Alex Gordon, UZR, and Bad Left Field Defense
Since Alex Gordon moved into first place in position player WAR (although he’s now second again), quite a bit of back-and-forth discussion has occurred on if he is this season’s best position player. Most of the talk revolves around how much stock should people put into defensive statistics. Our own Dave Cameron has already taken a stab at the subject earlier in the week. Alex Gordon is getting close to two wins of value from his defense, a considerable jump from his previous seasons. After looking at the inputs used for UZR, it is not Alex Gordon’s performance going to new levels, but the lack of talented defenders in left field making him seem better.
Since Alex Gordon has moved to left field, he has put up decent defensive numbers, but not the level of 2014.
Season: UZR, UZR/150
2010: 3, 6 (partial season)
2011: 12, 12
2012: 15, 12
2013: 9, 7
2014: 24, 28
Gordon has already saved more runs than in previous seasons and is on pace to save twice as many by the end of the season… kind of. UZR has two components: what a player does and what the rest of the league does. Let’s start with Gordon’s production and then move to the league-wide values.
Outfield defense has three components, Range (ability to field balls), Arm and Errors. 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. It seems like Alex is getting to more balls in play, but the reality is that he probably isn’t.
UZR uses Baseball Info Solutions numbers, so we can look to see if Alex is actually getting to any more balls in play. They track the number of balls a fielder gets to in his Zone and those out of his Zone. In 2014, Gordon is at a career average for getting to balls in his zone (.916). The value is not too far off his previous values between .913 to .923. The gain is not here. For making plays out of his zone, I have to do a little math. I have the raw number of out-of-zone (OOZ) plays, so I will divide them by the number of innings played each year to get a rough estimate of his OOZ ability.
Year: OOZ plays, IP in LF, OOZ/IP
2010: 21, 486, 0.043
2011: 74, 1309 , 0.057
2012: 89, 1424, 0.062
2013: 110, 1364, 0.081
2014: 87, 1059, 0.082
He is getting to a few more balls out of his area, but the numbers are close to his 2013 values when he had a -2 UZR Range value. In all, Gordon has done a bit worse with balls in his zone and a bit better with balls out of his zone which makes it all pretty much a wash. Basically, Gordon is getting to just as many balls as he did in previous seasons.
Now moving on to the second component of UZR, the rest of the left fielders. UZR is set so a zero value means the player is league average in defense for his position. Since we know Gordon’s defensive values for his arm and errors aren’t causing the jump, I will concentrate on the range values. 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
It may or may not be that the average left fielder’s performance has declined; it could just be an instance of year-to-year variance we’re seeing. In every season, however, Gordon has outperformed the league-average rate. Gordon is making fewer plays in his zone this season, but compared to the rest of the league, he is quite a bit better. The same is happening with OOZ plays.
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. The baseline values for hitting stats like wRC+ can move also, but the swing won’t be as dramatic with over 240 inputs. With offensive stats, position changes don’t matter, so only players not playing changes the inputs. Even if the top four to five hitters are lost for the season, the baseline will stay fairly constant. Now consider if wRC+ is based on just those players at one position. A couple players missing time could really change the baseline, just as it does with UZR.
So what happened from 2013 to 2014. A lot of stuff. Aging. Injuries. Position movement. Here are the top-ten left fielders ranked by there 2013 left field range values. Their 2014 range values and comments on the change are added.
Name: 2013 Range Runs, 2014 Range Runs, Comments
Starling Marte: 13, 3, Missed 16 games for injury and played some CF
Andy Dirks: 10, NA, Missed entire season on the DL (back)
Carl Crawford: 9, 3, Missed 41 games because of injury
Andres Torres: 8, NA, Minors
Alfonso Soriano: 8, -2, Can’t hit and more time in RF
Yoenis Cespedes: 7, 0
David Murphy: 7, -1, Playing RF > 95% of the time
Matt Tuiasosopo: 6, NA, Minors
Chris Heisey: 5, 7, Limited action, with drop in LF talent has a 32 UZR/150 in LF
Eric Young: 5, 5, Limited playing time because of bat
Total: 78, 15
What a contrast from the previous season. Three of the top eight haven’t played in the majors this season. Only one of the ten as seen his numbers go up (Chris Heisey).
