The Hot Hand in Actual Game Situations
share
Miller and Sanjurjo [8] showed that the majority of the hot hand studies were subject to streak selection bias, leading to an underestimation of any effect present. In this study, we extend the elegant permutation test recommended by Miller and Sanjurjo [8] to in-game situations, where consecutive shots of a player are not identical. We do so by developing a shot make probability model trained with optical tracking data from SportVU, and resampling the underlying non homogeneous Bernoulli process for the shots taken by a player to build the distribution under the null hypothesis (i.e., independence from the result of previous shots). Our results point to the presence of a sizable hot hand effect within game settings.
READ FULL TEXT