Impact of Defense

[Brock Beauchamp]

 

Gutierrez also went from 105 to 87 to 54 OPS+ those three seasons.  And 3.3 to 2.2 to -0.3 oWAR (does not include fielding).  Are those "insane fluctuations" too?

Only if you can find a legitimate reason why a guy who was healthy, played in the same ballpark, played roughly the same number of innings, and stole more bases (at a much more successful clip, to boot) suddenly became a mediocre defender after a season as GOAT.

 

And while you're at it, do the same for Torii Hunter and his multiple 20 run defensive swings from season to season. Also Ian Kinsler.

 

Defense is the equivalent of a skill position. People in this thread are basically arguing that it's possible Randy Moss lost a step (and his hands) in one season but regained both the following season with no rhyme or reason for why it happened... But we should trust the numbers because they said it happened. We couldn't observe Randy losing his hands or speed, we just have to rely on the numbers because they said it's true.

 

Defense isn't like hitting or throwing a baseball, both of which are far more nuanced... not to mention that another player has direct input in the result of both. In 2010, nobody said "Franklin can't catch a curve ball" and changed their approach to his defense.

[Mike Sixel]

Uh, there will always be noise in the data. There is noise in the data produced by Billion dollar science experiments. There is noise in the data Boeing used to build airplanes. That's the nature of statistics and science.

[Mike Sixel]

 

Only if you can find a legitimate reason why a guy who was healthy, played in the same ballpark, played roughly the same number of innings, and stole more bases (at a much more successful clip, to boot) suddenly became a mediocre defender after a season as GOAT.

 

And while you're at it, do the same for Torii Hunter and his multiple 20 run defensive swings from season to season. Also Ian Kinsler.

 

Defense is the equivalent of a skill position. People in this thread are basically arguing that it's possible Randy Moss lost a step (and his hands) in one season but regained both the following season with no rhyme or reason for why it happened... But we should trust the numbers because they said it happened. We couldn't observe Randy losing his hands or speed, we just have to rely on the numbers because they said it's true.

 

Defense isn't like hitting or throwing a baseball, both of which are far more nuanced... not to mention that another player has direct input in the result of both. In 2010, nobody said "Franklin can't catch a curve ball" and changed their approach to his defense.

 

I gave you some possible explanations. It's a game of inches, maybe he just barely missed a bunch of balls one year.......maybe there is luck on both sides,and true talent is more in the middle of those results.

 

this seems to be the crux of the issue. Neither side buys the other side's explanations for why the math works, or doesn't work.

[Otto von Ballpark]

Brock, I think you are WAY under-estimating how hard it is to measure/estimate anything with more than one or two components. Baseball fielding is not like a 40 yard dash. Do you think we could easily calculate your abilities as a web designer over multiple projects, multiple years and not see a similar kind of swings in the result?

 

I too welcome our SporTrac overlords, but I think you might be surprised how much noise will remain in calculating something so complicated.

[jimmer]

 

I too welcome our SporTrac overlords, but I think you might be surprised how much noise will remain in calculating something so complicated.

Truth is, metrics will never come close to being generally accepted.  There will always be a way to be skeptical of them whether that skepticism is due to the info not jiving with preconceived notions set in stone or not jiving with basic common sense, or whatever.  And some skepticism is good, but it will never go away.

 

No matter how refined metrics get, they will certainly never come close to being as accepted the way fielding % and errors were for over 100 years or so, even though that is based on subjectiveness as well and completely discounts things like range.  But the math is easy for all to understand and it got drilled into most of our heads for so long and most of us just accepted it, at least for a little while.

 

 

[markos]

 

Only if you can find a legitimate reason why a guy who was healthy, played in the same ballpark, played roughly the same number of innings, and stole more bases (at a much more successful clip, to boot) suddenly became a mediocre defender after a season as GOAT.

 

And while you're at it, do the same for Torii Hunter and his multiple 20 run defensive swings from season to season. Also Ian Kinsler.

