WAR has come a very long way in the past decade, it used to be a stat used only by nerds and hardcore baseball fans, but ever since the Mike Trout-Miguel Cabrera debate, WAR has been exposed to the average baseball fan a lot more, and even ESPN is carrying it. Position player WAR is a fantastic stat. The hitting portion of it is almost perfect. It's not perfectly predictive, but that's not what it's aspiring to be, it aspires to tell how much value a player produced for his team, which it does. The Fielding portion of WAR has some problems with hit location data, and should not be trusted as closely to the decimal point as hitting WAR, but someone who used cameras on the Cardinals stadium, who I cannot remember the name of, said that BIS hit location data was close to the real hit location data.
Pitching WAR is where there is a huge problem. Fangraphs and Rally war used by baseball-reference calculate pitching WAR differently, but both have huge flaws. Fangraphs has both FIP WAR and RA9 WAR. The good thing about FIP WAR is it only measures what a pitcher does (not completely true with pitch framing, and other things a catcher can do, but let's leave that out of this) but FIP excludes luck on balls in play. Now excluding luck is not necessarily a bad thing, but we want WAR to tell us how much value a pitcher provided for his team, and luck is included in that, therefore WAR with FIP is not a good stat. RA9 WAR is not that great either because it does nothing to separate a pitcher and his defense. Runs are saved by a pitcher and the defense, the pitcher should not be given all the credit. It has the same problem Wins and RBIs have in that RA9 WAR is heavily influenced by a pitcher's teammates. The Rally WAR used at Baseball-Reference tries to correct for this, but does it very wrong. It uses the team's defense as a whole and adjusts the pitcher's RA9 based on that, but there is a huge flaw with this. As we know, over the course of one season, because of luck and random variation, the position players on a baseball team do not always hit as well for one pitcher as another. The Red Sox scored 4.1 runs per game last year for John Lackey, and 5.4 runs per game for Jon Lester. It should reasonably be assumed that defense works the same way. The way Baseball-Reference calculates pitching WAR would be like adjusting wins based on team run support, and they would bump John Lackey down even more, when in reality, Lackey should have had more wins, not less.
The reason this problem exists though, is not because Fangraphs and Sean Smith are lazy or uneducated, it's just that we don't have the data to make a correct pitching WAR. The best way to make pitching WAR would be to use PZR by MGL which is UZR for pitchers, in which it measures the exact amounts of runs a defense saved for a pitcher. There are problems with UZR, but a pitching WAR with PZR would be far better than anything we have now. The problem is, MGL has not released this stat to the public or given it to Fangraphs or Baseball-Reference. Until he does, it will be very hard to measure exactly how much value a pitcher provided for his team. FIP is probably better right now than what Baseball-Reference does, so I would use Fangraphs pitching WAR, with the caveat that it is not measuring the exact value a pitcher provided, but it is measuring a pitcher's performance.