When small sample sizes matter

Miguel Cabrera is the planet’s best hitter, and he explained the other day why he doesn’t really draw much information from the written scouting reports available to all players: All of that is based on what has happened in the past and isn’t necessarily related to what’s happening today.

Cabrera watches some video of opposing pitchers before each game, but what he really wants to see is the pitcher throwing at the outset of a game -- in his warm-ups, in working to the first hitters of the game. Cabrera feels as if he’ll glean from that small sample so much usable information: how hard the pitcher is throwing that day, what pitches are working for him that day, how the pitcher might try to beat Cabrera that day.

"Small sample size" has become a common performance observation in dismissing particular results. It can be applied to players in September and October, but generally speaking, it’s probably heard more this time of year, as we try to wrap our brains around Josh Hamilton hitting .200 and Carlos Gomez hitting .360. "Small sample size" is employed as a cautionary phrase, as in: Be careful, don’t believe everything you see because it’s not really representative.

But here’s the funny thing about that. Small sample sizes are used in decision-making dozens and dozens of times during each game, before each game, after each game.