Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Baseball Newsstand > Baseball Primer Newsblog > Discussion
Baseball Primer Newsblog
— The Best News Links from the Baseball Newsstand

Friday, August 29, 2014

FG (Zimmerman): Alex Gordon, UZR, and Bad Left Field Defense

Duh, but also, thanks, Jeff.

Outfield defense has three components, Range (ability to field balls), Arm and Errors. [Alex] Gordon has a great arm and his LF values since 2011 have ranged from 8 Runs to 11 Runs, so no change there. His Error component has gone from 0 to 2 Runs. Again, not the reason for the jump. The key difference is in his Range value at 14 Runs because his previous high was just 4 Runs…

Here are the league average values for in and out-of-zone values compared to Gordon’s values and the difference.

In Zone
Year: League%, Gordon%, Difference
2010: 87.1%, 91.7%, 4.6%
2011: 90.4%, 92.3%, 1.9%
2012: 89.7%, 91.3%, 1.6%
2013: 90.6%, 91.8%, 1.2%
2014: 88.4%, 91.6%, 3.2%

Out-of-Zone
Year: League Rate, Gordon Rate, Difference
2010: 0.049, 0.043, -0.006
2011: 0.065, 0.057, -0.008
2012: 0.058, 0.062, 0.004
2013: 0.066, 0.081, 0.015
2014: 0.057, 0.082, 0.025

With individual position UZR values, the baseline values can move around quite a bit. There are only 30 inputs into the baseline values, so if just a few players move to a new position or get hurt, the zero value can change… Three of the top eight [fielders by LF range in 2013] haven’t played in the majors this season…

As I’ve noted before, this is one possible area of consideration for the UZR. A potential improvement to the system for the future, and one considered in that linked piece, is a baseline defensive value which is constant from year to year in addition to just UZR… The first 500 words of this article wouldn’t be necessary if [this] metric were available. It could show Gordon’s Range Value is at a +4 Historic Runs (hypothetically speaking) compared to +14 2014 Runs. Maybe each player doesn’t need two values, maybe just a comparison value for the entire season could be available.

Putting it all together, the jump in Alex Gordon‘s total WAR to a league leading value really has nothing to do with Alex Gordon. He is the same defensive player he has been over the past few season. What has happened is the league wide level in talent has fallen off in left field thereby boosting Gordon’s numbers substantially.

The District Attorney Posted: August 29, 2014 at 01:37 PM | 47 comment(s) Login to Bookmark
  Tags: alex gordon, defense, royals, sabermetrics

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. cardsfanboy Posted: August 29, 2014 at 05:19 PM (#4782005)
Excellent article. I like the idea of a non-seasonal adjusted stat so that we can get a clue about what the static performance of players was.

People have mentioned in the past that defenders on the corners today may not be as good as they were in the past, and defenders during the sillyball era was probably the worst collective group. A non-adjusted number would be a way of checking the veracity of something like those statements.
   2. PreservedFish Posted: August 29, 2014 at 05:36 PM (#4782012)
Wow. Short but a great read. We have all been wondering why the #### Gordon's UZR seemed so inaccurately high, and bam, here's the answer.
   3. cardsfanboy Posted: August 29, 2014 at 05:59 PM (#4782019)
Even though his numbers are high, when it's all said and done, it means that if the stat is accurate for the reasons the article gives, then his value is higher. Just like hitting 30 homeruns in 1967 is different than hitting 30 in 2007, same with his relative defensive performance by eras.
   4. The District Attorney Posted: August 29, 2014 at 06:22 PM (#4782027)
Just like hitting 30 homeruns in 1967 is different than hitting 30 in 2007, same with his relative defensive performance by eras.
Absolutely not. That would only be true if the conditions of the game itself changed, which is not the case. When Andy Dirks got hurt for the Tigers, that didn't make left field a more difficult position for Alex Gordon to play in Kansas City.

Now, I don't know how we would identify when changes in average defensive stats represent actual changes in the game of baseball. I probably am overreaching to say that I know that LF didn't become more difficult to play this year, rather than that the former top defensive LF got benched/injured/changed positions. The latter sure as hell seems like the far more likely theory, though.
   5. PreservedFish Posted: August 29, 2014 at 06:29 PM (#4782029)
Absolutely not. That would only be true if the conditions of the game itself changed, which is not the case. When Andy Dirks got hurt for the Tigers, that didn't make left field a more difficult position for Alex Gordon to play in Kansas City.


