WAR without the defense (oWAR) & WAR with just the defense (dWAR)
Posted by Sean Forman on September 14, 2010
Joe Posnanski's great post on Wins Above Replacement brought to the forefront for me an idea I've been kicking around now for a few months.
One of the big complaints about WAR that holdouts have is that they don't trust the defensive numbers. It is a valid concern. As is always pointed out, the defensive metrics can be a bit, umm, inconsistent in how they rate various players. This can cause players', like Josh Hamilton's for example, WAR total to range from 4th best in the AL to far and away the best depending on what the defensive measure says about the player. I think that FieldFX, should it become publicly available, could calm many of these concerns, but we aren't there yet.
So if you are one of the group of people who like the idea of WAR, but think the defensive numbers are garbage (for the record, I'm not one of those people), we are now presenting another output in the Player Value Tables and League Leader tables, oWAR or offensive-only wins above replacement. I hate to foist another acronym on folks, but I think it is a common-sense compromise for the large contingent leery of the accuracy of current defensive numbers.
oWAR assumes that every player is an average defender and therefore the stat utilizes the player's contributions from baserunning and batting, the positional adjustment (to account for the difference in batting stats at each position) and the replacement runs factor that counts playing time the player kept from going to a replacement level player.
I've also added a column for what we calculate the player's defensive performance is worth (dWAR). Unlike with batting, it's generally believed that the replacement level fielder is around league average. The idea is that there are a lot more guys who can field in the major leagues than can hit in the major leagues. There our dWAR value is just the player's total zone defensive runs saved divided by the runs to win conversion (generally ten runs to a win).
You can add the two together (oWAR + dWAR) and get our existing WAR values (which is still there and still our number of choice) or you can use somebody else's defensive measure you prefer (like the Fielding Bible's Defensive Runs Saved, which appear in our fielding stats), divide that defensive runs measure by ten and then add it to our oWAR for your very own WAR.
Perhaps you could call it myWAR.
A couple of lists:
*Career, Active, Single-Season and Year-by-Year Offensive Wins Above Replacement Leaders
*Career, Active, Single-Season and Year-by-Year Defensive Wins Above Replacement Leaders
For those who are very skeptical of our defensive metrics, I think our career and active defensive leaders pass the sniff test. Career: Brooks Robinson, Andruw Jones, Roberto Clemente, Ozzie Smith, Mark Belanger, Barry Bonds, Carl Yastrzemski, Germany Smith, Willie Mays and Cal Ripken. Active top 5: Andruw Jones, Ivan Rodriguez, Scott Rolen, Omar Vizquel, and Ichiro Suzuki.
As an example of how the numbers are presented for current seasons, here are the top 20 position players in MLB by overall WAR. Taken from our 2010 MLB Batting Value Page
| Rk | Tm | WAR ▾ | oWAR | dWAR | |
|---|---|---|---|---|---|
| 1 | Evan Longoria | TBR | 6.7 | 5.4 | 1.3 |
| 2 | Miguel Cabrera | DET | 6.5 | 7.0 | -0.5 |
| 3 | Adrian Gonzalez* | SDP | 6.5 | 5.5 | 1.0 |
| 4 | Robinson Cano* | NYY | 6.1 | 6.0 | 0.1 |
| 5 | Josh Hamilton* | TEX | 6.0 | 6.7 | -0.7 |
| 6 | Adrian Beltre | BOS | 5.8 | 5.2 | 0.6 |
| 7 | Shin-Soo Choo* | CLE | 5.7 | 4.0 | 1.7 |
| 8 | Albert Pujols | STL | 5.7 | 6.0 | -0.3 |
| 9 | Justin Morneau* | MIN | 5.4 | 4.3 | 1.1 |
| 10 | Troy Tulowitzki | COL | 5.4 | 4.2 | 1.2 |
| 11 | Jose Bautista | TOR | 5.3 | 6.8 | -1.5 |
| 12 | Joe Mauer* | MIN | 5.3 | 4.9 | 0.4 |
| 13 | Aubrey Huff* | SFG | 5.2 | 3.8 | 1.4 |
| 14 | Joey Votto* | CIN | 5.2 | 5.9 | -0.7 |
| 15 | Paul Konerko | CHW | 5.0 | 5.3 | -0.3 |
| 16 | Ryan Zimmerman | WSN | 5.0 | 4.8 | 0.2 |
| 17 | Brian McCann* | ATL | 4.7 | 4.2 | 0.5 |
| 18 | Carlos Gonzalez* | COL | 4.6 | 4.7 | -0.1 |
| 19 | Angel Pagan# | NYM | 4.5 | 2.3 | 2.2 |
| 20 | Mark Teixeira# | NYY | 4.5 | 4.1 | 0.4 |
On another note, I know that some of the issue here is a lack of clarity as to how these numbers are put together. We don't have an in depth look at where the numbers come from. I'm going to work on an indepth and detailed look at our WAR methodology in the coming months and hope to have an A-Z rundown suitable for non-sabermetricians during the offseason. I think you'll be impressed at how careful and detailed the methodology is. Nothing is slapdash about it and each step follows logically from the previous.
