As it turns out, I develop statistical models for a living. Sadly, the Phillies have not asked me to do so.
But one doesn't need a model to form opinions from data. And in this case, the question was not whether Nava would project to be a plus hitter, but whether he was worth keeping, among their other options, given the possibility of flipping him later. To be worth keeping, he needed to have a decent probablilty (maybe 25%) of hitting well enough to be worth something in a trade.
The data included Nava's past two years, but also his previous years, in which he was a decent hitter, and spring training, which, while not as important as other data sources, is not meaningless information. If the choice was only between keeping Nava or Saunders, the obvious choice would have been Saunders. But the choice was essentially choosing three of Nava, Coghlan, Stassi, Goeddel, and Saunders (or two of the first four if one assumes that cutting Saunders was not likely). Some comments were made at the end of ST that keeping Nava was a dumb choice. I thought the Phillies chose well (though I would have kept Goeddel over Luis Garcia myself), and so far so good.
It's not a matter of projecting correctly. The people who bravely publish projections of player performance each off-season are wrong on a large number of them, even if they have good models. Projecting point performance is less important than projecting ranges and probabilities. Because of uncertainty, Nava's range would have been large enough to include his current performance, which certainly has value to the Phillies.
