Human performance after success and failure: evidence from the NBA

Abstract

Based on play-by-play statistics of 10 NBA seasons, we identify players who are personally responsible for the overtime by taking the last shot of the game. Players who miss the shot when the game is tied (crucial failure), perform better in overtime than in the last quarter but not significantly differently to their game and season averages. Players who score the equalizer (crucial success), perform substantially worse in overtime compared to the last quarter as well as compared to their game and season averages. We discuss psychological explanations and argue that our findings can to some extent be transferred to behavior in team production.

Keywords:

• Success
• failure
• human performance
• psychological pressure
• overconfidence
• hot hand fallacy
• natural experiment
• basketball
• NBA

JEL CODES:

• C93

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 We thank two anonymous referees for encouraging us to do so.

2 Our dataset can be downloaded here: https://doi.org/10.6084/m9.figshare.6057755

3 Our data is not normally distributed – Shapiro-Wilk tests show that the data violates the assumption of normality with a p-value of <0.001. Parametric tests could nevertheless have been used because we have up to 2444 observations (of 377 players). However, for comparisons between the three groups further requirements for ANOVA are violated: Levene tests significantly reject the null hypothesis of variance homogeneity between groups and the number of observations varies by group. Therefore, we decided to use nonparametric tests in this section: Kruskal-Wallis tests to compare our three groups in one period, Mann-Whitney U tests to compare two groups in one period, and Wilcoxon signed rank tests to compare the field goal percentages in different periods within groups.

4 If not explicitly mentioned, we refer to the first overtime.

5 If not explicitly mentioned, we use Wilcoxon signed rank tests in this section

6 When calculating regression to the mean effects (as proposed by Linden, 2013), we find that part of the differences of overtime compared to 4^th quarter performance can be attributed to this phenomenon (Nesselroade et al. 1980). However, comparing overtime performance to game or season averages, the regression to the mean effects are virtually zero.

7 We exclude observations in which a player from our groups is responsible for the second or third overtime.

8 Table 3 uses field goal percentages as the performance indicator. In the appendix, the same specifications are conducted with the effective field goal percentage (Table 3a) and win score per minute (Table 3b).

9 Thus, we have to exclude the control group that we use in section 5.1.

10 Comparing the result to Table 2, we have to calculate the distance of the differences of field goal percentages: (53.16% - 41.02%) – (32.44% - 42.53%) = 22.23 percentage points.

11 We thank two anonymous referees for drawing our attention to these limitations.