NOTE: updated March 7, 2013 due to incorrect values for Runs for several teams.
In my last FanPost regarding WAR, commenters were suggesting that actual record isn't a good reference for WAR, since it is likely that a team could have many wins that were the result of close games, and then have blowouts losses, such that there is a large discrepancy between pythagorean record and actual wins (what BR calls "Luck"). I even noted that there was a slight correlation between RT Wins and Luck. OK, so let's forget about actual wins, and let's look at run differential.
WAR is attempting to measure runs created/saved by players, which can also be interpreted as adding to a team's run differential. A team's total WAR can be converted to this added run differential by multiplying it by 10. If we take a team's total added run differential from WAR and subtract it from its actual run differential, you would get a team's replacement level team run differential. I would expect this value to be pretty consistent among all teams, if WAR is representing what it is supposed to. But I'm finding this is not true.
|Team||Off fWAR||Pit fWAR||Total fWAR||R||RA||R-RA||fRT Runs||fRT Wins|
In this chart, sorted by total fWAR, fRT Runs is run differential minus 10 times total fWAR. fRT Wins = 81 + fRT Runs / 10.
You again notice strange deviations, such at the large values for Blue Jays, Mariners, and White Sox, and small values for the Diamondbacks, Rockies, and Brewers.
Perhaps it is as others have suggested, that Pitching WAR just needs some more work. That may be true, but until it is improved, or even after it is, I think there might be a way to handle this, and I'll have another FanPost coming soon with my idea.