You may have noticed that there are two commonly used Wins Above Replacementnt (WAR) statistics for both hitters and pitchers:
Today. I’m going to discuss the pitching version of each. There are pros and cons to both measures and the statistical community is still debating the merits of the two. fWAR is based on FIP and therefore only considers events which a pitcher essentially controls – K, BB, HR, IP. It also adjusts for a pitchers home park.
The fWAR statistic has some potential drawbacks. First, it considers BABIP to be completely out of control of the pitcher. It also gives a pitcher no credit for distribution of base runners or sequencing events. For example, a pitcher that pitched well with runners on base would not get credit for that.
The rWAR statistic was invented by Sean Smith and is found on Baseball-Reference. It starts with total runs allowed by a pitcher and then adjusts for team fielding behind that pitcher. It considers IP run average (RA) - which is the same as ERA except it considers all runs rather than earned runs - and the Total Zone statistic. It also adjusts for the pitcher’s home park. The advantage is that it tries to tease out defense rather than completely ignore balls in play.
A potential shortcoming of rWAR is that the measurement of team defense in a single season is still shaky. It’s also possible that rWAR gives a pitcher too much credit for limiting hits on balls in play, distribution of base runners and sequencing.
So, which one is better? I prefer the concept of rWAR better. I think it’s good to start any kind of run prevention evaluation with runs allowed. The next step would be to determine how much of run prevention is pitching and how much is fielding. rWAR attempts to do that but the measurement of team defense in a single season is still too tenuous for me to use rWAR by itself.
Over the course of a career or several seasons, rWAR becomes more reliable because variation in team defense evens out and we learn more about a pitchers ability to control BABIP, base runner distribution and sequencing over time.
fWAR is valuable because it tells us how good a pitcher was at events over which he has the most responsibility. Thus, it is better statistic than rWAR for identifying pitching talent and projecting into the future. However, it makes too many assumptions about events which a pitcher does not control by himself for me to use it as a stand alone measure.
So rWAR is best for career measurement and fWAR is good for projection. How about evaluating which pitcher did best last season? I would not use either by itself, but rather I would make a judgment using both. One way to do that would be to take an average or weighted average of the two statistics. For now, I’ll look at the Tigers starters in 2010 using a straight average (Table 1).
Table 1: WAR for Tigers Starters in 2010
We can see that Justin Verlander was 4.5 wins above a replacement player according to Fan Graphs and 3.4 using Baseball-Reference. That averages out to 4.0 WAR. Based on that, we would say that he has added an estimated four wins to the Tigers beyond what an average player would contribute. Max Scherzer is next in line with an average WAR of 3.1.
All of the Tigers did better on fWAR than rWAR except Armando Galarraga. Galarraga does poorly on f WAR because he has not done a good job on things that a pitcher most controls (fWAR). He does better on rWAR because he has a low BABIP and relatively low ERA.
In future posts, I am going to look at average WAR for pitchers on other teams.