All WAR methodology is provided by Sean Smith of Baseball Projection.com.
September 14th, 2010 at 9:45 am
Good stuff. There have certainly been a lot of blog comments from people who seem openminded about WAR, but just don't understand how the numbers are conceived. I think a more in-depth explanation will convince some of them it's worth looking at.
September 14th, 2010 at 10:11 am
Yay! I'm one of those people who's quite skeptical of defensive metrics (or at least of putting too much faith in them), so this is wonderful! Obviously, linear weights has a lot of evidence to support it, so I can definitely get behind this. Also, it's allowed me to try to better understand my Brewers: 3rd best offense in baseball, 4th worst defense, and pitching so bad that we're the only below-replacement pitching staff in all of baseball. Yikes.
Anyway, thanks for the innovation! Now if we could only get WAR/600 PA and WAR/200 IP, then I'd really be in heaven...
Thanks again!
September 14th, 2010 at 10:29 am
Great line from Posnanski:
"It’s amazing how irrational people get when it comes to making arguments against things they don’t like."
Yes indeed. Of course, it's a crime no one ever thinks they're guilty of.
September 14th, 2010 at 10:44 am
The same could be said about people making arguments about things they do like.
September 14th, 2010 at 10:57 am
Great stuff; defensive metrics are very iffy at best, but WAR is a cool concept.
September 14th, 2010 at 11:04 am
Sean,
If I am interpreting the linked leader boards correctly what you are presenting as dWAR is just the defensive wins above average for that player at his position. For example, Ozzie Smith's 21.6 defensive wins as a shortstop are being compared to Barry Bond's 20.4 defensive wins as an outfielder. (Mostly left field, he played some center early in his career.) Which means that you are not including the position adjustment as part of defense. Which means you are including it in oWAR. I'm not sure that this is wrong, but to paraphrase Bill James quoting someone else, "Ozzie Smith does not bat as a shortstop. He does not take his fielder's glove to the plate. He takes a bat." Much of Ozzie Smith's value as a defensive player comes from his position. A slightly below average shortstop, (say, -5 runs a year), is worth far more to his team than the greatest defensive left fielder ever. (Bonds) So I think the position adjustment, Rpos should be included in dWAR, not oWAR. I am open to persuation, but this is where I am now.
September 14th, 2010 at 11:06 am
I like it, but I strongly suggest moving the position adjustment into the DWAR part. An average fielding shortstop simply has more value than an average fielding 1st baseman - and that value is defensive value.
September 14th, 2010 at 11:06 am
Most people agree that partial year defensive data for both UZR and BIS is not reliable.
Since WAR uses that unreliable data, then how can WAR be used and trusted?
IMO it can't.
So I applaud the new stats. Thanks.
September 14th, 2010 at 11:06 am
Thankyou, Sean, you must have been reading some of our minds of late.
I'm not a believer in WAR, and the defensive metrics are a large factor in why. I've always wanted a seperate offense/defense list to see exactly who the stat over/underrates and where.
I'm not saying this won't in itself swing me over to the "believer" side of the fence, but it's a start.
