.

 

They used to rate defensive players. The max values a player could have were as follows:

 

C: 9

1B: 5

2B: 9

SS:10

3B: 6

OF: 3

 

Catchers also had arm ratings and passed ball ratings. OFs had arm ratings. I won't get into either, as they aren't relative to my point.

 

I'm mentioning this for a reason. I believe that this is an extremely accurate representation of the maximum impact a player has on defense. Now, these numbers are relative. In other words, a great defensive catcher is worth more than a great defensive firstbaseman, etc. The impact that these players have on the game is mainly dependent on the position they play.

 

This is why when you discuss the difference between say...Cameron and Bay...it's not as big as you might think...but when discussing Scutaro and Gonzalez, it might be bigger than you realize.

 

I have other points, but I want to see what you guys think of this concept first.

 

This is how UZR looks at it:

 

SS = 9

2B = 7

3B = 5

CF = 4

LF = -7

1B = -7

RF = -10

 

As you can see, UZR applies a similar but more "refined" numerical value to positional impact.

 

Here's the detailed explanation:

 

UZR,multiple positions.

[Gom]

Gom' date=' impact is dependent on opportunity. Opportunity can be accounted for. What don't you understand about this?[/quote']

 

I understand it perfectly. Their scale was based upon opportunity. They believed that the shortstop had the highest opportunity, hence he had the highest potential rating. Next came catchers and second baseman, followed by thirdbaseman, firstbasemen, then outfielders. The maximum impact an OFer could have was the smallest incremental change. Most OFers were "two's".

 

What I'm saying is that the impact that an OFer would have would be minimal on the team's overall defense.

 

Now look at this:

 

http://www.fangraphs.com/blogs/index.php/highs-and-lows-of-uzr-2007-9-dunn

 

This lists the best and worst fielders of the last three years. There are two glaring errors here.

 

First of all, I am of the fundamental belief that no OFer can impact a game, day in and day out, as much as an infielder. Look at the listing...Utley, Vizquel, Hardy..all middle infielders. I guess you could say that those guys would be 9's or 10's. [by the way, the only 10s in the whole 1985-86 seasons were Ozzie Smith, and Tony Fernandez I believe [if memory serves correct on the year]. The scale was a bell curve, with very few players getting the top or bottom numbers. Zimmerman however? How can a thirdbaseman impact the game as much as a shortstop or secondbasemen? The real head-scratcher was Gutierrez.

 

He's an OFer. He just doesn't get that much of a chance to impact the game. Nothing about how good or bad he is, his impact is just not that great due to chances. How is it possible that a Franklin Gutierrez means more to a teams defense than say....a Pedroia?

 

Now look at the other side. Who are the worst fielders? All OFers. Doesn't make sense. They have the smallest chance to impact the game. Yet, they are costing their teams tons of runs. Doesn't make sense.

 

This is why I think a gold glove caliber OFer as opposed to a terrible OFer would have the same impact [say 3 to 1] as say an Ozzie Smith would have over an average shortstop. The impact is not as big as you would think.

[Keeper]

I'd still like to know who rates UZR. Scouts? Scorekeepers? Or sabermetricians? I would think the best judges of defensive prowess would be scouts.

[Gom]

Keeper, it's a mathematical formula, as far as I know.

 

I'm starting to see the errors in this. This is what I was hoping for.

 

First of all, kudos to Dipre for the little primer on UZR, it was informative. There seem to be some fundamental errors in it. 1B seems way too low, CF seems way too high. The range is much greater than I would believe.

 

However...what seems to be the biggest problem is the lack of a "cap", or maybe better explained as a logarithmic jump rather than a linear one.

 

It seems to me that to get a true value of someone's defensive ability, it should be bell-curved. Let me elaborate further. I'm not sure what the major league baseball average, and the exact number is truly unimportant. Let's say the average batting average is .270. The farther you get away from .270, in either direction, the harder it is to maintain. The norm will drive you to that average. This is the reason that .400 is very tough to do, but .350 is done ever year for the most part. Even the worst pitchers don't bat .000. Opposite end of the spectrum.

 

Batting average for major league players falls into this curve..as will most stats that we derive offensively or defensively.

 

Realize I'm more or less thinking out loud here.