Exactly. If I was convinced that there was some reason that it was more difficult to find solid LFers in 2014 than it was in 2013, I'd be inclined to give Gordon the credit. Or if there were some environmental change that made it more difficult for both he and his colleagues to catch the ball. But neither seems plausible to me. If UZR is really using a yearly average based on these small rickety samples, then it makes the numbers even less trustworthy than I thought they were previously.
   6. Captain Supporter Posted: August 29, 2014 at 06:34 PM (#4782030)
GIGO. When stats don't make sense, rather than go through tortuous reasoning to try to justifying them just admit that they have significant flaws and should be used with a great deal of caution.
   7. cardsfanboy Posted: August 29, 2014 at 06:36 PM (#4782032)
Absolutely not. That would only be true if the conditions of the game itself changed, which is not the case. When Andy Dirks got hurt for the Tigers, that didn't make left field a more difficult position for Alex Gordon to play in Kansas City.


It's not about the game itself changing, it's about the relative value that those skills provide. War or other value stats are not saying his skills or ability is better, just that his skills or ability in the current context is more valuable.

Same with a right handed power hitter in a stadium with a short left field, nobody is saying that he is a better player than a spray hitter on the same team, just that his value might end up being more valuable.
   8. PreservedFish Posted: August 29, 2014 at 06:42 PM (#4782035)
War or other value stats are not saying his skills or ability is better, just that his skills or ability in the current context is more valuable.


But for me you need to define "more valuable." If the Padres' leftfielder hurts his ankle and it lowers the overall UZR baseline, explain to me how Alex Gordon is now more valuable in any real way.

I suppose the same issue is at play with OPS+ and other context-adjusted stats (if Miguel Cabrera is injured everyone else looks better), but the small sample for UZR seems to make it particularly vulnerable to such quirks.
   9. Robert in Manhattan Beach Posted: August 29, 2014 at 06:48 PM (#4782038)
Let's say an in-his-prime Andruw had played LF for a season. He plays it brilliantly and because he's now being compared to guys mostly in the lineup for their bat, he rates out far, far better than the LF competition. So Rfield or whatever spikes big time. WAR thinks he the second coming. So Andruw, who's the same guy he's always been, is now rated as one of the most valuable players in the league despite moving to a non-premium position. This is no good.
   10. PreservedFish Posted: August 29, 2014 at 06:53 PM (#4782041)
Let's say and in-his-prime Andruw had played LF for a season.

WAR is supposed to be able to handle this, but if Sullivan is correct that the whole year's average can be thrown off by the fact that guys like Dirks, Torres and Tuiasasopo haven't played, I'm skeptical.
   11. cardsfanboy Posted: August 29, 2014 at 07:12 PM (#4782048)
Let's say an in-his-prime Andruw had played LF for a season. He plays it brilliantly and because he's now being compared to guys mostly in the lineup for their bat, he rates out far, far better than the LF competition. So Rfield or whatever spikes big time. WAR thinks he the second coming. So Andruw, who's the same guy he's always been, is now rated as one of the most valuable players in the league despite moving to a non-premium position. This is no good.


It's not good in some respects, but it's accurate in regards to value. Let's say that in a year that the top five first baseman in the game all suffer season ending injuries early in the month. All five are replaced with pure replacement level players. The average offensive value for a first baseman has gone down, making what would be "average" in the year prior and the following year higher. So putting up a 3.0 war in those previous and next season would now be a 3.5 or even a 4.0 war, without one single change in performance.