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. Even if it is harder to hit .300 now than just a few years back, we at least have an idea what a .300 AVG means and can reference it. It would be nice to have a second simple stat which doesn’t change based on the year so we can compare how a player in vacuum is performing instead of having to bring all the yearly data for the league and player and compare them. The first 500 words of this article wouldn’t be necessary if the second 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. If Alex Gordon gets any MVP love, he may want to thank more than his Royal teammates. The real help came from the rest of the league.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
Mike Morse, LF.
Dom Brown, LF.
Matt Kemp, LF
Shin-Soo Choo and Michael Choice, LF
Josh Willingham, LF
Career paths can be so odd. I remember Mike Morse making his debut for the mariners as a shortstop.
LF is one of the few remaining positions where teams are willing to punt on defense to get a big bat into the lineup. Few teams view it as a two-way position and a lot of questionable fielders (the Jonny Gomes of the world) are stowed there because the DH slot has already been filled.
Teams should get an extra DH and a designated fielder.
AL teams actually can have the pitcher hit and DH for one of their fielders, if they choose.
No, they can’t.
Huh, you’re right. I guess I got turned around with the “losing the DH” rule.
But they could have A pitcher DH, just not the one pitching in the same game.
KC Davis: I’m too lazy apparently to actually look at the DH rule in MLB, but Danny Hultzen routinely DH’d in college for Danny Hultzen, which allowed UVA to keep his bat in the lineup after they pulled him from the game.
College DH rules are different than MLB ones
This seems a pretty big flaw with WAR calculations. Especially for OF, why not simply compare range factors to ALL outfielders and remove the positional adjustment? The positional adjustment is pretty static year to year, yet as you’ve shown above, the defensive abilities of the comparative fielders fluctuate drastically. By using all OF, the ability of 3 players being removed has less of an effect on the aggregate of the group.
I guess my big issue is this – say Gordon was playing CF all this year. He had the same range scores. He would gain ~10 runs for positional adjustment. If his UZR goes down by more than that (and I suspect it would significantly), then his WAR would go down.
UZR can’t use data from balls hit to center or right field to determine how likely an outfielder playing left is to catch a ball hit to left field.
On the other do we really need to change average every year on a year to year basis rather than using a 3 year running average or something similar?
You can increase the sample size simply by averaging over a longer time.
I like the idea of a running average. I would say over 3 or 5 years.
I think if the “zones” are defined the same for all outfield positions, you could do what he is saying. Sidenote: are they the same?
Nevermind, dumb comment, I did not properly understand how UZR works.
And even if it could, positional adjustments already kinda fuck up defensive scores: A, say, -19 defending shortshop is worth more than a 0 defense first basemen for example (SS +7.5, 1B: -12.5). Likewise, a +19 defender is only worth a league average SS. Positional adjustments tell us Miggy was more valuable at 3B than 1B the years he was there compared to the years he was at 1B. It is a system that needs work.
How do you know that this isn’t correct?
Yeah, Ruki, you’re just… telling us what the positional adjustments are. How is that supposed to convince anyone who isn’t already convinced?
Well, admittedly I don’t know, and I don’t know how I’d check for it. But that would mean there is a gulf of this messed up Alex Gordon or so between an average SS and the best defensive 1B. Pretty much every Sabermetrican (I feel I spelled that wrong) agreed Miggy was hurting the team a lot more playing at 3B, but his defensive stats when including adjustment were better at 3B than his previous years at 1B. It’d mean it is probably best to turn your average 1Bs into SS even if they’re like -15. I don’t know anything, but it seems like something really off and to look at.
I wasn’t really trying to convince people, I guess, just kind of stating that I felt positional adjustments already mess with what the OP is talking about.
http://www.fangraphs.com/library/misc/war/positional-adjustment/
Here’s what fangraphs says on the subject and it seems to agree with in that it isn’t really all that accurate, but it is a good start.
We might need variable position adjustments based on the relative defensive strength of each position each year.
I remember reading an article here in the past couple of years that suggested that if Jesus Montero was easily the worst defensive catcher in baseball, he would still be more valuable as a catcher than as a 1B or DH (maybe just DH was the comparison). That really doesn’t gel for me. Occupying space at a position you can’t play should not make you more valuable than DHing. Obviously whether Montero hits is the important question as to whether he should be playing at all, but just sticking him at C even if he absolutely without a doubt cannot play it, should not increase his value. I don’t know of a statistical approach that would illustrate this, if someone has a study design that would work I’d be interested, but this fact doesn’t gel with baseball 101 to me at all.
By the logic of the positional adjustment then you should just stick Adam Dunn at C and let balls roll to the backstop to increase his WAR (actually this may not be any worse than putting Jesus Montero at C).
because miggy is an awful defensive third baseman you don’t even need to look at stats for that just watch him play
Tip: Time on page would increase like 300% per GIF.
…because people are waiting for them to load, thus becoming annoyed and less likely to visit Fangraphs again.
Interesting, I like Alex Gordon and while I realized his UZR was high I thought it was more random variation than statistical error, but this is pretty hard to ignore. Also slightly amsuing because someone brought this up in Dave Cameron’s chat a while back and he dismissed it somewhat harshly.
This definitely seems like something that should be fixed in some way.
If you’re referring to what I think you are, someone in that chat kept insisting that Trout was at a relative disadvantage in CF. Was that it? The argument was dismissed on the grounds that Trout gets more positional runs in CF. I don’t this current analysis negates that. The kind of season-to-season changes in average that Jeff is emphasizing are probably random. The same thing could happen in CF, and for all I know, maybe Trout’s defensive value in 2012 was pumped up, relative to this year, because of that.
So maybe if Trout had played in LF all year this season his overall value would be greater, even taking into account the fewer positional runs. But that isn’t a bias against CFers, as I think that commenter was arguing. It’s just the luck of the draw for this year.
IOW, it is random variation. I’m not sure what you mean be statistical error.
I think what this data indicates is that UZR is off for all LF this year. The UZR baseline should probably be 5+ runs higher than it is, meaning, that all LF’ers should lose about 5 runs of value because LF’ers as a whole are 5 runs worse at fielding this year, but the adjustment doesn’t show up in any of the positional adjustments, so eventhough the group as a whole sucks, overall LF’ers are still netting out to 0 runs on defense, according to UZR.
Having a baseline for defense seems like a good idea. Couldn’t you just sum all the plays for each position for as far back as the data go, and use that as the average UZR? Then scale it to a baseline of 100 like wRC+ and adjust the defensive value of a player based on their relative performance to the average player at their position. This would be a much easier metric for casual fans to understand, and could help clarify to a broader audience the idea that there is a difference between defensive talent and defensive performance (i.e. defense can slump). The values would change year to year for a particular player, but the baseline would remain the basically the same–it would fluctuate slightly but not by much if you continually update it and include all the defensive plays made by the entire league.
Yes, but if you do that, assuming there are seasonal changes in the quality of defense at various positions, then the values will not sum to zero. For example, if you did that with LF this year, AG would have a lower defensive value, and LFers as a whole would have a negative value.
When values don’t sum to zero, you run into other problems. For WAR to hang together consistently, certain overall sums have to be either zero, or some fixed number (e.g., I believe the sum of all WAR of all MLB players is fixed at about 1000).
It’s not only fine for LFs as a whole to have negative fielding value, they obviously should given that a replacement level outfielder could do a better job than many of the guys trotted out there.
This is one of the coolest pieces I’ve read in a while. Poses a question, and answers it very concisely in a way that no one else has really done yet.
Great work Jeff.
Thanks.
Jeff,
It is interesting to see the falloff from last years rangiest LF; but doesn’t reporting their range runs confuse the issue here? why don’t LFs like Cespedes and Marte show some benefit from the supposedly weaker competition? Marte only missed 16 games right?
shouldn’t you also look at LFs who are putting up good range numbers this year, who did not play a lot of LF last year? There could be a balancing influx of fielding talent, correct?
thanks for an interesting post. I agree that SSS effects of having few fielders at each position does explain volatility of UZR numbers to some extent.
Correct me if I’m wrong…but doesn’t the league average baseline always matter in our calculations? If you’re far ahead of the baseline – whatever that is – you get credit for it. In other words, if no one hits .300, then someone who does so becomes more valuable and gets more credit for it. It is true that fewer players can affect the league average baseline for left-fielders versus hitters (thus adding variance), but that seems like something we have to live with in order to properly credit someone who’s (seemingly) ahead of the pack.
Yes, this is my understanding. You could say it all begins with the fact that the total WAR for any season has to be fixed, at 1000 now, I think. So you can’t have a situation where everyone is somewhat worse than everyone was in some past year. If the overall talent level goes down, players who maintain their talent level get extra credit for this, since the loss of WAR by some players has to be compensated by others.
MGL can verify this, but the “average left fielder” is not determined using a one year sample, I believe. I believe it uses six years of data.
This pertains to the value of the play, FYI. Wrote that horribly.
These are MGL’s words from the primer:
“One of the differences between UZR and linear weights is that with UZR the amount of credit that the fielder receives on each play, positive (if he makes an out) or negative (if he allows a hit or an ROE), depends on how often that particular kind of batted ball, in terms of its location, speed and several other factors, is fielded by an average fielder at the same position, measured over a time span of several years, in addition to whether the batted was a hit, out, or error (or FC).”
Here’s a little more from the primer for anyone interested:
“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.”
Yes, so it’s like hitting in that respect, and as I said in other posts, I think it has to be for total WAR to be fixed and the same every year.
But here’s one possible problem: If average hitting goes down, as it has in recent years, a given BA, wOBA or whatever is worth more, because it’s valued vs. average. But this can be rationalized by arguing that pitching is better, for whatever reason, so it takes more talent to achieve a certain batting proficiency in that environment.
But when it comes to fielding, that argument is not so obvious, to me at least. Why should it be any more difficult for fielders from year to year reach a ball in the same place or bucket. A batter goes one on one with a pitcher, so as pitchers get better, batters get worse. But a fielder goes one on one with a ball hit to some spot. That doesn’t change from year to year.
I guess you could argue over time, arms get a little stronger and legs a little faster. But these are very gradual changes over a long period of time, and should only go in one direction. This is very different from hitting vs. pitching, which can fluctuate from year to year, and in either direction.
Other factors that could change the difficulty of defensive plays include quality of equipment–bats, balls, and gloves–maintenance, surface type, and special features (e.g. Tal’s Hill) of parks in the league, leaguewide baserunning and hitting ability and strategy, and leaguewide pitching ability. Those last two categories require a bit of unpacking: if, for reasons we don’t entirely understand, pitching has become increasingly dominant, part of that advantage may manifest in a lower quality of contact. Fewer line drives, more soft vs. hard fly balls, etc. Moreover, tactics could change in response–teams could start favoring speedy groundball hitters over lumbering sluggers who are especially vulnerable to the rise in strikeouts, changing the degree of difficulty of corner infielders’ jobs.
Not all of these factors come into play every year, e.g., only rarely is a new special feature added to a ballpark. But when you start thinking about the multitude of factors that could influence how difficult it is to field a batted ball, you realize there’s no reason to expect that difficulty to be the same from season to season.
How do they determine if a ball is in the zone or not? Is it a fixed area regardless of the starting position of the fielder and the type of ball hit? Or do they use those factors for the determination?
UZR doesn’t care whether the player was in the bucket when the pitch was released or had to cover 90 feet while running faster than a frightened deer to get to the bucket.
Yeesh
One could use this to argue that there needs to be a larger positional adjustment between CFs and LFs than normal.
This is great- it helps us see clearly why the “it takes 3 years for Def stats to stabilize” is the mantra.
I’m confused as to why some are commenting that this needs to be fixed. Seems like a WAR should take into account the relative value of the defense, against an average, much like the offensive components do.
Right. Nobody thinks we need to adjust Pedro’s 1999 WAR down because pitching was worse in 1999. Or at least, they shouldn’t.
That’s because we’re familiar with a concept called “sample sizes,” not because we like to draw false equivalencies between UZR and literally any other stat used to measure value, by claiming UZR to be even close to their league in terms of reliability.
This is what I was thinking. If a fielder is an average fielder one year and maintains his performance the next while the league’s average performance gets worse, he has become above average at his position, and should be recognized as such. I just don’t see why this is an issue.
It just *seems* wrong in this case because Gordon has not played any better.
If one guy hits .300 in both 1968 and 1969, then I think everyone agrees he deserves different credit for each year, because each year represents an authentically different achievement. He was actually better in one of those years.
But in this case we are assuming that the average LF defense numbers haven’t dropped because of any large trends regarding the environment. They’ve dropped, and Gordon has been lifted, because of whatever quirky things are going on in the rest of the league at that position.