 

What happens is that there should be a range. Using Dipre's values that he quoted for UZR, there should be a "cap" or logarithmic jump. Say we look at a CFer. He is average, so he gets rated a 4. However to go to 5, to make him equal to an average 3B, he's got to be a good CF. To get to 7, an average 2B, he's got to be Wille Mays. It should be next to impossible to get him to equal an average shortstop. The values should be averaged, and put on a bell curve scale in logarithmic fashion. As of now, in linear format, it's next to useless IMO.

[Dipre]

Keeper, it's a mathematical formula, as far as I know.

 

I'm starting to see the errors in this. This is what I was hoping for.

 

First of all, kudos to Dipre for the little primer on UZR, it was informative. There seem to be some fundamental errors in it. 1B seems way too low, CF seems way too high. The range is much greater than I would believe.

 

However...what seems to be the biggest problem is the lack of a "cap", or maybe better explained as a logarithmic jump rather than a linear one.

 

It seems to me that to get a true value of someone's defensive ability, it should be bell-curved. Let me elaborate further. I'm not sure what the major league baseball average, and the exact number is truly unimportant. Let's say the average batting average is .270. The farther you get away from .270, in either direction, the harder it is to maintain. The norm will drive you to that average. This is the reason that .400 is very tough to do, but .350 is done ever year for the most part. Even the worst pitchers don't bat .000. Opposite end of the spectrum.

 

Batting average for major league players falls into this curve..as will most stats that we derive offensively or defensively.

 

Realize I'm more or less thinking out loud here.

 

What happens is that there should be a range. Using Dipre's values that he quoted for UZR, there should be a "cap" or logarithmic jump. Say we look at a CFer. He is average, so he gets rated a 4. However to go to 5, to make him equal to an average 3B, he's got to be a good CF. To get to 7, an average 2B, he's got to be Wille Mays. It should be next to impossible to get him to equal an average shortstop. The values should be averaged, and put on a bell curve scale in logarithmic fashion. As of now, in linear format, it's next to useless IMO.

 

You took it completely out of context.

[ORS]

I understand it perfectly. Their scale was based upon opportunity. They believed that the shortstop had the highest opportunity, hence he had the highest potential rating. Next came catchers and second baseman, followed by thirdbaseman, firstbasemen, then outfielders. The maximum impact an OFer could have was the smallest incremental change. Most OFers were "two's".

 

What I'm saying is that the impact that an OFer would have would be minimal on the team's overall defense.

 

Now look at this:

 

http://www.fangraphs.com/blogs/index.php/highs-and-lows-of-uzr-2007-9-dunn

 

This lists the best and worst fielders of the last three years. There are two glaring errors here.

 

First of all, I am of the fundamental belief that no OFer can impact a game, day in and day out, as much as an infielder. Look at the listing...Utley, Vizquel, Hardy..all middle infielders. I guess you could say that those guys would be 9's or 10's. [by the way, the only 10s in the whole 1985-86 seasons were Ozzie Smith, and Tony Fernandez I believe [if memory serves correct on the year]. The scale was a bell curve, with very few players getting the top or bottom numbers. Zimmerman however? How can a thirdbaseman impact the game as much as a shortstop or secondbasemen? The real head-scratcher was Gutierrez.

 

He's an OFer. He just doesn't get that much of a chance to impact the game. Nothing about how good or bad he is, his impact is just not that great due to chances. How is it possible that a Franklin Gutierrez means more to a teams defense than say....a Pedroia?

 

Now look at the other side. Who are the worst fielders? All OFers. Doesn't make sense. They have the smallest chance to impact the game. Yet, they are costing their teams tons of runs. Doesn't make sense.

 

This is why I think a gold glove caliber OFer as opposed to a terrible OFer would have the same impact [say 3 to 1] as say an Ozzie Smith would have over an average shortstop. The impact is not as big as you would think.

Short answer: You don't understand why the ability to account for opportunties is relevant.

 

The impact is what it is when you can account for opportunities. It doesn't matter if you are comparing two SS or two LF. In your video game they used the scale to compare between positions, but that isn't what you are talking about. You are using that scale to say that it's not possible to upgrade that much at an OF position. This is true within the context of the scale, but the scale is based on trends. Any individual comparison within a position can be outside the scale in reality, and current statistical methods, which account for opportunities (and quality of opportunities), capture this reality.

 

So, no, your video game does not invalidate modern statistics.

[Gom]

Yet you can believe that the 2nd biggest impact of the last three years is an OFer. Seriously?

 

It is impossible to have an intelligent conversation with you. You lack the ability to reason and think for yourself. You can never formulate an opinion on your own. The only think you can do is cut and paste.