This is a feature of a stat based upon an average(or theoretical replacement level) and not a flaw. It's something everyone probably already knows even if they don't voice it. I'm not sure what people are expecting for a value stat to be about. Should the overall average of all stats(offense and defense) be based upon a three year rolling average? Sounds like a good idea, but not for War. When measuring wins, you are talking about on a seasonal basis, since that is all that really matters. Nobody is asking that Cy Young awards be based upon a three year rolling average, why would any other stat be based upon that?*



*Park factors should probably be based upon three year average, simply because you don't have enough visitor variety of data to create an accurate rating. With defense, the data isn't too little, it's just subject to factors such as a fluke game or not, but it's still enough data to create a value stat, just not enough to create an accurate skill evaluation.
   12. BDC Posted: August 29, 2014 at 07:13 PM (#4782049)
In terms of wins and losses, though, all of sports is based on being better than your opponents. As noted by cfb, WAR is about winning, not about absolute athletic targets.

I'm still trying to see how Elvis Andrus cratered so much this year. It does not seem to be that bunches of great SS entered the league. Quite the reverse: Jeter replaced Brendan Ryan, yet Andrus rates worse than Jeter this year, though he was ahead of Ryan last year.
   13. Petunia inquires about ponies Posted: August 29, 2014 at 07:40 PM (#4782065)
And WAR is about replacement.

It doesn't seem to be what we want out of the stat, but I can see the logic there. To answer the question in [8], when the Padres' left fielder gets hurt, the average ability of a real replacement level LF just went down, because the pool of actual replacement players was just reduced by one. So Gordon's WAR goes up because the Padres' LF is now not available to replace Alex Gordon if he gets run over by a bus. It does hang together.
   14. PreservedFish Posted: August 29, 2014 at 07:49 PM (#4782069)
*Park factors should probably be based upon three year average, simply because you don't have enough visitor variety of data to create an accurate rating. With defense, the data isn't too little, it's just subject to factors such as a fluke game or not, but it's still enough data to create a value stat, just not enough to create an accurate skill evaluation.


Not seeing the distinction here.
   15. Bhaakon Posted: August 29, 2014 at 07:50 PM (#4782072)
This is a feature of a stat based upon an average(or theoretical replacement level) and not a flaw. It's something everyone probably already knows even if they don't voice it. I'm not sure what people are expecting for a value stat to be about.


I agree in principle, but the way WAR is used in most internet baseball discussions is not in the context of player contributions adding to team wins, but in things like "Who should be the MVP?" or "Is player X better than player Y?", in which case this flaw/feature of WAR is a new and problematic wrinkle.

No one wants to make the argument that so-and-so should be the MVP because four of the top five players at his position came up lame. Does that make him more valuable to his team in the context of that season? Most certainly. Does that make him a better player? No.
   16. cardsfanboy Posted: August 29, 2014 at 07:59 PM (#4782078)
Not seeing the distinction here.


Seasonal park factors are figured with very few games against opponents outside of your division, so there just isn't a large enough sample size to have any conclusive evidence. 81 games, with 40 of them being against 4 teams, doesn't present much data and the data could be skew dramatically by having the unfortunate luck of getting Kershaw/Greinke/etc on the road and Lincecum/and others at home during a couple of series. Eventually I imagine someone is going to create a formula for teasing out quality of pitchers etc but for now, most park factors don't drill down to that level. Park factors are an adjustment made to the run value of a player's offensive/pitching performance based upon what the expected average performance is in that park.

Defense is different, it's not trying to determine quality of a park or player, but on the value that the level of defense provided in comparison to other players. 3 years of data is probably good at providing an idea on the quality of the players defense, but isn't at all useful in determining whether the actual defense contributed to a win for that season.


Not sure if there is a good way for me to explain it, but to me park factors are an adjustment made to the numbers based upon "environmental factors". While defense is the exact same as offense in regards to a value stat.
   17. cardsfanboy Posted: August 29, 2014 at 08:01 PM (#4782079)
I agree in principle, but the way WAR is used in most internet baseball discussions is not in the context of player contributions adding to team wins, but in things like "Who should be the MVP?" or "Is player X better than player Y?", in which case this flaw/feature of WAR is a new and problematic wrinkle.


Who should win mvp is exactly what War is designed for. Being the best player doesn't mean much if you don't perform in a way that helps your team win.
   18. The District Attorney Posted: August 29, 2014 at 08:09 PM (#4782085)
Let's say that in a year that the top five first baseman in the game all suffer season ending injuries early in the month. All five are replaced with pure replacement level players. The average offensive value for a first baseman has gone down, making what would be "average" in the year prior and the following year higher. So putting up a 3.0 war in those previous and next season would now be a 3.5 or even a 4.0 war, without one single change in performance.