Thanks
September 14th, 2010 at 11:13 am
If I am interpreting the linked leader boards correctly what you are presenting as dWAR is just the defensive wins above average for that player at his position. For example, Ozzie Smith's 21.6 defensive wins as a shortstop are being compared to Barry Bond's 20.4 defensive wins as an outfielder. (Mostly left field, he played some center early in his career.) Which means that you are not including the position adjustment as part of defense. Which means you are including it in oWAR. I'm not sure that this is wrong, but to paraphrase Bill James quoting someone else, "Ozzie Smith does not bat as a shortstop. He does not take his fielder's glove to the plate. He takes a bat." Much of Ozzie Smith's value as a defensive player comes from his position. A slightly below average shortstop, (say, -5 runs a year), is worth far more to his team than the greatest defensive left fielder ever. (Bonds) So I think the position adjustment, Rpos should be included in dWAR, not oWAR. I am open to persuation, but this is where I am now.
Kds,
I would disagree. Smith's value on defense is how much better he plays defense than the player the Cardinals could find to replace him. In this case an average defensive shortstop (lots of those in AAA).
On offense, James is right that he doesn't bat shortstop, but there is a definite effect on the Cardinals' offense that Smith bats shortstop.
Think about Kevin Youkilis.
Is the Red Sox offense overall better with
1) a replacement player at first and Youkilis playing third, or with
2) a replacement player at third and Youkilis playing first base?
The two replacement players (in theory) should have the same overall value to the team, but the Sox score more runs in situation 1 (assuming Youk bats the same in each case) than in situation 2. Youk playing third adds runs to the Red Sox offense.
Albert Pujols playing shortstop adds a lot of offensive value to the Cardinals office, you replace Skip Shumaker's offense with a new first baseman's, but Albert will likely give back those runs by being a well below average defender (though in his case, he might have been able be a better than -20 fielder).
September 14th, 2010 at 11:14 am
Great idea! This cleans up some of the numbers I think and breaks them down a bit.
September 14th, 2010 at 1:14 pm
Sean,
Thanks for the quick and thoughtful answer! I am still not convinced.
The way you are doing it both of the comparisons of oWAR and dWAR are against other players of the same position only. This is canceled out when we add them together to get WAR.
I think the correct way to do this is to have oWAR compare the batter to all other players as batters only. And get the total defensive value by including the position adjustment there. I don't like arguments from "how it feels", but doesn't it feel wrong to have Ozzie as a much more valuable defender than Barry? Add the Rpos and you get about 36 vs 8. I think most analysts would agree that that is a much better reflection of their defensive value than 21.6 to 20.4.
Youkilis will (presumably) hit the same at either position, his batting won't be any more valuable. His overall play will be more valuable because of his position. The replacement 1st baseman will (presumably) hit better than a replacement 3rd sacker, so the team offense will be greater. You would assign that increased offense of Youk, who hasn't changed his offense at all, rather than to the better hitting of the replacement at 1st compared to 3rd. I'm surer of my position, (with adjustment), than I was before.
September 14th, 2010 at 1:22 pm
I like the breakdown between Offensive WAR vs Defensive WAR, as I think its an excellent idea.
One thing I've noticed so far...Derek Jeter goes from 55th best alltime (position players) in WAR to 25th best alltime in Offensive WAR (oWAR). Its not a surprise that he moves up via oWAR but I was surprised at HOW MUCH he moved up and how highly he is ranked via that metric.
September 14th, 2010 at 1:26 pm
I agree, my initial thought was that the positional adjustments should be included in dWAR. That's really your total defensive contribution -- being able to play a particular position (adjustment), and how well you play it (TotalZone).
September 14th, 2010 at 1:55 pm
"For those who are very skeptical of our defensive metrics, I think our career and active defensive leaders pass the sniff test."
Perhaps. But for current players you might want to take another sniff, Sean.
How do you explain Ryan Zimmerman ? One of the best defensive players in the game, and arguably the best at 3B. UZR, Dewan, Fielding Bible panelists, fans and coaches scouting reports all agree that he's a top-tier defender.
Your dWAR disagrees, calls him basically replacement level with the glove.
For 2010, UZR rates him at #1 among 3B and #3 overall. DRS ? The same.
Your dWAR ? #10 among 3B and 138 overall.
For his career, UZR has him at 59. DRS, 79. Your Rtot ? 2. That's not a typo. TWO.
