dWAR to end all WARs
Baseball Reference never stops improving, which is one of the reasons I love the place. The searches keep getting better. It gets easier and easier all the time to find cool stuff. Love it.
But I have to admit: dWAR baffles me.
dWAR is Baseball Reference’s “Defensive Wins Above Replacement.” It is such a convenient concept — a one-stop number that estimates how much value above replacement level a player brings with his defense — that I look at it all the time. But, like Inigo says in the movie, I don’t think it means what we think it means.
Take Keith Hernandez. Please.
Keith Hernandez was a great defensive first baseman over a long career. This is something everyone who saw him play agrees upon. We watched Hernandez all through the 1970s and 1980s as he reinvented how first baseman play the bunt. He was fantastic at every aspect of first base defense — great range, great arm, brilliant at scooping out bad throws — or anyway it certainly LOOKED that way.
Keith Hernandez’s CAREER dWAR is 0.6.
That seems to suggest that Keith Hernandez was just .6 wins better than a replacement first baseman over his entire 17-year career. That’s doesn’t sound particularly good. In fact, that doesn’t sound good at all. Is Baseball Reference really saying that Hernandez basically offered no defensive value above a replacement first baseman? Was all that we saw an illusion? Are we given the ugly choice of throwing out our memories of Hernandez or throwing out the dWAR statistic?
The answer is: No. That’s not really the choice we are given.
We have to understand what dWAR is.
First of all, the name is a bit of an illusion. Yes, WAR is Wins Above Replacement, and we’ve come to understand that “replacement” represents a rough estimate of a player you could find to replace a starter in case of injury or some such thing. Eliezer Alfonzo, Russ Adams, Wendell Magee, Mario Guerrero, these are replacement level players, and so it is useful to know just how much better any player is than replacement.
But that’s not how dWAR works. It is actually measuring a player against the AVERAGE Major League fielder. There are reasons for this — and it is logical to think that it is easier to find a replacement player who would be an average fielder — but it’s an important distinction.
What dWAR is saying is that Keith Hernandez was 0.6 wins better than AVERAGE not REPLACEMENT. That still doesn’t compute … until we get to the second point.
And the second point is this: dWAR includes what is called a “positional adjustment.” Apologies for telling you something you probably already know, but it’s easier to just lay it out there. To make WAR work well, you need to include a positional adjustment because a shortstop is naturally and universally more valuable defensively than a first baseman. A catcher is more valuable defensively than a left fielder. That’s obvious.
These players are more valuable simply by virtue of the POSITION THAT THEY PLAY. The circle of people who can play shortstop in the big leagues is a great deal smaller than the number of people who can play first base in the big leagues. WAR has to account for this. And so it does by this positional adjustment.
But what you might not know — I didn’t know — is that when Baseball Reference splits up WAR into defensive WAR and offensive WAR, they include the positional adjustment in BOTH. In other words, dWAR is not saying that Keith Hernandez is 0.6 wins better than the average FIRST BASEMAN. It is saying that Keith Hernandez is 0.6 wins better than the average FIELDER.
Baseball Reference founder Sean Forman makes his case logically.
“Originally I didn’t include it,” he says of the positional adjustment, “but it was really confusing to people when we did it that way. I wanted dWAR to show who the best fielders were and if you don’t include the positional adjustment, guys like Keith Hernandez and Barry Bonds show up as the best fielders of all time which isn’t really what that should be imo. I think you can also make the case that the worst shortstop is likely a better fielder than even the best first basemen, and the numbers now jibe with that.”
Sean brought up Keith Hernandez independently, by the way — I didn’t mention Hernandez in my question. But this is exactly his point: When you just look at WAR Runs Fielding, without taking into account position, Keith Hernandez ranks as the second-best fielding first baseman of all time (behind only Albert Pujols) and the 53rd greatest fielder of all time regardless of position.
So now the question: Which way do you prefer it? If you make the positional adjustment, Hernandez’s 0.6 dWAR looks puny and unimpressive for such a marvelouis fielder.
If you DON’T make the positional adjustment, Hernandez’s dWAR would be about 11.7, and he would be ranked as a better all-around fielder than Yadier Molina, Garry Maddox, Phil Rizzuto and Tris Speaker, which he certainly was not.
Now, I’ll give you my opinion: I personally would NOT include the positional adjustment in dWAR and I’ll tell you why: I don’t think people naturally compare defenders at different positions. I don’t think people often ask, “Who was a better fielder, Dave Concepcion or Barry Bonds? Joe Morgan or George Scott? Luis Aparicio or Andre Thornton?” For me, that isn’t how I think most people’s mind works.
I think when people look at dWAR w we would like to see a number that reflects how good they are AT THE POSITION THEY PLAY.