This is a feature of a stat based upon an average(or theoretical replacement level) and not a flaw.
No, no, no. If all the best hitters in the league got hurt, it would change the run environment. It would take fewer runs to win a game. Each run created would, therefore, indeed be more valuable.

What we're talking about here didn't change the run environment. David Murphy moves from LF to RF. If you want to say that that makes Alex Gordon a relatively better defensive LF and Nori Aoki a relatively worse defensive RF, I won't stop you. But I do have to stop you if you want to give Gordon credit for "runs saved" due to this move, when, on a team level, nothing happened! No runs were, in fact, saved! The Kansas City Royals did not benefit!

Do you really want to say, in that situation, that Gordon saved runs and Aoki gave those same runs back? That would be the ultimate example of the statistician who says you're comfortable because one foot is in a fire and the other foot is in an ice bucket.
   19. God Posted: August 29, 2014 at 08:12 PM (#4782089)
No runs were, in fact, saved! The Kansas City Royals did not benefit!

Don't they benefit by one of their opponents fielding a worse defensive lineup?
   20. Brian White Posted: August 29, 2014 at 08:15 PM (#4782092)
Let's say that in a year that the top five first baseman in the game all suffer season ending injuries early in the month. All five are replaced with pure replacement level players. The average offensive value for a first baseman has gone down, making what would be "average" in the year prior and the following year higher.


Is the offensive part of WAR dependent on year-to-year collective performances at a single position, in the same way UZR uses? I thought hitters were compared to the performance of all MLB hitters, which aren't nearly as subject to this kind of goofy variation.
   21. PreservedFish Posted: August 29, 2014 at 08:16 PM (#4782094)
Don't they benefit by one of their opponents fielding a worse defensive lineup?


Let's say no. Let's say the Indians just decided to shift their poor defender from one corner to another. There's no net effect on runs scored / allowed.
   22. Bhaakon Posted: August 29, 2014 at 08:20 PM (#4782097)
Who should win mvp is exactly what War is designed for. Being the best player doesn't mean much if you don't perform in a way that helps your team win.


It may have been designed for that, but this feature/flaw makes it less useful for that purpose. As far as I'm concerned, basing MVP votes on a flukey single-season positional adjustment skewed by a rash of injuries is about as legitimate as refusing to vote for players on losing teams because 'value' that doesn't get a team to the postseason is wasted. In both case you're rewarding/punishing a player for something he has no control over (namely, the performance of other individuals).
   23. Infinite Joost (Voxter) Posted: August 29, 2014 at 08:22 PM (#4782099)
People have mentioned in the past that defenders on the corners today may not be as good as they were in the past, and defenders during the sillyball era was probably the worst collective group.


I've never seen any evidence for this other than the reflexive nostalgia for the baseball of one's youth. It's also completely illogical, as in at least one league there is now a place to hide a terrible defender.
   24. Zach Posted: August 29, 2014 at 08:24 PM (#4782102)
The original sin of WAR is that offensive values are well established and uncontroversial, but defensive values are difficult to measure and require a lot of regression.

Splitting it up into two statistics makes a lot of sense. Why mess up the good statistic by adding it to the bad one?
   25. cardsfanboy Posted: August 29, 2014 at 08:28 PM (#4782104)
Let's say no. Let's say the Indians just decided to shift their poor defender from one corner to another. There's no net effect on runs scored / allowed.


There are a lot of flaws with the way defensive stats are figured, I'm not for one second arguing against that obvious point. Only thing I'm arguing about is that war is a value stat, it's based upon relative performance in comparison to an average/replacement player at that position as determined by the average performance that season.

I will support many adjustments to how defensive stats are figured, but once the methodology is determined, basing it upon individual positions and average is pretty much the only way to go, and in a value stat with war, single season is the only reasonable method. As I said in my first comment, I would love to see a "raw" stat like mentioned in this article, but it wouldn't be a useful component in a war type of stat that is based upon a theoretical replacement level.

