Every Second Counts: Using Big Data to Investigate Referee Accuracy in NBA Games (2023)

Undergraduates: Gavin McDermott, Matthew Hettleman, Leah Kozel, Virginia York

Faculty Advisor: Jonathan Jensen
Department: Exercise & Sport Science

---

Researchers have previously studied referee bias across many sports and different conditions and there has been a significant amount of findings about NBA referees specifically. Our focus was on previously-studied independent variables and their effect on referee accuracy. We used big data from official NBA Last Two Minute Reports, totaling more than 16,000 calls from 2017-2022. The model did not show significance in referee bias in making incorrect calls depending on crowd size or whether the call was benefiting the home team. However, our model displayed statistically significant results for two variables: time left in the game and the season in which the game was played. As the amount of time left in the game increases, the probability of a referee making an incorrect call increases. This means within the last two minutes, referees are more likely to make an error toward the two-minute mark than at the buzzer. As the season increases, which signifies later years, the probability of a referee making an incorrect call decreases. We also analyzed that this suggests referees are becoming more accurate year after year. Overall, although these correlations are significant, they are relatively small, and we can be confident in referee performance.

Link to Poster