A Quick Note On Those "Whoever Wins Game 3 Wins The Series" People

During the closing minutes of Heat-Mavs last night, Mike Breen said that the game would go "a long way towards determining who wins the series." That statement, while unnecessary, was at least true. I'd like to talk about the stat that can appear in any game of any playoff series: that the team that wins Game X wins the series Y% of the time. Yesterday, the stat de jour was that in all 11 previous 2-3-2 NBA Finals in which the series was tied 1-1, the team that won Game 3 went on to win the championship every time.

That's a fun piece of trivia, and there's nothing wrong with using it as such (the author of that particular piece, for instance, knows his stuff). But if you think there's any predictive value from past matchups, or, heaven forbid, you think it means that one game is more important than another, you're wrong. Here are the problems with this type of stat:

  • Small sample size. Like the batter vs. pitcher stats in baseball, even if the stat had legitimacy, looking into specific situations inevitably results in samples too small to do anything with.
  • Selection bias. if a team wins a given game in a series, they're probably the better team. Even if the series was tied 1-1 before, the team that has won two out of the three first games is, more likely than not, better than the other team. Now the better team has to win two games before the worse team wins three. Guess which will happen the overwhelming majority of the time.
  • Cherry-picking; related to the above. You might notice that when the announcers or writers use this stat to make a point about the importance of one particular game, they don't give you the same stat for the other games in the series. That's because...
  • Every game is important. Yes, believe it or not, in a short series, individual games have a large effect on the outcome of the series without any special precondition.
  • Every game is equally important. Here's the mathematical proof. While the author is a schmuck, the table still shows that no one game is more important than any other.

The counter-argument is that the games aren't independent, that momentum coming out of one game changes the outcome of the series. I don't dispute the possibility, but I'd challenge you to show that the effect exists. You'd run into the sample size and selection problems above. For now, let's stick with what we know and avoid dumb, teleological speculation on what we don't. Besides, I hear Game 4 is gonna be huge.