Is the offensive part of WAR dependent on year-to-year collective performances at a single position, in the same way UZR uses? I thought hitters were compared to the performance of all MLB hitters, which aren't nearly as subject to this kind of goofy variation.


I was under the impression that the positional adjustment to war is figured annually, so it's possible 1200 innings at short in 2007 is a slightly smaller value than in 2008. Of course that number is well below rounding error usually so we never see it on the bb-ref.

   26. cardsfanboy Posted: August 29, 2014 at 08:34 PM (#4782108)
It may have been designed for that, but this feature/flaw makes it less useful for that purpose. As far as I'm concerned, basing MVP votes on a flukey single-season positional adjustment skewed by a rash of injuries is about as legitimate as refusing to vote for players on losing teams because 'value' that doesn't get a team to the postseason is wasted. In both case you're rewarding/punishing a player for something he has no control over (namely, the performance of other individuals).


I guess it depends on which camp you are in, in regards to war for pitchers. I despise war for pitchers based upon fip, I would rather err on the side of actual results than theory.

Of course again, any "average" stat out there is going to be purely dependent on how the rest of the league does offensively or defensivey(and yes both of these determine the run environment--it's not some mystical thing that just happens) . If you are a shortstop(we'll call you Tejada) your value is going to be diminished if you are playing in an era of other great shortstops. These are things people who are arguing based upon these type of stats should know. (note: for argument sake, I'm using the word average instead of replacement, because in reality war is devised based upon average and adjusted downwards to the theoretical replacement)

   27. cardsfanboy Posted: August 29, 2014 at 08:36 PM (#4782110)
I've never seen any evidence for this other than the reflexive nostalgia for the baseball of one's youth. It's also completely illogical, as in at least one league there is now a place to hide a terrible defender.


Agree 100%, but it's something I do hear frequently. When someone sees a guy like Matt Holliday who plays like a wounded gazelle chasing down a ball and then being told he's a tad below average for his position, people will say "that is because the standards have dropped". The reality is of course teams frequently have punted in left and to a degree right field and the standards have always been fairly low.
   28. PreservedFish Posted: August 29, 2014 at 08:39 PM (#4782111)
I despise war for pitchers based upon fip, I would rather err on the side of actual results than theory.


Semi-related question. Given that BBref gives credit to pitchers for BABIP under- or overperformance, does that mean that fielders on BBref have less extreme numbers than those on Fangraphs? It would stand to reason, or else BBref would be double counting.
   29. cardsfanboy Posted: August 29, 2014 at 08:44 PM (#4782113)
Semi-related question. Given that BBref gives credit to pitchers for BABIP under- or overperformance, does that mean that fielders on BBref have less extreme numbers than those on Fangraphs? It would stand to reason, or else BBref would be double counting.


Seems reasonable, but I don't know, generally bb-ref doesn't seem to have as much extreme variance on defense as fangraphs, but both systems have the occasional head scratchers.
   30. The District Attorney Posted: August 29, 2014 at 08:58 PM (#4782115)
cfb, all that you're doing is asserting that your accounting system is reality. If your runs saved by Alex Gordon are not decreased run totals being allowed by the Kansas City Royals, then WTF are they?

What if we chose to define players by batting order position, rather than fielding position? That system would probably say that Jose Reyes is a better hitter than Buster Posey, since leadoff hitters hit worse than #3 hitters. And if Mike Trout moved from hitting #2 to hitting leadoff, Reyes would get worse. And... it would in fact all "add up"! The best teams would score best, the worst teams would score worst! But it would still be nonsense in terms of how it ranked individual players.

Your response will probably be that some players could have hit leadoff, but didn't. Well, great. And some players like Eric Young Jr. and Matt Tuiasasopo also could have played more in left field, but didn't. How could you even accept "replacement player" analysis and yet not accept that?
   31. cardsfanboy Posted: August 29, 2014 at 09:13 PM (#4782122)
One thing about defensive stats, as Bill James pointed out when discussing some of the extreme great defensive seasons, it's very unlikely you could stick 3 30 run save outfielders in the outfield and save 90 runs over the course of the season. The best designed defensive systems just can't handle the elective portion of defense. So an individual might have saved 30 runs, but not all of those runs are "really" saved, just that they have been adjusted from one fielder into another fielder by the methodology used to measure defense.

