Do It Again! Proof that the Hot Hand Exists
In scientific literature the hot hand has been defined as a tendency for good outcomes to cluster, or the perception that good outcomes are more likely after a streak of good outcomes.
The tendency for people to believe in the hot hand, even when it is not there, has often been used to explain behavioral anomalies in financial markets, sports, casino gambling, and hedge funds management. This behavior is known as the "hot hand" fallacy in scientific circles.
Interestingly, the origin of the hot hand fallacy, and still the strongest evidence of its costly impact on decision making, comes from the sport of basketball. In the seminal 1985 paper, Thomas Gilovich, Robert Vallone and Amos Tversky (GVT) show using in game data, and a controlled shooting experiment, that coaches, players and fans are incorrect to believe that some shooters get hot, and thus are more likely to score.
A field experiment by Joshua Miller (Department of Decision Sciences) and Adam Sanjurjo (Universidad de Alicante), summarized in the working paper A Cold Shower for the Hot Hand Fallacy, re-opens the question, showing that the belief in hot hand is not a fallacy. Perhaps even more remarkably, when Miller and Sanjurjo re-analyze the GVT's dataset with their own statistical measures, they find "that the hot hand exists in GVT's dataset as well".
The field experiment was done with a group of expert shooters---Spanish semi-professional basketball players. Miller and Sanjurjo invited players to participate in a study of basketball shooting with financial incentives. They were on average 24 years old and had roughly 14 years of experience in competitive basketball leagues. The experiment was carried out on their home court.
Two phases of the experiment were conducted six months apart. Phase One's goal was to test whether hot hand was noticeable in individual players and whether hot hand was an average effect among all the players.
In Phase Two, Miller and Sanjurjo checked whether players with a hot hand in Phase One had a hot hand again in Phase Two and whether an average from Phase One had a chance of reoccurring.
Miller and Sanjurjo find substantial hot hand effects in both phases. Also, they find that Phase One hot hand performance predicted Phase Two hot hand performance. Further, in a separate questionnaire, players' ranking of shooters from hottest to coldest, based only on experience from practice and games, predicted hot hand performance in the shooting task. This indicates that not only are players correct in believing in the hot hand, but they can actually detect it.
Using a novel empirical strategy to study the best available data (both from their own experiment, the original 1985 GVT study, and a lesser known earlier study), Miller and Sanjurjo find that contrary to nearly 30 years of research, substantial hot hand shooting really does exist.