WAR vs. Win Shares

A little while ago, I wrote a THT Live post that compared a couple of relatively new and advanced sabermetric stats: Win Shares* and Wins Above Replacement, or WAR (the Sean Smith version).

*Actually, I used Win Shares Above Bench (WSAB), my own interpretation of how to best implement Win Shares. In this article, I’ve divided all WSAB figures by three, to make WSAB directly comparable to WAR.

You see, I have an unhealthy obsession with baseball stats denoted in wins. I think they’re cool, and I enjoy the fact that each one is a “theory” of how baseball works. That’s why I like having many win-based stats; I think there is something to learn from each one and I’m in no hurry to anoint one particular stat “the best.” That would take away the fun.

Still, I’d like further compare WAR and WSAB. We can learn from the way the two systems compare players, and we can also learn some things about the current state of baseball analysis. And, mostly, it’s just fun.

Let’s start with a graph. This is the one I posted before, of all position players (not pitchers) who played between 1900 and 2008. It compares each player’s career WSAB and WAR. There are a lot of data points here, but as you can see, they line up pretty well.

You might think that the two stats are in close agreement, and you’d be right. The correlation is high, with an R squared of .96 (although it appears the best fit isn’t completely linear). But there are still serious differences between the two systems. To make the point, I’m going to insert a line where WAR roughly equals 70 and highlight the difference in WSAB. As you can see, the WSAB results range from from about 40 to 80. That’s a “big diff.”

There’s a reason I chose 70 WAR as the cutoff. If you look to the right of that line, you’ll see that the data points are more spread out—there are less of them. It’s pretty easy to spot the top 40 or 50 best ballplayers in baseball history, regardless of which system you use. However, once you reach a certain threshold, the playing field gets crowded and the two systems start to seriously disagree.

Sometimes I’ll hear someone say that he is a “large Hall of Fame” person. That’s a generous position, but the larger the Hall the more intense the disagreements. For some of us, that’s the fun part.

Here are the players with the largest positive variances between WAR and WSAB. These are the players that WAR values more highly than WSAB does.

First    Last        WAR   WSAB  Diff
Cal      Ripken       90     64    26
Brooks   Robinson     69     45    24
Kenny    Lofton       65     42    23
Ozzie    Smith        65     42    22
Mark     Belanger     33     10    22
Buddy    Bell         61     39    22
Omar     Vizquel      42     20    22
Wade     Boggs        89     67    22
Devon    White        41     20    21
Roberto  Clemente     84     63    21

I think the Orioles should file a discrimination suit against Win Shares. Three of the top five players on our list played on the left side of the Oriole infield. The difference in ranking is most extreme in Cal Ripken’s case. He rises from the 59th highest-ranked player according to WSAB to the 23rd-best according to WAR. I’ll admit that, at times, I’ve wondered if Ripken truly deserves all the plaudits he’s received for his contributions on the field (I’ve never doubted that he was a Hall of Fame player). WAR answers a resounding yes, ranking him between Al Kaline and Nap Lajoie. That’s impressive company.

There’s an obvious trend at work here. WAR values good fielders more highly than WSAB does, particularly good-fielding middle infielders. I believe this is primarily the result of several differences in the systems.
{exp:list_maker}Win Shares puts a cap on the fielding impact a player can have. Bill James didn’t completely trust his fielding numbers, so he added limits to the credit fielding could be given.
The “top/down” approach that Win Shares uses to apportion fielding impact probably further limits the impact any one fielder could have, particularly because the system doesn’t allow individual fielders to have negative fielding values.
WAR’s fielding system, like UZR, is based on play-by-play data. James never uses play-by-play data in Win Shares.
WAR includes a position adjustment that gives credit to “high-skill” positions, such as catcher and shortstop. Win Shares does that too, but only for fielding impact. {/exp:list_maker}We can legitimately disagree about the fourth point, but I don’t think there’s any denying that the first three points are a mark against Win Shares and for WAR.

Baseball fans and pundits have always appreciated great fielding but it was nearly impossible to estimate the impact of fielding with any kind of accuracy. The first serious attempts, such as Pete Palmer’s Fielding Wins system, had serious flaws. That’s why James took a conservative approach to the interpretation of his fielding Win Shares, and why he has actively discouraged his readers from comparing players with the fielding Win Shares scale.

Now, however, thanks to improved data such as that provided by Retrosheet and Baseball Info Solutions, and analysis by my esteemed colleagues MGL, John Dewan and Sean Smith (and many others), we do have some confidence in our results. There’s no longer a need to interpret our fielding stats conservatively. Fielding stats for the Retrosheet era (from the mid 1950’s on) are valid enough for legitimate historical comparisons and Hall of Fame arguments. And there’s no need to hold back.

Let’s get into one of those arguments.

