Kelly Gaffney writes,

The problem with WAR in my mind is it fails to accurately account for the scarcity of opportunities.  While you can build a roster of 25 player with 1 WAR ability for basically $15 M, you only get to use 9 of the hitters at one time.  You need to focus acquiring the best 15 players, not the best 25 (stars and scrubs). 

At the micro level, Bill James' most important insight was appreciating that the second most important counting stat in baseball is outs (following runs).  Players that make a lot of outs to generate runs were hideously overvalued in the 70's and 80's.  

It probably took James ten years to convince people of the importance of OBP, that "the most important difference in baseball is between safe and out."

In roster construction and WAR, the value of outs must also be fully appreciated.  While this is acknowledged at the on-base percentage level, it is not fully accounted for with regard to great hitters in my opinion.  Special hitters in the corners, allow you to carry Brendan Ryan and Franklin Guittierez -- special defenders at the positions that get the most opportunities.  If you have a great SS and 2B and a ground ball staff, you can afford to run Manny Ramirez out there every night.  While I understand that WAR 'accounts' for these effects with replacement level and positional adjustments, the defensive scale is generally -2 to +2 wins, while the offensive scale is -2 to +7 wins.  The roster flexibility that Albert Pujols generates is not accounted for with WAR.  WAR undervalues great players, because they make the job for the GM, the manager, and the banjo hitting shortstop easier.

To zeroth order, baseball is a game where the contributions of the individuals are additive. At present, no one has a compelling theory for first order effects, though winshares and Matt's quantitative descriptions of defense certainly try with some success. That the theories are hard to construct and challenging to understand, doesn't make the effects irrelevant or even small.

The difference in real value between +2 and +4 and +6 WAR players is not linear; it's more like geometric.  The GM's have always understood this.  Any time you find a player on the FA market who genuinely is a safe bet for +6 WAR going forward, you are going to see him go for much, much more than two +3 WAR players ... 

...

WAR is a great stat, an improvement on James' Win Shares concept, which probably defined 25 years of his work.  There is no overstating the value of WAR.

Well, there is overstating it :- ) because a few sabertistas want to represent themselves as having 990 out of 1,000 light bulbs on.  The tone of the comments in this thread made me quite sad.  

Some guys work very hard to --- >position themselves to --- > sneer at alternative intellectual positions.  To them it's not enough just to call the other side wrong; they must always call the other side worthy of ridicule.

If WAR were used as Bill James intended Win Shares to be used -- as one of many tools, or even the best tools, for analyzing trades of players at different positions -- that would be great.  

The problem comes when WAR is used as Truth, to mock and deride decisions made for different reasons.  As James said, "We do not have near-perfect measurements of baseball players.  It is foolish to assume that we do."  

No baseball team has ever used Win Shares or WAR, even primarily, to win the pennant.  Jack Zduriencik didn't trade Putz for Gutierrez, Carp, and Vargas because of WAR.  And the Rangers didn't trade Smoak for Cliff Lee because of WAR.