If you're anything like me, the idea of advanced metrics appeals to you, but the thought of doing real math is a major turn-off. For me, the threshold is simple: if the equation can't be easily solved while also bonging a beer, I'm not interested.
Run differential just happens to be one of those advanced metrics that can be understood, and quite quickly and easily translated into something useful (predicting future win/loss records) without having to come within 20 feet of a cosine. (Whatever the hell that is.)
Run differential itself is just a matter of basic subtraction.