WAR ranks Kenny Lofton as the 68th best player in this particular dataset, just ahead of Willie McCovey, Tim Raines, Ernie Banks and Ozzie Smith. WSAB ranks him 190th, immediately behind Al Oliver and Ken Boyer. What should we believe? Just how good was Kenny Lofton?

Before we begin, go back and look at Lofton’s pages at Baseball Reference and Fangraphs. Given his nomadic tendencies at the end of his career (made memorable by those DHL commercials), you may have forgotten that Lofton was a great player, finishing in the top 20 in MVP voting three out of his first five seasons and winning the Gold Glove in four of them. He played a premium position and did everything very well except hit for power.

Now let’s look at a table of how our two systems rank Lofton each year (compared to all major league position players):

    Year     WAR    WSAB    Diff
    1992      17      38      21
    1993       6      29      23
    1994       2      11       9
    1995      33      24      -9
    1996      36      39       3
    1997      17      37      20
    1998      34      70      36
    1999      16      93      77
    2000      75      94      19
    2001     125     177      52
    2002       7      60      53
    2003      53      92      39
    2004     235     248      13
    2005      56      78      22
    2006     207     156     -51
    2007     104     125      21

In nearly every year, WAR ranks Lofton much more highly than WSAB does. WAR says that Lofton was one of the ten best players in the majors three different times—WSAB says he never was.

Lofton’s best case for greatness was the strike-shortened 1994 season, when he batted .349/.412/.536 with 60 stolen bases. He made the All-Star team, won a Gold Glove and finished fourth in the American League MVP voting. According to BR, he was fourth in the league in Runs Created and fifth in Batting Wins. According to Fangraphs, he was fifth in the league in wOBA. So all the advanced stats agree on his offensive production. Put the two leagues together, and he ranks about tenth in offensive production, which is where WSAB ranks him overall.

According to Sean’s analysis, however, he was 42nd in Total Zone (Sean’s range estimate) and second in outfield arm outs, behind only Raul Mondesi. I pulled out my Baseball Scoreboard for 1994 (those were the books in which John Dewan published Zone Rating each year) and found that Lofton had the second-highest Zone Rating among all center fielders (Total Zone may actually be a bit conservative). And he did have a great arm—Baseball Scoreboard notes that he led all major league outfielders in assists and was the third-best center fielder in holding runners. The bottom line is that Baseball Scoreboard corroborates Sean’s rankings.

I don’t have all the detail behind Lofton’s Win Shares figures—James includes range, arm and errors in his calculations—but I can tell you that he ranked 103rd in the majors with only three total fielding Win Shares in 1994. This seems way off, for all the reasons I listed in the above paragraph.

The bottom line is that WAR appears to be the more accurate ranking system due to its more advanced methods for measuring the impact of fielding. And it appears that Kenny Lofton does have a case to enter the Hall of Fame.

Not that I expect him to receive serious consideration. The Hall voters are the guys who just voted Jim Rice in (Rice ranks 225th in our WAR database). If the BBWAA can’t appreciate the importance of OBP, or the impact of playing in certain parks, I don’t expect them to value our new-fangled ways of evaluating fielding and baserunning and all the other things that make a ballplayer great.

But let’s see if we can spread the word anyway—if you value players for all the skills they bring to the game, Kenny Lofton has a Hall of Fame case.

Here is a list of the players that WSAB values relatively more than WAR.

First    Last        WAR   WSAB  Diff
Gary     Sheffield    64     79   -16
Frank    Howard       39     53   -14
Bill     Bergen      -18     -4   -14
Gabby    Hartnett     50     64   -13
Honus    Wagner      124    137   -13
Bobby    Murcer       33     45   -13
Yogi     Berra        62     74   -12
Dante    Bichette      2     14   -12
Manny    Ramirez      63     75   -12
Pete     Rose         75     87   -12

It’s hilarious that Bill Bergen is on this list; evidently, Win Shares just doesn’t dislike him enough. I’m surprised that Honus Wagner is on the list—after all, he was a shortstop and an outstanding fielder to boot. However, there may be some issues with Sean’s system in the “pre-Retrosheet” years—those years for which we don’t have play-by-play data.

There are also three catchers on this list, which implies that Win Shares is giving more credit to those who can adequately handle the tools of ignorance than WAR is. Ranking catchers among other positions is one of the most difficult aspects of ranking players in general, because the demands on catchers are so far outside the demands on other positions.

So I think there are some legitimate open questions here. I’m left wondering about the ranking of players who played in the first half of the last century, and I’m wondering about catchers. I still like Win Shares for several things it does, such as forcing individual player wins to equal team wins. I think there is value in its top-down approach. And I know Bill is reconfiguring Win Shares as we speak, rolling it out in portions on his website.

There is still plenty of room in my little head for these two systems. But if you’re looking at players from the 1950s forward, WAR should be your stat of choice.