It pains me to say it, but he's at least partially right. The aggregate data allows us to challenge established (and possibly inaccurate) strategies in general, but it is not correct to blindly apply it to specific situations. What matters in an individual situation is the likelihood for the players coming up in the game to get the runner in. I suspect the data would show that sometimes in individual situations it is correct to bunt, and most of the time it is not.