Either you've doscovered that the other metrics (and subjective opinions) are wrong and Zimmerman is truly replacement level defensively, or there's a serious flaw in the way you come up with your defense #'s.
I tend towards the latter.
September 14th, 2010 at 2:11 pm
Either you've doscovered that the other metrics (and subjective opinions) are wrong and Zimmerman is truly replacement level defensively, or there's a serious flaw in the way you come up with your defense #'s.
We don't have him as replacement level, we have him as average. 0 runs is average. I would have to delve more deeply into the numbers to see why that is the case for Zimmerman. Perhaps there is one area that he does really well that TZR isn't picking up.
September 14th, 2010 at 2:20 pm
Youkilis will (presumably) hit the same at either position, his batting won't be any more valuable. His overall play will be more valuable because of his position. The replacement 1st baseman will (presumably) hit better than a replacement 3rd sacker, so the team offense will be greater. You would assign that increased offense of Youk, who hasn't changed his offense at all, rather than to the better hitting of the replacement at 1st compared to 3rd. I'm surer of my position, (with adjustment), than I was before.
I can see the case you are making, but I have to go back to the statement that the replacement defender is a league average one. Ozzie's replacement in the field is not Keith Hernandez or Lonnie Smith, it is Rafael Belliard or Rey Ordonez (or guys somewhat more average than them in the field). There are lots of guys who can pick it at a league average level, but the scarcity comes when looking for a guy who can field shortstop and then hit as well.
I think a very strong argument can be made for putting the positional adjustment on the offensive side. The Red Sox score more runs with Youkilis at third base than they do with Youkilis at first. That directly improves their offense.
I'm willing to be convinced I'm wrong, but I think the way I've stated it is at worst defensible and perhaps more importantly it is the more informative of the two possible formulations.
September 14th, 2010 at 2:44 pm
"We don't have him as replacement level, we have him as average. 0 runs is average. I would have to delve more deeply into the numbers to see why that is the case for Zimmerman. Perhaps there is one area that he does really well that TZR isn't picking up."
By your own words, replacement level ~= average for fielding: "it's generally believed that the replacement level fielder is around league average".
Whether you call it replacement level or average, your defensive #'s for Zimmerman come up wildly different than everyone else's. You've gotta admit, that smells fishy.
There is indeed one thing he does really well that TZR isn't picking up - which is turning balls hit anywhere near him into outs. But that's pretty basic and any defensive metric should pick that up.
If you do delve more deeply into the #'s, I'd be interested to see what you come up with. So would many ohers, I'm sure.
September 14th, 2010 at 2:46 pm
Thanks for this, Sean. Personally I don't like the defensive contribution to WAR at present (but I don't throw WAR out the window either), and am confident that fielding metrics will be much better in the near future and turn the saber-haters into a more obvious Flat Earth Society than they are already.
Anyway, I'm interested to see how the position adjustment discussion plays out. On the one hand it seems counter-intuitive that dWAR only ranks players against average fielders and loses the positional adjustment. It does seem weird that Smith and Bonds are so close.
On the other hand, what you said makes perfect sense. The whole concept of WAR is based upon the replacement player, and if you separate the two offensive/other contributions and the defensive contributions, you are probably right to assert that in trying to replace Ozzie's defense he should be compared to an average fielder, because that is exactly what a team could get in short order to replace him.
However, the more I think about it, the more I don't like the positional adjustment falling under oWAR. Personally, I think I'd prefer if it was omitted entirely from both of the new stats. It somehow seems wrong falling under either of them.
I need more convincing.
September 14th, 2010 at 2:52 pm
Is it just me, or is oWAR and dWAR not showing up for every player.
September 14th, 2010 at 3:25 pm
Sean has the right idea in including the positional adjustment in oWAR. If you just want to compare two players' offensive value regardless of the position they play, well, you can already do that by simply comparing their batting wins. The whole point of oWAR is to describe a player's offensive value relative to the position he plays. If the positional adjustment weren't included in oWAR, there wouldn't be a point in having oWAR to begin with. It would be redundant, since we already have the numbers for doing a position-blind comparison of two hitters.
September 14th, 2010 at 3:26 pm
It somehow seems wrong falling under either of them.
