Saturday, July 12, 2008

Phil Birnbaum's Sabermetric Research blog reports on a new study (originally from an L.A. Dodger fan site) of winning streaks in Major League Baseball and their consequences. Specifically, the new study asks whether short winning streaks raise a team's probability of winning its next game, beyond the team's established (baserate) winning percentage from large numbers of games.

Such a comparison is one of the earliest approaches to defining and testing for the existence of the "hot hand," as used by Gilovich, Vallone, and Tversky (1985, Cognitive Psychology). Gilovich and colleagues studied basketball shooting (for details, see the following PowerPoint lecture, especially slides 20-23).

The recent baseball winning-streak study -- like Gilovich and colleagues' earlier basketball-shooting study -- failed to find evidence of a hot hand. Specifically, quoting Phil, "...after winning three, four, or five consecutive games, [teams] won the subsequent game far less often than you'd expect from their record."

These findings also are similar to those for an older NHL hockey study that I discovered recently. The authors of this study concluded that, "NHL teams do indeed play better than normal after a few los[s]es and worse after a few wins."

Thus, contrary to the impressions of many athletes and sports fans, a number of studies suggest that putting together a string of successes (e.g, games won; shots made) does not appear to give performers momentum, in the sense of raising their probability of success on the next trial beyond established baselines.

No comments: