Winning streaks more than random chance

Madeleine_Witt_Scitech
Photo by Madeleine Witt.

Winning streaks may be more than pure chance, a recent Yale study has shown.

Gur Yaari, a Yale pathology postdoctoral associate, and Shmuel Eisenmann, CEO of HIL Applied Medical Ltd, investigated the “hot hand phenomenon,” in which athletes’ performance appears to improve during winning streaks. Published Oct. 5 in the journal PLoS ONE, the study found that free throw statistics in National Basketball Association seasons supported the phenomenon.

“The phenomenon has been controversial, and many studies have dismissed it as a myth,” Yaari said. “Many people were skeptical about its existence.”

The report showed that, on average, professional basketball players are 1-5 percent more likely to make their second free throw if they also make their first one. Although some previous studies have found similar results, not all of them have accepted those findings, Yaari said. He added that before his study, few researchers had conducted experiments of sufficient scale to determine whether the effects were statistically significant.

Yaari recorded data on over 300,000 free throws over the course of five years. The large data set and the controlled conditions of the study ensured its free throw shot ensured the significance of its results, he said.

But the psychological interpretation remains to be resolved, since this positive difference could represent one of two effects, Yaari said. It could be a positive feedback loop in which the success of making one shot breeds success for the next. Alternately, it could involve “hotstreaks,” with the player randomly alternating between “hot” and “cold” periods. Either effect could explain the results of the study, so the evidence for hotstreaks is not conclusive.

Yaari said he will need to collect data from sports such as bowling, where the given task is repeated continually, in order to be certain that hotstreaks exist, but added that his analysis of other factors makes hotstreaks a more likely explanation than a positive feedback loop.

One other factor that supports the concept of hotstreaks might be variation between seasons. The study found an individual’s difference between post-hit and post-miss shot percentages varied across seasons, which suggests the “hot hand” is not a characteristic of the player. By contrast, if the hotstreak effect were due to psychological feedback loops, then it would remain constant for players over their careers. The variation between seasons gives more support to the hypothesis that hotstreaks exist, the report stated.

“[The results] set the stage for further physiological and psychological investigations of the origin of this phenomenon,” Yaari said.

But other experts were less confident about the strength of the study. Sidney Redner, a physics professor at Boston University, said the study seems legitimate, but the measured effect was so small that it might be explained by some subtle statistical flaw.

Jonathan Koehler, a professor at Northwestern Law School, said more complex tasks in sports, such as dribbling a ball around a defender, do not show the same trend as that found in this study.

“As far as the common understanding of “hotness” is concerned, people don’t typically regard a shooter who makes relatively easy shots over an extended time span to be hot,” Koehler added.

Yaari said he remained confident that the study was significant, adding that its implications may extend beyond basketball into other sports and activities. He said he is working to improve his statistical methods so that he can detect weaker manifestations of the hotstreak phenomenon in bowling.

This study found that, on average, players are 4-8 percent more likely to make their second shot than their first shot, regardless of whether the first went in.

Comments

  • The Anti-Yale

    *Yaari said he remained confident that the study was significant, adding that its implications may extend beyond basketball into other sports and activities*

    Bernie Madoff had a long, good run.

  • An_Observer

    The margin of measurement error in the summary statistics reported seems awfully wide. For example, the article says that “players are 4-8 percent more likely to make their second shot than their first shot, regardless of whether the first went in.” Why is the range four points wide? Given that NBA players generally make about three-quarters of their free throws, the measurement error here is 5-11 percent of the underlying statistic being measured, a huge error range. Shouldn’t there be one single number for the percent of first shots made and one number for the second shots? The difference should be a precise number. Why is wide range given instead?

    As it is, the 1-5 percent differential for “the hot streak” of a player who made his first shot is well within the 4-8 percent error range reported for ALL second shots. It seems as if the observed “hot streak” could easily be measurement error.

  • Referee5

    > Why is the range four points wide?

    The 4-8 percent refer to the variation across seasons of the average increase in individual success rates – it is not a “measurement error”.

    > “Shouldn’t there be one single number for the percent of first shots made and one number for the second shots? ”

    Yes there is. One number for each player for each season according to the paper.

    > Why is wide range given instead?
    As mentioned above – it reflect the variation in this observation across seasons.

    > could easily be measurement error

    It is not so clear to me what do you refer to when you write “measurement error”…the variation between players and between seasons ? I guess this is exactly why the authors analysed the data at the individual level and for each season separately.