To be honest, my main goal here was to remove the offending (to some) defensive stats from WAR and give a defense neutral WAR number, similar to VORP if you are familiar with that.
It is only on there for player who are active. I am planning a big javascript update in the next week or two and will rerun everyone then.
September 14th, 2010 at 3:31 pm
I could see the positional adjustment fitting with either oWAR or dWAR, really, depending on how you look at it.
It feels as though it should go more with defense at first, but ultimately I have to agree with Sean that it's more logical to classify it under oWAR.
The reason for this is that the numbers for rPos are based on the measurable differences between the typical offense for players at each position. While the true difference between the players is on the defensive side, unless I'm mistaken, the positional bonus or penalty is determined based on offensive differences.
The other thing is that while it feels weird to have Bonds listed as being about equal to Ozzie in dWAR, it would feel equally weird to imply that Keith Hernandez cost his teams games defensively.
The last thing is that this breakdown strips out the most contested aspect of WAR to make it easier to adjust your thinking. If you prefer UZR or Plus/Minus, or if you prefer to include a subjective adjustment to the defensive aspect, or some homebrew system that averages a bunch of metrics, all you really need to do is throw away the dWAR value and divide your metric by 10 and slap it to the oWAR. With most defensive metrics, it would mean one more layer of math if you included the positional modifier with the defense. It wouldn't be a big deal either way, but it's a bit more practical the way Sean has it for people who want to tinker, and at any rate, there's nothing stopping anyone from looking at the components and organizing the data in a way that's more intuitive to them.
September 14th, 2010 at 4:23 pm
Feels weird to me, but maybe with some sleep it'll seem less weird. I get the logic though, and don't contest it.
I really like the idea of being able to easily access oWAR. I don't think there is much criticism to be had toward batting or pitching metrics.
Sean, I know the site is stat-oriented, and that it is very good at that; this is the first site I go to for stats. However, when are you gonna implement some more community/social features? I think the volume of discussion on the blog shows that people are very willing to discuss things right here instead of elsewhere. I'm not going to ask you to set up an IRCd or anything (though one can hope), but do you have any plans for forums?
September 14th, 2010 at 4:36 pm
I'm withholding judgement until I see how this affects Fred McGriff. I don't understand why no one has brought him up yet.
September 14th, 2010 at 5:41 pm
I think Beavis and Butthead want to know if there is going to be a gWAR.
September 14th, 2010 at 5:44 pm
"Albert Pujols playing shortstop adds a lot of offensive value to the Cardinals office, you replace Skip Shumaker's offense with a new first baseman's, but Albert will likely give back those runs by being a well below average defender (though in his case, he might have been able be a better than -20 fielder)."
Is this really the case though? The career totals indicate plays potentially accruing 150 oWAR and only 25 dWAR? Wouldn't this imply that the new focus on defense is perhaps a bit much? Yes, defense matters, but still FAR less than offense, because the defensive variance is, generally speaking, a lot smaller. Shouldn't teams just find the best hitters possible who are evenly mildly competent at a position and go with that?
Maybe I'm missing something, but I'm struggling to understand the immense gap that exists between oWAR and dWAR numbers. Unless, of course, defense is generally overstated. Which *MIGHT* explain the struggles of defensively-oriented teams like the Mariners.
September 14th, 2010 at 6:47 pm
BSK I don't know what you mean by overstated, but elite offensive performers cannot be matched by elite defensive performers. However, many offensive performers whom are lauded as big-time hitters (but whom are a bit weak defensively) by analysts who rely on counting stats and simple stats can often be matched in overall WAR by relatively meek hitters who are defensive studs. But when you're talking about the Albert Pujols types, their dWAR is going to pale in comparison to their oWAR. For the majority of MLB players, though, having a great defensive season makes them immensely more valuable, because they might be able to put up 1.5 or 2 WAR on defense but not be able to do more than 2 WAR in more than a couple of seasons in a relatively long career.
A team that had available on the free agent market, through trades and their own system an absolutely stellar cast of fielders, coupled with good pitching would be very competitive even if they had a below-average offense.