If you look at it from a team-building perspective, nothing happened in the winter of 2014 to make it more difficult for teams to find competent LFers. Gordon is not more valuable than he was previously.
To put it really simply, if 17 managers went insane and started their backup catchers at shortstop, it wouldn’t make the remaining starters better in any real way. But their defensive numbers would skyrocket.
It seems to me that hitting .300 in both 1968 and 1969 is a perfect example of exactly the same thing that happened with Gordon. The guy had the same average, the hitting environment changed. He was better relative to the league. Gordon is making almost exactly the same number of plays as he did last year but the fielding environment changed. He is better relative to the league.
But he’s not better relative to the league. He’s better relative to 1/9 of the league. When you’re compared to everyone in the game who swung a bat for an entire season, that’s one thing. When you’re just compared to other left fielders, who have a defensive rating that comes from making or not making a relatively small number of plays, that’s not exactly similar.
Because the theoretical “replacement level” has not changed. It’s not as if all capable defensive outfielders died off, leaving no competent alternatives. Teams are consciously choosing to sacrifice defense in favor of offense.
It could be a concern of reliability. Every year, over 500 Pitchers will take the mound, logging some 22,000 Innings of work, each one being involved in every play. There just aren’t that many sinking flies to Bucket 23 (or whatever they’re labeled) in Left Field and only a couple dozen guys logging significant time there.
WAR is considering the performance of a theoretical replacement player whose numbers are derived from evaluation of actual events. If we don’t have enough instances of a given event, we lack confidence in saying “Replacement LF would have gotten to that.” Hence, the multi-season examination mentioned above.
This is interesting, but also overstates the problem. The upgrade from Morse/Ibanez to Ackley in LF for Seattle is substantial, as one counter example. Also, some level of regression to the mean should be expected for any list of top performers at any aspect of the game….
For Oliver I use a constant baseline of data from Gameday, starting in 2005 (grows with each season), and have never ranked Gordon below 2nd best in LF
2011 2nd +22
2012 2nd +14
2013 1st +15
2014 1st +18
However, I only rated him as the 8th best position player this year (3.8 WAR) due to differences in calculating offensive value. I have Gordon +18 batting, +1 stealing, 0 base running and +18 on defense for a total of +38.
And what about Gerardo Parra, going from a +26 Def to his current -8 Def? It’s just begging for a gif-laden Sullivan article.
my numbers have a smaller spread than UZR, and I believe better YTY consistency, but some guys can have bad years.
I have Parra’s otufield totals (all 3 positions) each of the last 6 years as -2, +12, +18, +2, +11, -3
I also have Starling Marte -8 this year after averaging +16/150 in his career
So are the CFers this year maybe better than they were last year and the year before? Mike Trout wants to know.
down a little, MLB CF catch rate from .375 to .369 to .367 last 3 years, from high of .403 in 2006.
Especially considering that UZR has him as being almost as bad as Derek Jeter defensively, despite his being well above the average CF per Inside Edge.
way to get a jeter reference in there, douche
I would like to see an estimate of plus/minus uncertainty on these numbers. Saying Gordon got much better or everyone else just happened to get worse at the same time makes me skeptical either way.
The positional adjustments are calculated based on players who have played multiple positions in the same approximate time span. So to that extent they are exactly correct. They are designed to answer the question, for example, “If an average fielding LF plays CF, what would we expect his defensive talent/performance to be in CF (as compared to other CF rather than as compared to other LF)?” The answer to that question is the positional adjustment difference between CF and LF.
Now, because each position gets a different number of opps per season, there is a certain threshold below which a player is better off playing the “easier” position and another threshold above which the player is better off playing the “harder” position. If that were not the case, then it would not matter where a player played. He would have exactly the same defensive value once you include the positional adjustments.
For example, Miggy, because he is around an average 1B, defensively, his value at 3B has to be worse. In order for it to be better, he would have to be an exceptional first baseman and a near average 3B.
Now, all that being said, every player obviously has their own unique translations from one position to another for various reasons, so the positional adjustments are a one-size-fits all construct which technically does not apply equally to each and every player.
In addition to that, we are not even sure of the exact values of the positional adjustments for several reasons. That is especially true when we compare infield to outfield and when we include 1B and catcher.
So we live with what we have even though it is not nearly perfect.