 

Defense is the equivalent of a skill position. People in this thread are basically arguing that it's possible Randy Moss lost a step (and his hands) in one season but regained both the following season with no rhyme or reason for why it happened... But we should trust the numbers because they said it happened. We couldn't observe Randy losing his hands or speed, we just have to rely on the numbers because they said it's true.

 

Defense isn't like hitting or throwing a baseball, both of which are far more nuanced... not to mention that another player has direct input in the result of both. In 2010, nobody said "Franklin can't catch a curve ball" and changed their approach to his defense.

There are three points that I'd like to bring up in response to this.

 

First, opportunity plays a huge role in the value that a defensive player is able to provide. As has been pointed out earlier in this thread, the vast majority of balls-in-play are either clearly hits or clearly outs. The gray middle ground between those two is where most of the defensive value comes from. And it is a lot of value! For an outfielder, the difference between making and missing a catch in that gray area is often the difference between an out and an extra base hit. Using the same linear weights used to determine batting values, that difference is at least a run, maybe even 1.25 runs. Even the difference between a single and an out is 0.75 runs. So making or missing these plays is huge. However, the amount of plays a player makes in the gray area is dependent largely on the number of opportunities. Unfortunately, that information isn't public, but I have to imagine that there is a fair amount of variability from season to season. Making up some numbers, a typical CF has roughly 500 chances in a given season. One year he might have 50 (10%) that fall in that high-value gray area. The next season he might only have 25 (5%). Let's say that he converts those plays into outs during both seasons at an 80% clip. That is 40 plays in the first season but only 20 in the second. That could be a difference in 20 runs, even though the player has demonstrated no difference in actual ability, only in opportunity.

 

Second, UZR values are relative to the league average for the position in question. So a +30 shortstop in one season may have very different absolute defensive ability than a +30 shortstop a different year. A player may not have any change in ability but score higher or lower in separate years based on the quality of defenders on other teams. I think this makes sense, as they are trying to capture the relative value that the defender provides the team at that position. If each team had a clone of prime Peter Bourjos, no team would have an advantage and the defensive value that the clones provided would be essentially be zero (notwithstanding the opportunity differences mentioned above). It doesn’t make any sense that every team could have a +20 UZR centerfielder. The value is relative. In Gutierrez's example, 2009 had both Vernon Wells and Gary Matthews Jr. log significant time in centerfield. I’m sure that brought down the average.

 

Third, (and probably least important for this) adjacent defenders matter. Having two good defenders next to one another allows both defenders to potentially accumulate more value. First, it allows them to position themselves more aggressively knowing that the other defender is able to cover more of the gap between them. Second, there should be fewer balls that get between the defenders. The penalty when balls fall between defenders is generally split between the defenders, so a bad defender can drag down the players next to them if they miss plays.

 

Brock, I get the sense that you expect UZR and the other fielding metrics to quantify defensive ability, but they do not do that. They accumulate the value of the plays made or missed by a defender, and how that value compares to the other defenders at the same position in the league. These are kind of related, as one hopes that the players with more defensive ability will make more high-value plays (and mess up fewer easy plays), and subsequently have higher scores. However, amazing defensive ability is neither necessary nor sufficient for having a high UZR. It is just an accounting system that knows that a specific liner in the gap falls for a double 90% of the time, and it frankly doesn't care at all how much effort or skill was needed to make the play. If the play was made, it is going to give that defender a bunch of credit regardless if that specific play ended up as a highlight reel catch. And if a defender doesn't get an opportunity to catch that liner because it is never hit, then tough.

[jay]

By WAR, the gap between the best and worst fielding teams is comparable to the gap between the best and worst hitting or pitching teams, true. Which seems wrong -- fielding isn't nearly as difficult as hitting or pitching, right?

 

But you have to consider there is immense competition between teams for hitting and pitching talent. Basically, if you live in North America and are capable of hitting or pitching in MLB, you are almost certainly in MLB. Fielding? Not so much. While fielders are scouted and graded too, very very few players make MLB for their fielding alone.