[ORS]

Gom, that post is a critique of your analytical method. I spoke only to the conceptual errors. I made no mention of a specific example, so my belief is not in question because I haven't stated it. You are making things up again.

 

Cut and paste? This is a critique of your thought process as in applies to a video game from 1980. If you can find me one article on this subject from which I can cut and paste, please provide me a link.

 

Wash, rinse, repeat. It's hilarious you think you've struck a point here.

[Gom]

You fail to see errors in a system that has an OFer as the second biggest impact player defensively, one that undermines Tex's and AGon's defense, and has the worst defensive players as OFers, who have the smallest impact of all the defensive players.

 

I'm trying to see how and why it's flawed, and to figure out if there is a way to fix it. Any ideas?

 

By the way, in the 80's it was a board game.

[ORS]

Gom, you still fail to understand how the stat is measured. It is measured relative to average. If an LF is +10, that means he's +10 compared to the rest of the league at his position. If a SS is +5, same thing. It doesn't mean the LF has more impact on the game, the SS still does. It means the OF was better than his peers at the same position by a margin larger than the SS.

 

To compare them for overall impact, like you want to, you'd need to apply the adjustments Dipre posted. A SS is 9, to a LF's -7. There's a 16 run difference in positional impact. When you apply the adjustments for position it goes from 10 to 5 in favor of the LF, to 14 to 3 in favor of the SS.

 

Get it?

[Gom]

Makes more sense now, thank you.

 

However, why is it so flawed? Also, that dismisses my notion that each point should be exponentially harder to get than the one before. If this is scales logarithmically, it would lead to a greater degree of accuracy.

[ORS]

Why should each point be harder to get than the next? It's a linear process.

[Gom]

Why should each point be harder to get than the next? It's a linear process.

 

Because of the sheer randomness and inaccuracies inherent in the metric. Also, when comparing players, it should be noted that a difference of say, 2.0 on a UZR scale is much more important for a shortstop than it is for a LFer.

 

Let me attempt to illustrate. I'm not a mathematician, but here is the concept. To make this a more accurate description of a player's total value defensively, it should be multiplied by a factor. SS is the most important defensive position. So going with the above metric, he would get a value of say 1. Each other position would receive a percentage of that.

 

Say a SS is 2.0 UZR better than his counterpart. Now let's say you have a LFer that is also 2.0 better than his counterpart. To actually be 2.0 better at shortstop is a huge impact defensively at shortstop, but minimal at LF. Let's say the factor for LF is .10.

 

Where a shortstop might save you 2.0 runs per whatever, over the same factor, the LFer will save you .2 runs in that time period. Therefore, even a marginal increase or decrease at shortstop will be more substantial than a major increase or decrease in the OF.

 

I'm not really interested in the math, just the concept. What do you think?

[Dipre]

Because of the sheer randomness and inaccuracies inherent in the metric. Also, when comparing players, it should be noted that a difference of say, 2.0 on a UZR scale is much more important for a shortstop than it is for a LFer.

 

Let me attempt to illustrate. I'm not a mathematician, but here is the concept. To make this a more accurate description of a player's total value defensively, it should be multiplied by a factor. SS is the most important defensive position. So going with the above metric, he would get a value of say 1. Each other position would receive a percentage of that.

 

Say a SS is 2.0 UZR better than his counterpart. Now let's say you have a LFer that is also 2.0 better than his counterpart. To actually be 2.0 better at shortstop is a huge impact defensively at shortstop, but minimal at LF. Let's say the factor for LF is .10.

 

Where a shortstop might save you 2.0 runs per whatever, over the same factor, the LFer will save you .2 runs in that time period. Therefore, even a marginal increase or decrease at shortstop will be more substantial than a major increase or decrease in the OF.

 

I'm not really interested in the math, just the concept. What do you think?

 

This concept, in itself, makes no sense. UZR does not attempt to illustrate the defensive impact of the players per se, but the efficacy of a player when compared to what could be defined as the positional "average". As ORS said, it's a linear process, the reason the "impact" aspect of the formula was incorporated in the first place is to eliminate the variables that appear from the difference in sample sizes that appear from one position to the next.

[ORS]

You are still woefully short of understanding what it is measuring....on a conceptual level.