Again, I'm all for debating the methodology of defensive metrics, or how to better design a system, or just making people aware of the flaws in the system etc. I just think that once you have the system set up, you are stuck with going relative to average at that position on a seasonal basis in a war like stat. And just like the Trout argument from years past, being aware of the flaws in the system is useful in an argument, but every stat out there is going to have flaws, it's just a matter of how much you think it reduces it's usefulness.
   32. cardsfanboy Posted: August 29, 2014 at 09:17 PM (#4782124)
cfb, all that you're doing is asserting that your accounting system is reality. If your runs saved by Alex Gordon are not decreased run totals being allowed by the Kansas City Royals, then WTF are they?


I do not subscribe 100% to any of the defensive metrics out there, as I have mentioned they are flawed, only thing I'm defending here, is using whatever defensive system you come up with and basing it upon average relative to that position on an annual basis.

My argument is not about asserting that the accounting system is reality, my argument is that if you are using a war methodology, you have to use 1. seasonal data for defense 2. relative to position. The rest of the defensive system doesn't matter for this discussion.

   33. PreservedFish Posted: August 29, 2014 at 09:20 PM (#4782128)
cfb, all that you're doing is asserting that your accounting system is reality. If your runs saved by Alex Gordon are not decreased run totals being allowed by the Kansas City Royals, then WTF are they?


Someone got this right above. In reality the runs saved by Alex Gordon actually show up on the scoreboard as runs scored by the Royals - because they are facing poor defensive LFers compared to previous years. The accounting is weird but it does make some sense.
   34. villageidiom Posted: August 29, 2014 at 09:21 PM (#4782129)
Jeff Sullivan is awesome and deserves a lot of credit. But this article was by Jeff Zimmerman.
   35. cardsfanboy Posted: August 29, 2014 at 09:22 PM (#4782130)
Someone got this right above. In reality the runs saved by Alex Gordon actually show up on the scoreboard as runs scored by the Royals - because they are facing poor defensive LFers compared to previous years. The accounting is weird but it does make some sense.


I like the way that is worded.
   36. BDC Posted: August 29, 2014 at 09:24 PM (#4782132)
To take a stylized example, let's say four teams are in the league, and their starting LFs are Alex Gordon and three other good fielders. The three good ones get hurt and are replaced by Kevin Reimer, Pete Incaviglia, and Nelson Cruz. The Royals allow as many runs as before. Their opponents allow a lot more. That helps the Royals win games.

Does that make Gordon a better player? Not in absolute terms, not vis-a-vis Stan Musial, but for that year, of course it does. People win Olympic medals every fourth year because their top competition got hurt. They set no records, but they win.

Now, the estimate of runs saved may be incorrect, the formula suspect, illusions unaccounted for, observations biased, or the effect much smaller than the one I'm proposing, even nugatory. But if it's correct and large enough, Gordon is exceedingly valuable.

   37. BDC Posted: August 29, 2014 at 09:26 PM (#4782133)
Edit: Coke to PreservedFish unless it would overhydrate him.
   38. The District Attorney Posted: August 29, 2014 at 09:39 PM (#4782138)
In reality the runs saved by Alex Gordon actually show up on the scoreboard as runs scored by the Royals - because they are facing poor defensive LFers compared to previous years.
And are promptly cancelled out by the runs scored Aoki used to be contributing that no longer show up on the scoreboard?

Jeff Sullivan is awesome and deserves a lot of credit. But this article was by Jeff Zimmerman.
####, awful mistake. Thank you, fixed.
   39. PreservedFish Posted: August 29, 2014 at 09:41 PM (#4782140)
Does that make Gordon a better player? Not in absolute terms, not vis-a-vis Stan Musial, but for that year, of course it does.


Allow me to quibble with this. It only makes Gordon a markedly better player if you base your entire system on positional value.