This is an interesting point, particularly of some teams don't value defensive data in the current state.

 

Could part of the measured gap in defense be due to a wider skill range between best and worst?

 

I think the effect could be very noticeable at "bat-first" positions like LF/RF (hence 2014 Gordon/Heyward WAR), but I'd think less so somewhere like CF/SS.

[USAFChief]

There are three points that I'd like to bring up in response to this.

 

First, opportunity plays a huge role in the value that a defensive player is able to provide. As has been pointed out earlier in this thread, the vast majority of balls-in-play are either clearly hits or clearly outs. The gray middle ground between those two is where most of the defensive value comes from. And it is a lot of value! For an outfielder, the difference between making and missing a catch in that gray area is often the difference between an out and an extra base hit. Using the same linear weights used to determine batting values, that difference is at least a run, maybe even 1.25 runs. Even the difference between a single and an out is 0.75 runs. So making or missing these plays is huge. However, the amount of plays a player makes in the gray area is dependent largely on the number of opportunities. Unfortunately, that information isn't public, but I have to imagine that there is a fair amount of variability from season to season. Making up some numbers, a typical CF has roughly 500 chances in a given season. One year he might have 50 (10%) that fall in that high-value gray area. The next season he might only have 25 (5%). Let's say that he converts those plays into outs during both seasons at an 80% clip. That is 40 plays in the first season but only 20 in the second. That could be a difference in 20 runs, even though the player has demonstrated no difference in actual ability, only in opportunity.

 

Second, UZR values are relative to the league average for the position in question. So a +30 shortstop in one season may have very different absolute defensive ability than a +30 shortstop a different year. A player may not have any change in ability but score higher or lower in separate years based on the quality of defenders on other teams. I think this makes sense, as they are trying to capture the relative value that the defender provides the team at that position. If each team had a clone of prime Peter Bourjos, no team would have an advantage and the defensive value that the clones provided would be essentially be zero (notwithstanding the opportunity differences mentioned above). It doesn’t make any sense that every team could have a +20 UZR centerfielder. The value is relative. In Gutierrez's example, 2009 had both Vernon Wells and Gary Matthews Jr. log significant time in centerfield. I’m sure that brought down the average.

 

Third, (and probably least important for this) adjacent defenders matter. Having two good defenders next to one another allows both defenders to potentially accumulate more value. First, it allows them to position themselves more aggressively knowing that the other defender is able to cover more of the gap between them. Second, there should be fewer balls that get between the defenders. The penalty when balls fall between defenders is generally split between the defenders, so a bad defender can drag down the players next to them if they miss plays.

 

Brock, I get the sense that you expect UZR and the other fielding metrics to quantify defensive ability, but they do not do that. They accumulate the value of the plays made or missed by a defender, and how that value compares to the other defenders at the same position in the league. These are kind of related, as one hopes that the players with more defensive ability will make more high-value plays (and mess up fewer easy plays), and subsequently have higher scores. However, amazing defensive ability is neither necessary nor sufficient for having a high UZR. It is just an accounting system that knows that a specific liner in the gap falls for a double 90% of the time, and it frankly doesn't care at all how much effort or skill was needed to make the play. If the play was made, it is going to give that defender a bunch of credit regardless if that specific play ended up as a highlight reel catch. And if a defender doesn't get an opportunity to catch that liner because it is never hit, then tough.

Help me with the math, please.

 

 

Which CFers had 500 chances last year?

 

The run value of a single is .75?

 

Where did you get the "opportunity" numbers from?

 

Help me with theory, please.

 

Are we sure a season of Gary Matthews Jr and Vernon Wells can really change the baseline? I thought UZR was based off a rolling average of plays made and not made over 6 years..is that incorrect? If so, a season of those two should have very little impact on Franklin Gutierrez, correct? Maybe no impact at all, unless you believe CF never had players like Matthews and Wells before, and never will again?