But yeah, if your pitching is decent, especially if they aren't too bad at missing bats, you could just trot out Prince Fielder, Dan Uggla, A-Rod, Jeter, Manram, Dunn, etc. and call it a day. That these guys are net negatives in the field doesn't take away from the fact that they can all put up 4 5 6 7 8 WAR seasons without a problem, even if their defense is -1 or -2 WAR.
September 14th, 2010 at 7:31 pm
Fireworks-
My point is that the most career dWar is in the 20s and the most career oWar is in the 150s. Clearly, it is easier to generate oWar. So, I guess technically speaking, 1 dWar is harder to get than 1 oWar, but since every WAR is equal, just get what you can. I'm sure it's a lot easier to find guys with oWar of 6+ and dWar of -2 than guys with the inverse.
September 14th, 2010 at 8:24 pm
Sean,
The value in Mike Piazza playing catcher had nothing to do with the quality (or lackthereof) of his defense.
It was about the difference in offensive production between Eric Karros and a typical hitting catcher.
September 14th, 2010 at 8:29 pm
BSK,
You are including both the replacement runs and the position runs in the oWAR values. True apples to apples would compare dWAR to Batting Wins.
September 14th, 2010 at 8:34 pm
@29
A good chunk of oWAR for non-star players is just "rRep" which you just get for showing up. Its ostensibly a durability bonus. Take that out of oWAR and there is a bit more balance with dWAR. Star hitters will still rack up huge amounts of oWAR though.
September 14th, 2010 at 9:04 pm
Sean-
Thanks. I knew I was probably missing something.
September 14th, 2010 at 11:32 pm
BSK's point about offensive vs. defensive WAR and the magnitudes of the totals is rooted in the fact that there are only so many runs (and WAR by extension) that can be saved by a defender, but the number of runs that a great offensive player can add is theoretically limitless.
Case in point, Pujols is having a down year by his standards, but nevertheless is still a very strong hitter. The stat Runs Created per Game (RC/G - available under the more offensive stats tab on the player page) estimates that a team of 9 players of Pujols' hitting performance this year would score 8.1 per 9 innings played (this is his lowest number since 2002 and he was exceed 10 and 11 several times during his career). The NL average runs per game per team is 4.36. There is some distortion here because not all games are exactly 9 innings, but roughly speaking a team full of Pujolses would double offensive output of the average NL team. On the concept of a run scored is equal to a run saved and fielder would have to be of such great ability that 9 players fielding at his level would eliminate just about every run scored by the opponents to equal what Pujols has done offensively in a poor season for him. Of course it is impossible for a fielder to do this. In reality no fielders are good enough to get to every single baseball, and even in a theoretical situation where you had nine defensive players that could record outs on every ball in play, the offense would still score runs on walks and home runs.
So simply put the best offensive players in the game are contributing more than even the theoretically greatest defensive player could.
I don't think this gets directly to team construction in that a balance needs to be crafted based upon available players, financial resources and the abilities of other defensive players (e.g. a great defensive CF might make up for some of the shortcomings of a poor LF, RF and even SS/2B who ar bad at going back on popups). Besides this, players who can make an offensive impact comparable to Pujols are exceedingly rare.
September 15th, 2010 at 9:54 am
Is it just me, or has oWAR disappeared from the player value tables as of this morning? I was going to look at it, but couldn't find it.
September 15th, 2010 at 9:59 am
Aargh, not sure how that happened. It will be back shortly.
September 15th, 2010 at 1:30 pm
Sean, I'm pleased and amused to see you finally returning to the formulation that was used twenty years ago (with a different calculational method for the defensive component, of course) in Brock Hanke's version of THE BASEBALL ABSTRACT 1989, which continued on into THE BIG BAD BASEBALL ANNUAL through 2001. Back in the day we didn't do the little trick with the lower-case letter, however, we just called them "OWAR" and "DWAR."
Joe Posnanski's elliptical tour through WAR has some nice nuggets. However, I'd like to see a breakout of MVP awards that shows the distribution in ten, twenty and forty-year increments from the beginning to the present, to see how that pattern may (or may not) be changing. (I also wonder what the seventy-year distribution is, since Joe didn't decide to use that!!) What's also interesting is that Joe's claim for middle IF MVP winners has at least one anomaly: Dustin Pedroia in 2008, who's the eighth-best player in the AL that year according to WAR.