Forget stats. It's just you and me talking baseball. I would say that Gordon is not better, because I prefer the absolute view of achievement. But if you press me to consider his relative value, then I would argue that although Gordon "improves" after the injuries, all of his teammates improve by an equivalent measure. Why would I only compare him to leftfielders? WAR does it for reasons that potentially wouldn't enter into our conversation.
   40. PreservedFish Posted: August 29, 2014 at 09:51 PM (#4782144)
And are promptly cancelled out by the runs scored Aoki used to be contributing that no longer show up on the scoreboard?


Sorry, I forgot we were talking about just David Murphy switching. In this case I think you have properly identified a small absurdity.
   41. BDC Posted: August 29, 2014 at 09:52 PM (#4782145)
I agree, Fish. I think that WAR is internally consistent, but I wonder if it tends to overrate some players & underrate others because of its narrow concept of position.
   42. cardsfanboy Posted: August 29, 2014 at 10:22 PM (#4782151)
Forget stats. It's just you and me talking baseball. I would say that Gordon is not better, because I prefer the absolute view of achievement. But if you press me to consider his relative value, then I would argue that although Gordon "improves" after the injuries, all of his teammates improve by an equivalent measure. Why would I only compare him to leftfielders? WAR does it for reasons that potentially wouldn't enter into our conversation.


and


I agree, Fish. I think that WAR is internally consistent, but I wonder if it tends to overrate some players & underrate others because of its narrow concept of position.


I have my issues with war, one of the basic assumptions of war is that baseball has evolved to the point that all the position players(except catchers) have been min/maxed to the point that an average player at one position is equivalent on a per inning basis as an average player at another position (this has nothing to do with defense except in regards to positioned assigned to a player) I personally don't think that is correct, and think war makes mistakes because of that assumption (we've talked about catchers to death on that particular point, but war actually does rate catchers higher on a per inning basis than other position players on average, although it may not be enough of an adjustment) but there is no reason that a 150 games played average third base man is equally valuable as 150 games played average first baseman.

But without a ream of data, that we don't have yet, it's not going to be really possible to argue one way or another without dealing with "feels like..." type of arguments. (to me personally if I had a player play 150 games as an average player at his position, that they would be ranked 1. catcher 2. first base 3.shortstop 4.centerfield 5.right field 6. second base 7. left field 8. third base but that is "my opinion type of argument, and not really a valid argument to make except as a thought provoking discussion.)

War absolutely has it's flaws, but it's still a good system, and much better than anything we have had previously and it might be years before we get anything with as much of a consensus in the community of providing values for the arguments it's designed to help answer. But everyone participating in these discussions should have a basic feeling for the flaws and know how it might color the argument in one way or another.
   43. BDC Posted: August 29, 2014 at 10:22 PM (#4782152)
And in fact that Murphy/Aoki example shows a possible flaw in treating the outfield as three highly discrete positions.

Edit: that was an afterthought to #41, but I find #42 very insightful, fanboy. Thanks!
   44. epoc Posted: August 30, 2014 at 12:39 PM (#4782379)
This whole conversation about value w/r/t UZr is forgetting that defensive value in the WAr schema is a combination of UZr and the positional adjustment. I agree with cfb when he says that for a value stat, comparing defense to the average at a given position is the only way to go. But WAr doesn't stop there. It also applies a positional adjustment so that you can compare the value of a left fielder's defense with the value of a center fielder's defense. Whether or not Gordon is super valuable this year depends not only on the reliability of the defensive numbers (which, like, huge grain of salt there) but also on the reliability of the positional adjustment, which validity is literally impossible to tell, since it's an empirical stat based on a theoretical assumption.

If it's true that left fielders as a group are worse defenders this year, that needs to be reflected in the positional adjustments or all the defensive values in WAr are going to be effed up. But you can't change the positional adjustments based on one year of data, because the sample is way too small. So WAr sort of just throws up its hands about the whole issue, and I'm not sure I see any other reasonable course of action, other than to just severely mentally regress whatever the single-year defensive numbers are telling you.
   45. Barnaby Jones Posted: August 30, 2014 at 01:55 PM (#4782409)
Semi-related question. Given that BBref gives credit to pitchers for BABIP under- or overperformance, does that mean that fielders on BBref have less extreme numbers than those on Fangraphs? It would stand to reason, or else BBref would be double counting.