 

EDIT: thanks Jimmer, for posting the primer below and confirming the six year rolling data used for comparison. That seems to rule out "other bad CFer" as a possible cause for the wild swings in Gutierrez' UZR, no?

[jimmer]

http://www.fangraphs.com/library/defense/uzr/

 

'For the details on how UZR is calculated — i.e. how we can attach a run value to defensive events — see the FanGraphs UZR Primer. It’s very thorough. The specifics can be a little overwhelming, but at it’s most basic level, it’s a measure of the average amount of damage that batted ball would do and how often it is converted into an out, relative to average at the position. So if the average left fielder makes a player 40% of the time on the ball in question and that batted ball (based on location, speed, etc) is worth 0.8 runs on average, fielding it cleanly earns you 0.48 runs toward your UZR (0.8*0.6). It’s a little more nuanced than that, but that’s the view from 30,000 feet.'

 

And if one wants to read the primer

 

http://www.fangraphs.com/blogs/the-fangraphs-uzr-primer/

 

'Now, let’s say that we want to compute a UZR for every player in 2009. For every batted ball, either it is caught by a fielder and turned into an out (either the batter or a base runner is out, or both of course), it is scored as a hit, or the batter reaches on an error or a fielder’s choice (and no out is made). One or more fielders will receive positive or negative credit depending on the outcome of the play and depending on how often that same batted ball in that same situation (outs, base runners, attributes of the batter, etc.) was successfully fielded by each fielder from 2004-2009.

 

Let’s say that that same batted ball in the example above was caught by the CF’er on the first play of a game. Since typically someone will catch that same ball only 25% of the time (see above), this particular CF’er will get credit for an extra .75 plays – 100% minus 25%. We then convert .75 plays into runs by multiplying .75 by the difference between an average hit in that location and the average value of an air ball out. A typical outfield hit is worth around .56 runs and any batted ball out is worth around -.27 runs, so the difference between a hit and an out is worth around .83 runs. (We don’t vary the value of the hit or out based on the outs or base runners because we want “game situation-neutral” defensive evaluations.) Since our fielder gets credit for .75 extra plays, we give him credit for .75 times .83 runs, or +.6255 runs for that play.'

[jimmer]

Why UZR:

 

'This isn’t the right place to debate UZR versus another similar metric, but you should use a metric like UZR or DRS because it is a better representation of defensive value than something like fielding percentage. Even your eyes aren’t going to do a great job measuring defensive performance because you simply can’t watch and remember enough plays a year to have a good sense of exactly how well a player stacks up against the competition. You might be able to judge a single play better than the metrics (although that’s debatable), but your ability to recall every play and compare them is limited. Run value defensive stats like UZR provide you with the best estimate of defensive value currently available  and allow you to estimate how much a player’s defense has helped his team win.'

[USAFChief]

Why UZR:

 

'This isn’t the right place to debate UZR versus another similar metric, but you should use a metric like UZR or DRS because it is a better representation of defensive value than something like fielding percentage. Even your eyes aren’t going to do a great job measuring defensive performance because you simply can’t watch and remember enough plays a year to have a good sense of exactly how well a player stacks up against the competition. You might be able to judge a single play better than the metrics (although that’s debatable), but your ability to recall every play and compare them is limited. Run value defensive stats like UZR provide you with the best estimate of defensive value currently available  and allow you to estimate how much a player’s defense has helped his team win.'

Your honor, the prosecution will stipulate to the fact the proponents of UZR think its great.

[jimmer]

 

Your honor, the prosecution will stipulate to the fact the proponents of UZR think its great.

Quite true, quite true. :-)

 

But the reasoning seems sound since one person can only watch so much baseball, and most do it for enjoyment.  Unlike the people who do the metrics, an individual fan doesn't watch the games as their job to gauge how each player's defense compares to the other  29 teams' players at the same position.

 

He doesn't say UZR is perfect, he says it's the best option currently available.

[Otto von Ballpark]

 

Help me with the math, please.


