It's a constant human tendency to see evidence in anecdote and to extrapolate from insufficient data. Nowhere does this happen more than in baseball. Managers change pitchers and swap in pinch hitters for the flimsiest of reasons. Players imagine that they're on a streak (or in a slump) when they experience a few good (or bad) breaks. And after a handful of weeks and a couple dozen games, fans and sportswriters think have already identified the best teams for the year—as if one could tell which gambler is a better coin-tosser based on a handful of flips.

The simple rule remains: don't pay attention to fluctuations smaller than the square root of the number of events. (There are mathematical footnotes to sharpen that rule of thumb, but ignore them for now.) Over a 162-game season nothing less than a 13 game difference is worth taking seriously. Last year, during regular season play one club won 105 games—but fully eight teams had 92 or more wins and were statistically indistinguishable. Several more were close to that threshold. Sure, the World Series and the playoffs leading to it are great fun. But one shouldn't believe that the "winner" is the "best". There's far too little evidence to decide.

And not just in baseball ...

(cf. HumanDiffusion (19 Jan 2000), ...)

TopicRecreation - TopicScience - TopicEntertainment - 2005-05-13

(correlates: SuboptimalPerformance, WeightOfOffice, Normal Distribution, ...)