Hope that you'll find time to add breakouts that create something akin to the position-based value lists that used to appear in BBBA. Right now I see those only on your "player value" directories, were the sorting doesn't really allow for such a breakdown. Keep up the good work!
September 15th, 2010 at 10:48 pm
Sean,
I am very happy to see that you are planning to work on a breakdown of how WAR numbers work. I'm most interested in the defensive component, since that often seems to conflict with my perception of who is or is not an effective fielder.
Sometimes the numbers make no sense to me. For instance Ryan Howard Rtot = 15; Rdrs = -10. It's great that you carry two figures, but why do they differ that way?
================
@#3, Thanks for bringing up Posnanski. I read his article and some of the responses. In the responses I found what bothers me about WAR:
Ryan Howard WAR=1.8
Jamey Carroll WAR=2.4
Maybe Jamey Carroll is a better player than Ryan Howard this year, but you really have to understand how the numbers are put together to back that one up in an argument! Assuming salaries and ages were not an issue, what GM would feel confident that he could trade Howard for Carroll and be sure that his team got the better of that deal?
September 16th, 2010 at 2:08 am
Sean,
One issue that might be contributing to your thought that the position adjustment should be included with offense and not defense is how you think the position adjustment is, (or should be), computed. Some of your comments lead me to believe that you think they are properly figured from the average differences of offensive performance between different positions. This is incorrect. They are properly figured by looking at players who have played multiple positions and seeing how well they defended at each. Tom Tango has done a lot of work on this, shown and discussed at insidethebook.com. The offensive differences are due to positional scarcity, as others have noted. For some years in the 1950's center fielders hit better than right fielders. (Willie, Mickey and the Duke). This would not make it correct to give right fielders the bigger position adjustment because they hit worse. They still were worse fielders, and that is all that matters here. One expects players at the harder defensive position to hit worse, but it is because there are fewer players who can defend well enough at the harder position. If you took 1000 players who were good enough at defense to play 1st, probably less than 100 of them would be competent at short. You would obviously expect the top and the average to be much higher in the 1000 than in the 100 drawn from that for their defense. Yes, the average differences in offense are about the same as the defensive differences between positions. This just shows that there are no huge market inefficiencies. You may reasonably use the offensive differences to approximate the position adjustment, but that is not the cause of the differences and it is technically the wrong way to measure it.
I think that if you took a reasonably long career, say 12000 defensive innings, that you would have to conclude that the worst defensive shortstop, who was good enough obviously to last at short, provided more defensive value than the best first sacker. Putting the position adjustment with the offense hides this, as most of the additional defensive value is measured in the position adjustment, not in wins above or below average.
Just about the only times you might really prefer to do it the way you have is when you are comparing players at the same position. But then the positional adjustment would only differ when playing time differed. You already have the replacement wins to do this on the offensive side, so even here you want to have the PA added to the defense above average to get the total defensive value. You don't want to say that a SS who was average in 100 innings provided as much value to his team as one who was average in 1000 innings. People have already commented on how very much smaller dWAR is compared to oWAR. Since offense is equivalent to pitching plus defense one expects it to be bigger than defense alone. There is much value in being average. If it takes 93 wins to have a good chance to make the playoffs, then average is 70% of the way from replacement to playoff. The replacement wins added in the oWAR take us to average for offense, but without the PA in dWAR we are missing much of the value, especially for average or below players. (It is true that the way I want to do this an above average player at an easy position such as LF or 1B will come out negative in total dWAR, but that is just a result of the empirical fact that defensive replacement at a position is essentially average and that the PA's are set up to zero out for a NL team. (The DH unbalances the equation for AL teams)).
September 16th, 2010 at 8:03 pm
DavidRF Says:
"A good chunk of oWAR for non-star players is just "rRep" which you just get for showing up. Its ostensibly a durability bonus."
Nope, rRep is not earned by just showing up. rRep is earned by playing better than a replacement player. Since the fielding and batting components of WAR are compared to average you need to add the amount of runs added by the player for playing better than a replacement player.