Which CFers had 500 chances last year?

The run value of a single is .75?

Where did you get the "opportunity" numbers from?

Help me with theory, please.

Are we sure a season of Gary Matthews Jr and Vernon Wells can really change the baseline? I thought UZR was based off a rolling average of plays made and not made over 6 years..is that incorrect? If so, a season of those two should have very little impact on Franklin Gutierrez, correct? Maybe no impact at all, unless you believe CF never had players like Matthews and Wells before, and never will again?

EDIT: thanks Jimmer, for posting the primer below and confirming the six year rolling data used for comparison. That seems to rule out "other bad CFer" as a possible cause for the wild swings in Gutierrez' UZR, no?

 

Chief, don't let the math and markos's one erroneous point cloud the issue here.  The rest of markos's post was pretty great.  And yes, the run value of hits to the outfield is pretty high.

 

His point about chances is great.  Say, for example, that in 2009 Gutierrez got 20 balls hit to him in very difficult but not impossible places (in the gaps, near the wall), and he is pretty darn good so he converted 15 and only missed 5.  That's going to really increase his UZR because those are very difficult, valuable plays.  But they are so infrequent, it is quite possible that the following season, he only got 10 similar balls hit to him due to any number of factors (random chance, weather, pitchers, hitters, etc.).  And if again he misses 5, that leaves 10 fewer big value plays that he converted.  Through no fault of his own, he may not have had the same chance to add that value in 2010 as he did in 2009.   That alone can be responsible for a big part of his UZR swing.

 

More factors: say Gutierrez in 2009 had an arrangement with his outfield mates and coaches that he could position himself to get more of those very difficult, big value plays.  But in 2010, maybe his teammates changed, maybe his coaches changed, or maybe the teammates and coaches stayed the same but just wanted to play differently.  More lost chances at big value plays.  (I wonder if this is a factor in Hunter's UZR drop from LA to Detroit.)

 

And while none of these should cause a 30 run swing alone, maybe one or more of these factors was at play, plus perhaps Gutierrez uncharacteristically missing a few more balls in 2010, plus the noise/error inherent in the samples, and all of a sudden, Gutierrez might be an outlier.

 

That's definitely an issue with small samples of UZR.  Very few people suggest trusting a single season for a single player.  3 seasons for a player, or a season for a whole team, though, and the sample gets a lot larger.  And not only do some of the sample oddities tend to even out a bit, but over the larger sample, the player/team should be expected to adapt to some of these circumstances (i.e. by his second season in Detroit, perhaps Hunter should have borne some responsibility for better tailoring his defensive approach in the absence of Trout and Bourjous).

[Otto von Ballpark]

As for why team UZR doesn't necessarily match on-field run differentials (your original question that kinda sparked the discussion), remember the UZR is a using average run values for plays.  It's not based on actual game state or sequencing.  The Royals may have made a lot more theoretically valuable fielding plays than the Indians in 2014, but perhaps more of them came to lead off an inning or with fewer men on base (sequencing noise).  And/or the Indians had the misfortune of making some miscues of equal value as the Royals at inopportune times.

[Otto von Ballpark]

As to the general tone, I think most of us are in agreement, but some of us are more accepting of what an outlier value means.  Outliers should invite some skepticism about that specific value, but not necessarily about the metric in general (which is the more common skepticism being expressed here).

[markos]

 

Are we sure a season of Gary Matthews Jr and Vernon Wells can really change the baseline? I thought UZR was based off a rolling average of plays made and not made over 6 years..is that incorrect? If so, a season of those two should have very little impact on Franklin Gutierrez, correct? Maybe no impact at all, unless you believe CF never had players like Matthews and Wells before, and never will again?

EDIT: thanks Jimmer, for posting the primer below and confirming the six year rolling data used for comparison. That seems to rule out "other bad CFer" as a possible cause for the wild swings in Gutierrez' UZR, no?