Ok it has always seemed odd to me that you have fixed positional adjustments by position for offense. I hate that. Suppose some manager wins the world series with an inept defensive CF who hits 60 homers and that changes the perspective of the league about offense in CF. Then your positional adjustments are now off.
The best system, I feel, is to use a 3 year or maybe a 5 year moving average for offense and defense at each position. That gives you an adjustable baseline that has many more observations and allows your positional adjustments to change with the times.
The positional adjustments are based on defensive ability. They’re calculated based on the actually performance of play multiple positions. They are applied to offense only in the sense that everything is added with offense for player WAR
Right, positional adjustments are NOT based on offensive output at each position, for the reason that you give as an example.Just because in any one year or two a particular position has a high or low offensive output, does not mean that the positional adjustments should change. And they don’t. They are fixed. Forever. Unless someone does a better job estimating them. As I said, they are based on the question, for example, “What would an average CF’er do in LF (or at 2B or whereever). Or as Tango likes to say, what would the defensive value of Bloomquist or Zobrist be at each position?
It seems to me that this highlights a big problem with comparing players to their position groups. It seems that it makes WAR really misleading. So say that both Gordon and Mike Trout are both +5WAR players. If it stands true that the overall quality of LFers is worse than the quality of CFers, then wouldn’t that make Mike Trout more valuable because the replacement level he is going up against is higher? That is to say, Trout is starting at a higher baseline than Gordon.
Put another way: say over the course of 50 PAs a pitcher has an OPS of .700 and a 1B has an OPS of .950. WAR is going to imply that that the pitcher’s offense has been more valuable to his team, maybe doubly so. But that simply wouldn’t be true. Maybe “more valuable than the expected contribution from his relative position group”, but in absolute terms the 1B was easily more valuable.
If the resplendent level for a group is very high, performing slightly better than that level is not worth especially much because it is easy to replace that production by definition. Value and therefore WAR is directly related to scarcity, basically because of supply and demand. If a skillset is common it is not worth much. If you don’t think that this is the right way to evaluate a player, than just choose a different metric.
Sorry replacement level not resplendent. Autocorrect…
If you truly believe that, then do you also have a problem with defensive and offensive runs being equally weighted for WAR? Seeing that an out is a much more common occurrence than a non-out. Outs happen what 60-80% of the time, so following the scarcity argument, wouldn’t an offensive run be more valuable since hits occur way less often than outs?
Every model is imperfect. WAR is so much better than everything that came before it (especially win shares!). And yet, all we end up doing in comment sections is griping that the model isn’t a perfect match to reality.
On an unrelated note, does the shift effect any of the defensive numbers? If three infielders are between 1st and 2nd on many more batters, doesn’t the left field end up trying to run down more balls (especially line drives)? Does that cross up the model at all?
Really interesting. The main thing I take away from this is that WAR is appropriate for what it is, and not for what it isn’t. That is, this does outline Gordon’s value relative to a replacement level leftfielder in a given year. But for people who are looking at the MVP candidates, then you have to ask yourself, am I looking for someone’s true talent? Or am I looking at how they’re doing this year relative to the whole pool of players? It’s interesting that people (implicitly) don’t want Gordon for MVP because they want to look at his true talent, but by definition the MVP is based on who’s doing really well THIS year. It’s like the point that Dave has made about the All-Star game. Are we electing people who started off the season well, or are we electing the best players year in and year out?
Of course MVP should be based on actual production. Exactly what production is not that clear though which can be illustrated with a few examples. If an offensive player has a really high WAR because his context neutral batting is great but relatively few of his hits create runs and wins, should he be the MVP? If a pitcher has a great FIP (and we use fWAR for him), but his ERA and a W/L record is terrible, is he an MVP or Cy Young candidate?
I do agree that RE24 should be considered more heavily than wRAA for the purposes of MVP voting.
MVP is about helping your team win games, not about having the most outstanding context-neutral stats.
(Trout, of course, leads the AL by a full win in RE24 this year, further cementing his case.)
I would also agree that RA9 is a far more important consideration for Cy Young and MVP than FIP, because context-independent performance doesn’t help your team win if it doesn’t stop the other team from scoring.
It’s why Ricky Nolasco isn’t “undervalued.” He just sucks at preventing the other team from scoring runs.