But of course a cell-phone is different from a statistic. You can test a cell phone to see if it works. How do you test a new mysterious statistic? Self-validation is not enough. (I believe I recall such a test from Moneyball for why OPS was preferable to, say, BA). But have we studied those tests? (Some are simple--I remember mathematical demonstrations in the 60s as to why bunts were not useful--of course, MLB managers paid not attentioin). One does not need to understand, say, dWAR to 'use' it. But one is required, I think, to demonstrate (or to describe such a demonstration) of why it is preferable to other stats.