Tangotiger Blog

I understand why there are differences in WAR between Baseball-Reference and Fangraphs, but I would prefer they not have differences as large as three wins, as in the case of Drew Hutchison, Andrew Cashner, and Chris Sale, three pitchers mentioned in the article.

An extreme example for this year is Zack Greinke versus Chris Sale. According to bWAR, Greinke was a nine-win pitcher (9.3 bWAR) and Sale was at 3.3 bWAR. That’s a huge difference. According to fWAR, Sale provided more value with 6.2 fWAR to Greinke’s 5.9.

Greinke pitched 222.7 innings with a 1.66 ERA and 2.76 FIP.
Sale pitched 208.7 innings with a 3.41 ERA and 2.73 FIP.

I understand that bWAR is based on actual runs allowed and fWAR is based on FIP and I can keep that in mind when I look at their numbers, but it is difficult to get people who are just getting interested in a more analytical look at baseball to understand this.

Unfortunately (or fortunately, or neither), there are many ways to look at and evaluate past performance and call it “wins”. That is especially true with pitchers.

When you use one or two words to try and describe or quantify something that requires many words, you often run into this problem. Not unlike the “skill” and “luck” dichotomy we are discussing on the DFS threads.

I agree with MGL, as it’s both fortunate and unfortunate.

It simply shows that, given the data being used, we are acknowledging the uncertainty present.  Just because Greinke has a low BABIP doesn’t mean that Greinke has a low BABIP: it means the DODGERS WITH GREINKE have a low BABIP.  How much that is the fielders and how much that is the pitcher is up for grabs.

Similarly, BABIP with men on base would need to be split.  Do we know how well the Dodgers fielders actually played with Greinke?  How about when the RF throws out a runner at home?  Well, guess what, rWAR says that Greinke did that, while fWAR says he didn’t.

Therefore, we should be grateful that Fangraphs and Baseball Reference have taken such polar opposite views, since the truth MUST be somewhere in the middle.  But if one or both move too much in the other direction, they may go past the middle!

So, if they are too close, it’s going to be worse. 

Just split the two in half, and move on.

Tango, that’s a good way of explaining it, that the truth must be somewhere in the middle between Fangraphs and B-R. I was looking at the AL Cy Young Award and the different WARs for the top six pitchers:

7.2 bWAR, 6.1 fWAR, 6.7 combinedWAR—Keuchel
6.0 bWAR, 6.4 fWAR, 6.2 combinedWAR—Price
4.2 bWAR, 5.5 fWAR, 4.9 combinedWAR—Kluber
5.8 bWAR, 3.8 fWAR, 4.8 combinedWAR—Gray
4.3 bWAR, 5.3 fWAR, 4.8 combinedWAR—Archer
3.3 bWAR, 6.2 fWAR, 4.8 combinedWAR—Sale

I mentioned in the top post how Gray and Sale were valued so differently by the two versions of WAR, but it’s also interesting to look at Archer and Gray. If you split the difference between bWAR and fWAR, they were equal. The numbers the voters like to see, wins and ERA, both favor Gray, while Archer had four more innings pitched and many more strikeouts. Gray finished 3rd and Archer was 5th in the voting. Sale snuck in there at 4th with one more win than Archer, one less win than Gray, and a higher ERA than both. Kluber was way down the ballot, with just 9 wins and a 3.49 ERA, but with a 4.9 WAR when you split the difference between Fangraphs and B-R, slightly better than Gray, Archer, and Sale.

I don’t subscribe to the “truth is somewhere in between” theory. Especially with pitchers. There is no “truth” because there is no way to precisely define what the hell you’re measuring. It’s like all the duscussions we’ve had about MVP awards. There are so many reasonable ways to describe retrospective value. The only thing the various WAR tools have in common are the definition of replacement (which can never be precisely measured) and the runs to wins conversion (and that is approximated).

And then there is the issue of UZR. As we’ve discuses a million times, the “truth” is a regressed version of UZR.

So lots of issues that I don’t think are properly addressed with a “truth is somewhere in between” construct.