Research, analysis, evaluation

The Truth about Baseball

People are leery of statistics, except for sports statistics. Sports statistics are accepted uncritically. Even the most basic sports statistics may be questioned, though.

For an example we'll look at won-lost records. As I write this, major league baseball teams have played over a hundred games. The Chicago White Sox seem to be doing exceptionally well in the American League, and the Tampa Bay Devil Rays exceptionally poorly. However, a little statistical analysis finds differently.

If we take the won-lost records of all the teams in the American League (I used the records through August 4, 2000) and perform a test for goodness of fit, we find that the distribution of won-lost records (and consequently of winning percentages) is not significantly different from a distribution you would expect to find if all the teams were equal in ability. That is, none of the teams has demonstrated superiority over the teams it has played.

Of course, the schedule is unbalanced and there is some chance, probably a small one, that the best teams' records are worse than they should be because they are playing each other more often than they would in a balanced schedule. Nevertheless, all that proves is that when you have an unbalanced schedule, winning percentage is not a good guide to the relative ability of teams. In particular, it is not a good criterion for selecting the wild card team in the playoffs.

The distribution for the National League does vary more than you would expect by accident, although the difference is not great. Individual tests on each team's records find Houston doing exceptionally poorly and Atlanta doing exceptionally well. That means the rest of the teams aren't doing significantly better or worse than breaking even.

By the end of the season, when each team has played 162 games, additional significant differences may appear because of the increase in sample size. You wonder, though, about the validity of the standings in sports which have shorter schedules than baseball.

The Truth about Baseball © 2000, John FitzGerald
Home page | Decisionmakers' index | E-mail