ONCE UPON A time, sabermetrics was easy. Really. In the first few decades of sports statistical analysis, researchers almost exclusively studied baseball, a game of individual performances that add up to a team result. And the simplicity made for straightforward conclusions: If Carlos Beltran is a better player than Vernon Wells, we can reliably project the Yankees will be better playing Beltran, regardless of whether Jacoby Ellsbury or Brett Gardner mans centerfield. One plus one equals something pretty close to two.
Now that analytics has moved into more dynamic sports like basketball, hockey and soccer, researchers have to reckon with games in which teammates can affect each other's play far more than they do in baseball. Just because the Bruins' David Krejci has more assists than fellow center Patrice Bergeron doesn't necessarily make him a better fit to play with winger Brad Marchand, Boston's leading goal scorer. In fluid sports, one plus one might equal two or four, or minus-three.
And here's the thing: It's in the interaction effects among players that we find the true impact of coaching. To win in continuous sports, coaches must deploy athletes in real time who fit together and augment one another's skills. They must be masters of chemistry. And statheads are now figuring out how to grade them on their lab work.