Saturday, March 30, 2013

WARs Move Closer Together

One of the biggest complaints from fans and media about the Wins Above Replacement (WAR) concept is that different sites (such as Baseball Reference and FanGraphs) have different results for the same statistic.  This has never been a problem for fans of sabermetrics who understand that each WAR is just one site's estimate of a player's overall value.  The goal is the same in each case, but calculations differ according to the the creator's perspective.  The lack of agreement has never sat well with the less mathematically-inclined however.

One of the differences between Baseball-Reference WAR (rWAR) and FanGraphs WAR (fWAR) had been different replacement levels.  Baseball-Reference had always assumed that a team of replacement level players would win 52 games over the course of a season, while FanGraphs assumed 43 wins.  In other words, a Baseball-Reference replacement player was better than a FanGraphs replacement player.  Thus, the two statistics were on a different scale with rWAR tending to produce lower values than fWAR.

The owners of the two sites - Sean Forman of Baseball-Reference and David Appleman of FanGraphs recently got together and arrived at a unified replacement level.  The decision was that both sites would now assume that a replacement level team would win 48 games.  So, now the two models are on the same scale with rWAR going up a bit and fWAR coming down some.

There are still differences between the WARs - batting and base running runs are calculated differently and defensive runs remains the biggest discrepancy.  However, replacement level which is probably the most confusing and misunderstood part of WAR is now consistent which makes it easier to explain other differences between the rWAR and fWAR.

Table 1 below compares rWAR and fWAR in 2012 for current Tigers starters.  Most of the players are within one win on rWAR and fWAR.  The biggest discrepancy is for shortstop Jhonny Peralta who is 1.2 on rWAR and 2.5 on fWAR.  The 1.3-win or 13-run difference can be completely explained by defensive runs.  Peralta was -1 runs on Defensive Runs Saved or DRS (Baseball-Reference's preferred fielding metric) and +12 on Ultimate Zone Rating or UZR (FanGraphs).

The 0.7 win gap Andy Dirks can also be totally explained by fielding - +2 on DRS and -5 on UZR.  Not all the variation for all players can be explained by defense, but it is the main source of disagreement and can easily be pulled out when making comparisons.

Table 1: Comparison Between WARs for Tigers, 2012

Player
Baseball-Reference WAR
FanGraphs WAR
Difference
Cabrera
7.3
6.9
0.4
Hunter
5.8
5.1
0.7
Jackson
5.5
5.2
0.3
Fielder
4.9
4.9
0.0
Infante
2.8
3.0
-0.2
Peralta
1.2
2.5
-1.3
Avila
2.5
2.4
0.1
Dirks
2.3
1.6
0.7

Data Sources: Baseball-Reference.com and FanGraphs.com

So, the WARS are now a little more similar which I think is a good thing in the big picture.  There is still the issue of how complicated the WAR calculations are, but that is a topic for another post.  For now, Forman has a good article on that on Baseball-Reference.       

No comments:

Post a Comment

Sabermetrics Book

Sabermetrics Book
One of Baseball America's top ten books of 2010

Blog Archive

Subscribe

501 Baseball Books

501 Baseball Books
Recommended by Tiger Tales

Stat Counter

Site Meter