September 16th, 2010 at 11:37 pm
I just thought I would mention, as one of the commenters that expressed doubt over the defensive metrics, I jut finished reading the very long article in the SABR Baseball Research Journal regarding just this idea, and I just kept coming back to the same conclusion: nothing is certain, nothing can be shown over a long period of time (since so much is based on info we only have from 2006 on, and most of the rest dates from 1989 on), and a lot seemed to be based on adjustments that never seemed to be fully explained. As in, we've penalized 1st basemen because 1st base is traditionally an offensive position.
Anyways, I'm really hoping to see a defensive metric that truly seems to capture everything come up in the next few years, and I'm sure that between all the amazing researchers at terrific organizations out there, we will have one. I love the idea of WAR, and I love the execution for the most part. The turning point for me was the Kenny Lofton vs Manny Ramirez blog post, where I realized that Manny had been penalized TWELVE WINS over his career for the simple fact that he plays right field. Not HOW he plays, but THAT he plays.
I really wanted to go the the PitchFX Summit this year, but I just didn't get a chance to, maybe that would have helped me have a better appreciation for fielding metrics. I think it's a super important area that has never gotten the recognition that it deserves over the history of baseball. I really appreciate the new numbers. And one of these days, I'll gain a fuller understanding of fielding metrics and learn to accept them more.
September 17th, 2010 at 2:13 am
Jeff, Manny plays left field not right but that number sounds correct. We have lots and lots of data of players spending some time at center and some time at left. If you switched an average CF with the average LF we would expect the new LF to be about +10 runs/season and the new CF about -10 runs/season. So the positional difference between center and left is about 10 runs. Compare every position to every other in this way and we get the defensive spectrum. This puts the average CF at about +2 so that is his position adjustment. The average LF comes in 10 runs worse or -8. Give Manny 15 years worth of games at -8, (and when he is not in left he is DH, with an even bigger PA.) and he gets docked 12 wins. Manny is a horrible defender at an easy, and therefore less valuable position. If he were just as bad but at an average defensive position he would be worth 12 wins more to his teams.
September 17th, 2010 at 12:56 pm
I am not sure I understand how WAR works, I thought I did but after looking "Taken from our 2010 MLB Batting Value Page" and sorting by oWAR it seems to be telling me Ryan Howard (.279/30/100 .862 OPS) is only the 79th best offensive player in baseball behind the like of Posada and Stephen Drew?
1. Please tell me I am going about this the wrong way.
2. Please tell me how to understand this tool/stat better.
P.S. Love this site!
September 17th, 2010 at 2:56 pm
The oWAR I am showing here includes the positional adjustment.
September 17th, 2010 at 3:08 pm
Yeah, BA, he ranks 79th offensively when adjusted for position.
If you just rank players by WAR batting, disregarding position, he's around 50th. He is 36th in OPS+, so that's not very different. OPS+ overrates him slightly because it undervalues OBP a bit, which WAR batting corrects for.
September 18th, 2010 at 9:56 am
[...] WAR without the defense (oWAR) & WAR with just the defense (dWAR) (Baseball-Reference). Sean talks about everyone’s favorite new controversial stat. He also explains replacement level (fascinating discussion ensues in the comments). [...]
September 18th, 2010 at 11:04 pm
Yeah, left field, sorry, must have been upside-down when I typed that. Anyways, my point is fairly well summed up by your statement "If you switched an average CF with the average LF we would expect the new LF to be about +10 runs/season and the new CF about -10 runs/season." And that's what I don't get. WHY would we expect that? How did we arrive at that number? In my mind, that's penalizing guys twice. Once for being a bad fielder, and once because you expect them to be a bad fielder. And the guys who actually are GOOD fielders, it's as if they're average, because they're expected to be bad. It just seems like a superfluous and spurious adjustment. I'm probably wrong, but that's how it comes across to me.
September 19th, 2010 at 1:49 am
Jeff, Mostly because enough players play both positions within a season or in consecutive seasons that this has been measured, and the average is about 10 runs per full season. In economic decisions one should never expect complete efficiency, because information is often lacking or misinterpreted, but you will find in general that most free agent signings line up well with WAR, (actually with the predicted WAR of the future years under contract). Specifically including the position adjustment. And of course while the position adjustment comes from the defensive difference between positions, it can usually be approximated by
the difference in offense, so the markets are very roughly efficient. But only very roughly.