While a 1985 study attempted to refute the “hot hand” phenomenon citing basketball statistics, Ya’ari claims that it lacked an adequate sample size. Courtesy of Dr. Gur Ya’ari.
When an NBA basketball player seems to be making shot after shot in a game, we naturally expect that the probability of his next shot being successful will be higher than normal. However, in 1985, Cornell professor Thomas Gilovich conducted analyses of basketball free-throw statistics to report that the idea of a player being “on fire” is nothing more than a misperception of random sequences. Yet, 27 years later, the work of Dr. Gur Yaari, Postdoctoral Associate in Department of Pathology, Yale School of Medicine and Dr. Gil David, Postdoctoral Associate in Applied Mathematics, has shown that the “hot hand” may not just be luck after all.
In 2011, Yaari and colleague, Shmuel Eisenmann, took another look at free-throw statistics and published an article suggesting that the “hot hand” was not simply a matter of chance, as previously supposed. Noting that free-throw shooting is too rare in basketball to draw conclusions of causility or correlation — a main problem with the 1985 study — he and David turned instead to statistics on strikes in bowling. Bowling, Yaari observed, offered “much more data, up to 100 frames per day,” for analysis.
The pair analyzed nine seasons of Professional Bowling Association data at the game, season, tournament, and career levels. There was a strong correlation between successes (strikes) at all levels, except for at the game level. Throughout a season or career, a player’s successes were concentrated into distinct periods of time, but within a single game, successes seemed to be randomly distributed. However, while the bowler’s performance in a prior frame had no causal effect on his next frame, his performance in the prior few frames could be used as a predictor. That is, a strike in the fourth frame alone did not mean a higher chance of a strike in the fifth frame, but if the bowler had been successful for eight frames, it indicated a “good game” and predicted a higher chance of success in the ninth frame. Thus, Yaari and David concluded that the “hot hand” does exist as something more than luck, but in a correlative, not causative, role. Hot streaks are not a matter of “success breeding success,” but rather one of having a “good” or “bad” day.
Bowling data for Michael Machuga. The more strikes he has had over the first 8 frames, the higher the probability of a strike in the ninth. Courtesy of Dr. Gur Yaari.
The significance of these results extends well beyond the sporting world. “While the examples we studied came from sports,” said Yaari, “the implications are much more far-reaching.” The results indicate that there is some component of the “hot hand” that could be under our control. If one can isolate the conditions that lead to better performance, sequences of higher success may be achieved. While it may seem like common sense, this radically contradicts the previously held statistical belief from Gilovich’s 1985 study.
Yaari plans to expand his studies with computer games, testing the existence of the “hot hand” in the mental realm of concentration. In addition, Yaari plans to use similar techniques to model a number of financial trends. Such work could better predict the time scales of certain events and determine, for example, whether a certain stock’s value fluctuates monthly or yearly. If Yaari’s hypotheses prove correct, it seems there may be no end to human endeavors in which “luck” can be made.