 

I still think "other bad CFers" matters. If you look at the UZR primer, it has the following section (note the part I bolded):

 

"Again, the baseline “catch rates” for all of the various “buckets” (batted ball types, speed, locations, etc.) are based on 6 years of data. That is an arbitrary number. It could be 3 years and it could be 10. I chose 6 years in order to accumulate fairly large samples of data in each bucket. The UZR numbers (each player’s runs saved above or below average) are initially presented as “as compared to the average player at all MLB parks over the 6-year baseline.” The numbers you see on Frangraphs, however, are scaled to an average player at each position for that league and year. So if you add up everyone’s UZR (the seasonal to-date or end-of-year totals – not the “per 150” rate) in any given year and league, it will sum to zero (or close to it because of rounding errors). Because of that, there is no guarantee that an average player in any one league or year is equal to an average player in any other league or year, even the same player. For example, let’s say that a certain player was zero in 2009 and 2008, thus you consider him to be an average defender for those two years, and you assume that his defensive ability or performance did not change from one year to the next. However, if in one of those years, the overall quality of defense in that player’s league was better or worse than the other year, the player may have actually gotten better or worse himself even though his UZR is zero in both years. That is a minor point, but it is something to keep in mind."

[Brock Beauchamp]

To be clear, I think defensive metrics do the best job they can with the data provided. Most of the time, I think they're probably reasonably accurate over multiple seasons when used in relative terms (eg. the Twins had one of the worst outfields in baseball over the past several years).

But there are enough issues with the metrics that I don't trust them enough to assign run values to defense or believe them when they say things about single seasons like "the Royals defense was 140 runs better than the Indians in 2014".

I applaud those who work in metrics and all the progress they've made with defensive numbers in the past decade... But it doesn't take much digging into the numbers to see that a lot more work needs to be done.

[Otto von Ballpark]

But there are enough issues with the metrics that I don't trust them enough to assign run values to defense or believe them when they say things about single seasons like "the Royals defense was 140 runs better than the Indians in 2014".

 

I applaud those who work in metrics and all the progress they've made with defensive numbers in the past decade... But it doesn't take much digging into the numbers to see that a lot more work needs to be done.

The run values are just estimates for results that happened. Just like there would be run values assigned for Joe Mauer's hits in a metric. It doesn't explain all of the factors behind those hits, but it is really the only way to "score" the results outside some kind of Win Probability Added scheme, which isn't all that useful.

 

And fundamentally, any defensive metric based on FieldFX data is going to do the same thing. They will just be able to break down the result further, like isolate positioning, first step, etc. The total result will likely be arrived at in much the same way as UZR. All you can do is grade on the opportunities that happen, and compare to the average result of such opportunities.

 

And given the sample sizes and the varied human factors involved, you will probably see similar noise/outliers/ranges in the total result. (And I also suspect the new FieldFX system will be even more proprietary and "black box" than current metrics, fueling even more of your skepticism.)

[Otto von Ballpark]

Thanks Markos. I still have a lot to learn about UZR.

 

I think it would be incredible if someone would list all of a player's chances, the average run value, and the assigned run value. Maybe a further analysis could identify the types of plays that were costing Hunter, helping Gutierrez, etc. I am sure a lot of the zone data isn't free, though.

[jay]

And fundamentally, any defensive metric based on FieldFX data is going to do the same thing. They will just be able to break down the result further, like isolate positioning, first step, etc. The total result will likely be arrived at in much the same way as UZR. All you can do is grade on the opportunities that happen, and compare to the average result of such opportunities.

Spot on. Even with better input data, there really isn't a better way to score it. We'll never know what would have happened after a different outcome of that exact play. However, there's been enough baseball games played to have a very precise idea of the average of what would have happened.

[USAFChief]

Quite true, quite true. :-)

 

But the reasoning seems sound since one person can only watch so much baseball, and most do it for enjoyment.  Unlike the people who do the metrics, an individual fan doesn't watch the games as their job to gauge how each player's defense compares to the other  29 teams' players at the same position.

 

He doesn't say UZR is perfect, he says it's the best option currently available.

I agree judging defense is difficult for fans. Perhaps too difficult.

 

I think it is much less difficult for team officials--managers, GMs, even players--who spend a great deal of time around the game and probably have a pretty good idea of what good defense looks like.

 

I also don't take on faith that UZR is "the best option currently available." Maybe it is, but how would we know that? Compared to what?

 

Even if I grant that UZR is the best option available, then what? What does that tell us? What should it be used for?

 

If I knew that 20 MLB teams were using UZR to make playing time decisions, I'd take it more seriously.

[USAFChief]

Chief, don't let the math and markos's one erroneous point cloud the issue here.  The rest of markos's post was pretty great.  And yes, the run value of hits to the outfield is pretty high.

 

His point about chances is great.  Say, for example, that in 2009 Gutierrez got 20 balls hit to him in very difficult but not impossible places (in the gaps, near the wall), and he is pretty darn good so he converted 15 and only missed 5.  That's going to really increase his UZR because those are very difficult, valuable plays.  But they are so infrequent, it is quite possible that the following season, he only got 10 similar balls hit to him due to any number of factors (random chance, weather, pitchers, hitters, etc.).  And if again he misses 5, that leaves 10 fewer big value plays that he converted.  Through no fault of his own, he may not have had the same chance to add that value in 2010 as he did in 2009.   That alone can be responsible for a big part of his UZR swing.

 

More factors: say Gutierrez in 2009 had an arrangement with his outfield mates and coaches that he could position himself to get more of those very difficult, big value plays.  But in 2010, maybe his teammates changed, maybe his coaches changed, or maybe the teammates and coaches stayed the same but just wanted to play differently.  More lost chances at big value plays.  (I wonder if this is a factor in Hunter's UZR drop from LA to Detroit.)

 

And while none of these should cause a 30 run swing alone, maybe one or more of these factors was at play, plus perhaps Gutierrez uncharacteristically missing a few more balls in 2010, plus the noise/error inherent in the samples, and all of a sudden, Gutierrez might be an outlier.

 

That's definitely an issue with small samples of UZR.  Very few people suggest trusting a single season for a single player.  3 seasons for a player, or a season for a whole team, though, and the sample gets a lot larger.  And not only do some of the sample oddities tend to even out a bit, but over the larger sample, the player/team should be expected to adapt to some of these circumstances (i.e. by his second season in Detroit, perhaps Hunter should have borne some responsibility for better tailoring his defensive approach in the absence of Trout and Bourjous).

I don't think it was one error, but whatever.

 

It seems at odds with the whole "it's science' argument to argue "don't let the math cloud the issue," no?

 

"Maybe this" or "possibly that" are at odds with the entire premise of saber metrics.

[Otto von Ballpark]

I don't think it was one error, but whatever.

 

It seems at odds with the whole "it's science' argument to argue "don't let the math cloud the issue," no?

 

"Maybe this" or "possibly that" are at odds with the entire premise of saber metrics.

The math comment was about your questioning markos's "500 chances" example, the fitting estimated UZR into unadjusted real world run differentials, etc. You seem to not understand what UZR is trying to do, which sort of precludes detailed mathematical discussion of it.

 

If you think sabermetrics has no room for maybe's or possibly's, you are probably railing against a caricature of a sabermetrician than the actual field. (Admittedly a few people probably resemble that caricature, but they are not here.)

[USAFChief]

The math comment was about your questioning markos's "500 chances" example, the fitting estimated UZR into unadjusted real world run differentials, etc. You seem to not understand what UZR is trying to do, which sort of precludes detailed mathematical discussion of it.

If you think sabermetrics has no room for maybe's or possibly's, you are probably railing against a caricature of a sabermetrician than the actual field. (Admittedly a few people probably resemble that caricature, but they are not here.)

What IS UZR trying to do?

 

By all means, please include all the detailed math you feel necessary. I think it's fair for me to examine the math, though, and if I think it doesn't add up, to ask for clarification.

