SPRING 2022, VOLUME 4

AUTHOR

Kevin Babitz, University of Pennsylvania (Capstone)

ABSTRACT

Common across many domains, and especially prevalent in basketball, the ‘hot hand’ suggests that a person who has experienced a recent period of success has a greater likelihood of future success. In basketball, the player with the hot hand who has recently made a series of baskets is more likely to continue their recent success. The seminal research on the hot hand, identifies it as a fallacy after observing 26 Cornell basketball players’ shooting sequences (Gilovich et al., 1985). The shooting percentages observed after a hot streak should be less than those after a cold streak when holding the probability of success and length of the sequence constant. This work instead looks at how much different what we observe is from what we would expect. We call this expected negative difference between the hot and marginal or cold shooting probability a bias adjustment (Miller and Sanjuro, 2018). We extend the hot hand literature in three ways: we add permutation tests and apply more robust hypothesis testing to the resulting p-values to the Cornell dataset, we demonstrate the effectiveness of a new formula to approximate the bias term in the literature, and we apply the hot hand analysis framework to PGA putting data from the 2019 season.

The results suggest that the Gilovich et al. paper was close to accurate in determining if the hot hand was true, even with their failure to recognize the bias adjustment needed in the analysis. The hot hand appears to be hard to detect statistically in the Cornell basketball player dataset and PGA putting data using the simulation framework discussed in these analyses. Additionally, we have found a relatively accurate approximation of the bias term used in correctly assessing if a player is significantly hot.