 

It's linear because each play is judged relative to what an average fielder would do with that opportunity. Let's suppose that on the first play and last play an OF has during a season, it is the exact same type of hit ball, hit by the same hitter, same azimuth, trajectory, velocity, etc, and let's suppose the fielder makes the play each time. You want to limit the amount of credit he is given on the last ball because you think subsequent credit should be harder to attain than initial credit. Do you not see the error here? It's the same play, it's still being judged relative to average, so it should receive the same credit. Thus, it is linear (on a conceptual level).

 

I realize why you think this should be the case. You think there should be a limit to the amount of impact an OF can have over the course of a season. Realize this, the limits established by your board game are based on trends....averages. Averages are composed of individual data points that are both above and below the mean, with outliers in each direction. You do not prohibit an outlier from occurring, because they are real. If a fielder performs to an outlier level, he should get credit for that, and be recognized for greatness or abject failure depending on the direct of the outlier.

 

To the rest, you are suggesting an application of a second adjustment. The adjustment scale has already been determined. You don't need to, and more importantly, shouldn't, adjust the data twice. You compile, then adjust. Not adjust, compile, and adjust again. If you use the data and adjustment for position right, then you can compare two fielders for overall impact.

[Gom]

I understand your point, ORS, but adjusting it on a logarithmic level would eliminate many of the statistical deviations. Since the data itself is very limited, the variations are great.

 

Conceptually, I understand it. However, it's application is very limited.

 

Case in point...an established player, say a Jeter, or a Drew...his offensive numbers will basically stay the same with little variation. Similar OPS from year to year, with little deviation. However, looking at Jeter, there is major deviation in his UZR. This makes no sense to me. There is an element of luck in a players batting average. Things tend to even out over time, but one year..hey, those bloops find holes. With UZR, this variation is more or less the norm, not a deviation.

 

I believe that there is less of a deviation for fielding on a year to year basis. You don't suddenly lose the ability to track fly balls, or the ability throw across a diamond. It deteriorates at a much slower rate, than say, the ability to turn on a fastball. This is one of the biggest problems I have with it.

 

The concept may be linear, but the effect is not linear IMO. It's logarithmic in it's application. It's not that I do not thing that a fielder should not get credit/failure. He is limited to the amount of plays he has an impact on. The number of plays should be factored in when determining a players overall defensive impact. Currently, it does not.

 

Not to mention that the statistic is mathematically flawed. I just don't see why you accept everything you read without question.

[ORS]

I understand your point, ORS, but adjusting it on a logarithmic level would eliminate many of the statistical deviations. Since the data itself is very limited, the variations are great.

 

Conceptually, I understand it. However, it's application is very limited.

 

Case in point...an established player, say a Jeter, or a Drew...his offensive numbers will basically stay the same with little variation. Similar OPS from year to year, with little deviation. However, looking at Jeter, there is major deviation in his UZR. This makes no sense to me. There is an element of luck in a players batting average. Things tend to even out over time, but one year..hey, those bloops find holes. With UZR, this variation is more or less the norm, not a deviation.

 

I believe that there is less of a deviation for fielding on a year to year basis. You don't suddenly lose the ability to track fly balls, or the ability throw across a diamond. It deteriorates at a much slower rate, than say, the ability to turn on a fastball. This is one of the biggest problems I have with it.

 

The concept may be linear, but the effect is not linear IMO. It's logarithmic in it's application. It's not that I do not thing that a fielder should not get credit/failure. He is limited to the amount of plays he has an impact on. The number of plays should be factored in when determining a players overall defensive impact. Currently, it does not.

 

Not to mention that the statistic is mathematically flawed. I just don't see why you accept everything you read without question.

You eliminate the "variations" (outliers) from the average, from the trend, not from the measurment.

 

You are looking at the numbers wrong in your comparison to the stability of offensive stats. The offensive stats look so stable because the opportunity is a part of the output of the measurement. Over a full season, a hitter that goes .333 then .300 the next year looks stable because his (approximated) 600 AB's are part of the measure. It's only a difference of 20 hits (1/30th of the data varied from one year to the next). In UZR, you are viewing it around average. A SS can go from a -5 to a +5, which seems like drastic variation, but a SS will see about 600 chances in a season, and with a single valued at 0.5 runs (approx) you are talking about 20 plays. Again, the 1/30th of the data varied from one year to the next, about the same as BA

 

It looks wrong to you because the perspective is different.

[Gom]

Not really. I stated before that I do not believe that there is that much variation in the field from year to year as there is in hitting.

 

We've made our points here..but now, how to tackle the inefficiency in firstbasemen?