 

For example, if you try to convince me that there's nothing odd about numbers that appear odd at first blush, I will examine the use of hypotheticals to explain those odd numbers. Particularly when they are not only hypotheticals, but hypotheticals based on numbers that are themselves at best questionable, at worst complete exaggeration s used to make a point seem stronger.

[jimmer]

 

Even if I grant that UZR is the best option available, then what? What does that tell us? What should it be used for?

If I knew that 20 MLB teams were using UZR to make playing time decisions, I'd take it more seriously.

What I normally use defensive metrics for is simply rankings, whether it's teams or players.  Same thing with offensive metrics and WAR overall.

[USAFChief]

What I normally use defensive metrics for is simply rankings, whether it's teams or players.  Same thing with offensive metrics and WAR overall.

Fair enough.

 

I will wait for something I trust.

[TheLeviathan]

 

I don't think it was one error, but whatever.

It seems at odds with the whole "it's science' argument to argue "don't let the math cloud the issue," no?

"Maybe this" or "possibly that" are at odds with the entire premise of saber metrics.

 

I'm not sure they are at odds Chief, but it seems like it at times.  From my vantage point SABR stats are meant to better approximate value by looking deeper into what a player can control.

 

I think that's great, but where I think many SABR people lose those that they want to join them is with forgetting to differentiate that they are approximating value, not precisely or perfectly measuring it.  That's an important distinction IMO.  In sports, many of the stats that we see are not approximations.  A QB's completion rate, a player's RBI total, a player's shooting percentage, etc.  I don't think enough time and effort is made to make clear that these two sorts of statistics are not in quite the same family.  It may seem trivial or ridiculous, but I disagree.  People bristle that they have to have some sort of caveat or disclaimer everytime...but maybe that's what it has to be.  The SABR community wants to be more largely accepted by the sport they follow, so maybe they need to make it clear what they are actually doing.  They are using science/math to approximate what is largely thought to be subjective, not necessarily giving a precise measurement.

 

At the same time, that community (and the post that started this comes to mind) has to be careful about the manner in which they state their statistics to not make them sound so concrete.  It isn't that hard to phrase your use of UZR or WAR or whatever else in a way that won't make people get stuck on 120 runs vs. 100 runs and yet there is virtually no effort to do so.  That's what confuses people IMO.  

 

It also doesn't help when you openly acknowledge in your primers that UZR doesn't factor in positioning but others claim it does.  Or that you should use UZR in larger samples for the most reliability and then have that very same site post about defensive rankings and gush about a one year difference between KC and CLE.  Or to say UZR is less reliable the smaller the sample size but use WAR over short periods with little thought to how that might be impacting the numbers. It sends mixed messages about the stats and how to properly interpret them or use them.  As a way or ranking players over periods of time they are great, but that isn't the limits of their use and that's what causes some confusion.

 

Just my two cents.

[Otto von Ballpark]

 

It also doesn't help when you openly acknowledge in your primers that UZR doesn't factor in positioning but others claim it does.

It really doesn't help when people can't distinguish "isolate the effect of positioning" from "factor in positioning"

[Otto von Ballpark]

 

 

I think that's great, but where I think many SABR people lose those that they want to join them is with forgetting to differentiate that they are approximating value, not precisely or perfectly measuring it.

Also, I am pretty sure that very very few people involved in sabermetrics think they are precisely or perfectly measuring anything.  It's widely acknowledged that the only precise, perfect measurement is the one that shows up on the scoreboard; sabermetrics is all about estimating the myriad factors that go into that simple number.

[TheLeviathan]

 

It really doesn't help when people can't distinguish "isolate the effect of positioning" from "factor in positioning"

 

Straight from Fangraphs:

 

In general, UZR isn’t perfect because it doesn’t factor in shifts, positioning, and can’t perfectly measure everything it needs to, but it’s still among the best options out there.

 

So, please, take your issue up with them and their phrasing.