Pitching rWAR has an adjustment for team defense, so in theory there is no double counting.
   46. haggard Posted: August 30, 2014 at 06:53 PM (#4782525)
In reality the runs saved by Alex Gordon actually show up on the scoreboard as runs scored by the Royals - because they are facing poor defensive LFers compared to previous years. The accounting is weird but it does make some sense.

No. WAR isn't designed to measure actual runs saved for the team Gordon actually plays for, its an estimation of how many wins he would be worth to a theoretical team under theoretical conditions. The theoretical conditions have changed.
   47. Ron J2 Posted: September 02, 2014 at 03:25 PM (#4783917)
#20 Positional value isn't precisely a constant, but it's not a huge factor.

Repost:

I looked at players who had 300+ PAs in consecutive seasons and looked at the variation of WAR oWAR and dWAR. dWAR is the most stable of the three.

Players with 300+ PAs in consecutive seasons, 1955-2011 (4536 players). In case it's not clear, I'm comparing the same players in (say) 2011 and 2012.

Stat   Correlation  Standard deviation
WAR       .52             2.2
oWAR      .55             1.9
dWAR      .65             1.3
dWAR2     .61             1.3 



Note that there two sets of dWAR numbers. The earlier one uses TotalZone -- an adjusted range factor method since PBP based metrics were not available were not available until 2002. Year to year variation is about the same.

That 1.3 win standard deviation is sufficiently high to be skeptical of the precision it's displayed to. I mean I understand why it's displayed to tenths of wins. But basically it's really best to think in terms of letter grades.

Still they're internally consistent. Note that the offensive component of WAR is actually more volatile than the defensive part.

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Support BBTF

donate

Thanks to
BFFB
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Hot Topics

NewsblogOT: NBC.news: Valve isn’t making one gaming console, but multiple ‘Steam machines’
(1177 - 7:07pm, Nov 27)
Last: JJ1986

NewsblogOTP Politics November 2014: Mets Deny Bias in Ticket Official’s Firing
(5046 - 6:47pm, Nov 27)
Last: Rickey! trades in sheep and threats

Newsblog2015 Potential Hall of Fame Ballot | Baseball-Reference.com
(33 - 6:45pm, Nov 27)
Last: shoewizard

NewsblogSandy Alderson says Mets can move quickly if a shortstop becomes available - NY Daily News
(35 - 6:33pm, Nov 27)
Last: Lassus

NewsblogBoston Red Sox prove (once again) that competitive balance in baseball will never exist | cleveland.com
(37 - 6:31pm, Nov 27)
Last: caprules

NewsblogDave Cameron: A proposed three-way swap for Red Sox, Mariners, Nationals
(46 - 6:22pm, Nov 27)
Last: cardsfanboy

NewsblogOT: NFL/NHL thread
(8715 - 6:00pm, Nov 27)
Last: Jolly Old St. Nick Is A Jolly Old St. Crip

Newsblog[Cricketer NOT baseball player] Phil Hughes dies after “pitch” to the head
(11 - 5:56pm, Nov 27)
Last: the Hugh Jorgan returns

NewsblogOT:  Soccer (the Round, True Football), November 2014
(547 - 5:26pm, Nov 27)
Last: Ron J2

NewsblogSource: Tomas agrees to six-year deal with D-backs | MLB.com
(22 - 4:49pm, Nov 27)
Last: McCoy

NewsblogOT: Wrestling Thread November 2014
(85 - 4:43pm, Nov 27)
Last: Tubbs & Minnie Miñoso don't fear Sid Monge

NewsblogJon Lester has plenty of options in addition to Red Sox - Sports - The Boston Globe
(9 - 3:56pm, Nov 27)
Last: Digit

NewsblogNotable Players Available In The Rule 5 Draft - BaseballAmerica.com
(10 - 1:34pm, Nov 27)
Last: Dock Ellis on Acid

NewsblogWhat's cooking? MLB shares Turkey Day recipes | MLB.com
(3 - 1:20pm, Nov 27)
Last: Mike Webber

Hall of Merit2015 Hall of Merit Ballot Discussion
(56 - 11:56am, Nov 27)
Last: Patrick W

Page rendered in 0.3693 seconds
52 querie(s) executed