This plays into something I wrote recently: The defense of free agent Eric Hosmer. He has won three of the last four Gold Gloves even though defensive metrics across the board suggest he’s just an average-to-below-average defender, somewhere in the C- range. His agent Scott Boras recently took a shot at the defensive statistics, saying that managers and coaches (who vote for the Gold Gloves) obviously see something more in him.*
*I certainly would not tell Scott Boras how to do his business, but if the long history of Gold Glove voting tells us anything it is that managers and coaches are pretty shaky on the subject of other teams’ player defense.
Thing is, dWAR has Hosmer a minus-1.6 WAR, which sounds really terrible. And I don’t think that’s fair to Hos. More than half of that negative value comes from the simple fact that he plays first base. He was minus-7 runs defensively, which isn’t good isn’t that far from average. But then subtract NINE runs for the positional adjustment, and that makes it look really bad, worse than he deserves.
Anyway, that’s just my view. I also tend to agree with others that Baseball Reference certainly should not include positional adjustment in BOTH dWAR and oWAR. This double counts the stat and makes it so you cannot just add the two together to get WAR. Once again, Sean explains why:
“I suppose the argument could be made that oWAR should not include the positional adjustment,” he says. “But for that stat I thought it made sense to include it as you then have a fielding independent WAR stat. So you can assign your own fielding value to the player and rank them that way.”
If you’d like to chime in below in the comments, that would be great … Sean has always been open to suggestions and thoughts.
In the meantime, for fun, here are the five highest rated players at each position by fielding runs since the end of Deadball and, in paretheses, their dWAR.
- Albert Pujols, 139 runs (-2.2 dWAR)
- Keith Hernandez, 117 runs (0.6)
- John Olerud, 103 runs (-2.0)
- Mark Teixeira, 99 runs (-0.4)
- Geoge Scott, 85.1 runs (-1.8)
- Joe Gordon, 150 runs (22.4 dWAR)
- Bill Mazeroski, 147 runs (23.9)
- Chase Utley, 142 runs (17.8)
- Frankie Frisch, 138 runs (21.3)
- Mark Ellis, 136 (17.6)
- Brooks Robinson, 293 runs (38.8 dWAR)
- Adrian Beltre, 230 runs (27.8)
- Scott Rolen, 175 runs (20.6)
- Buddy Bell, 174 runs (23.0)
- Clete Boyer, 159 runs (21.5)
- Mark Belanger, 241 runs (39.4)
- Ozzie Smith, 239 runs (43.4)
- Cal Ripken, 181 runs (34.6)
- Andrelton Simmons, 163 runs (21.9)
- Luis Aparicio, 149 runs (31.6)
- Carl Yastrzemski, 184 runs (0.5 dWAR)
- Barry Bonds, 175 runs (6.7)
- Brett Gardner, 122 runs (10.8)
- Alex Gordon, 102 runs (7.6)
- Luis Gonzalez, 91 runs (-1.4)
- Andruw Jones, 236 runs (24.1 dWAR)
- Willie Mays, 185 runs (18.1)
- Paul Blair, 174 runs (18.6)
- Jim Piersall, 174 runs (15.3)
- Devon White, 133 runs (16.2)
- Roberto Clemente, 205 runs (12.1 dWAR)
- Jesse Barfield, 161 runs (11.8)
- Jason Heyward, 158 runs (12.5)
- Brian Jordan, 153 runs (9.7)
- Al Kaline, 152 runs (2.5)
This Tweetstorm from Mitchel Lichtman is very helpful in understanding the whole concept of dWAR and WAR:
I don’t think you’re really understanding how this works. I’m not sure Sean Foreman does either. There’s no such thing as defensive wins or runs above “replacement” because replacement is defined by offense and defense. And as it turns out (not that it really matters) replacement players on the average are around average on defense. All their deficiency is in offense.
There is also no such thing as positional adjustment offense. You CAN give an offensive value relative to other players at that position but you have to specify over what time period.
All that being said, you are 100% correct in that most people want to know how good defensively a player is compared to others at their position and presenting a position adjusted number confuses that. Bottom line is this: Defense should be presented relative to the average at that position AND serious people SHOULD know how to compare players ACROSS positions (by applying positional adjustments).
Offense should be presented compared to average of ALL positions. Most people intuitively know that different positions have different offenses because of the size of the pool of players that can play that position (and the physical characteristics necessary to play them). But it’s not at all necessary to know those differences or include them in the offensive numbers.
Finally it IS necessary to include positional adjustments in the “final comprehensive number” (like WAR) in order to be able to compare all players AND because when we want to know “how good” a player is we MUST incorporate his defensive position.