Has the NBA regular season gotten less exciting? Part 2: The impact of parity (or lack thereof)
Introduction
This is a continuation of a previous post. Using inpredictable’s, excitement, and tension data from the 1996-97 to 2024-25 seasons, I want to explore the question: Has the NBA regular season gotten less exciting?
Last time, we made a cursory attempt at comparing excitement and tension across seasons by graphing the number of games in the top 100 and top 500 of excitement and tension. Based on these two graphs, we have a signal that perhaps tension has decreased, while excitement has increased. This potential change might be attributed to how the NBA has shifted to a faster pace, relaxed defensive rules, and an emphasis on three point shooting. Leads can balloon quickly (which would reduce tension), and disappear just as quickly, leading to more comebacks (and therefore hire excitement) and closer games. While hypothesizing why this signal exists, another potential factor I thought of which could affect tension and excitement across an entire season, would be parity. Theoretically, if teams were more evenly matched (reflected by closer standings), games might be closer, resulting in higher tension, and perhaps higher excitement, due to their correlation. This blog post will attempt to answer how much does parity affect a season’s daily median excitement and tension?
Measuring Parity
The Gini Coefficient is a statistical measure that has commonly be used to estimate the parity of a sports league via season standings. It’s origins are in economics, where this measure is used to evaluate inequality among income levels. To explain it simply, a Gini coefficient of 0 means perfect inequality, where all wealth would be equal. For the NBA, that means every team would be 41-41. A Gini of 1 would mean that one person holds all the wealth(a basketball equivalent doesn’t exist, since it’s impossible for one team to have all the wins). Basically, a higher Gini coefficient means less parity in a season.
Gini coefficient can be calculated as: \(\dfrac{1}{2} \cdot \text{Relative Mean Difference} = \dfrac{1}{2} \cdot \dfrac{\text{Mean Difference}}{\text{Mean Win Percentage}}\)
In our context, let \(w_t\) be the win percentage for team \(t\). We can calculate the Gini Coefficient as:
\[\dfrac{1}{2} \cdot \dfrac{\sum_{t=1}^n \| w_t-\overline{w} \|}{\overline{w}},\]where \(\overline{w}\) represents the average win percentage and \(n\) is the number of teams in that season.
Using this, we have the Gini’s for each season sorted:
| Original | Sorted | |||||||
|---|---|---|---|---|---|---|---|---|
| Season | Gini | Best Team | Worst Team | Season | Gini | Best Team | Worst Team | |
| 1996 | 0.161 | CHI (69 - 13) | VAN (14 - 68) | 1996 | 0.161 | CHI (69 - 13) | VAN (14 - 68) | |
| 1997 | 0.153 | CHI (62 - 20) | DEN (11 - 71) | 1997 | 0.153 | CHI (62 - 20) | DEN (11 - 71) | |
| 1998 | 0.128 | SAS (37 - 13) | VAN (8 - 42) | 2008 | 0.143 | CLE (66 - 16) | SAC (17 - 65) | |
| 1999 | 0.13 | LAL (67 - 15) | LAC (15 - 67) | 2007 | 0.141 | BOS (66 - 16) | MIA (15 - 67) | |
| 2000 | 0.132 | SAS (58 - 24) | CHI (15 - 67) | 2023 | 0.137 | BOS (64 - 18) | DET (14 - 68) | |
| 2001 | 0.111 | SAC (61 - 21) | GSW (21 - 61) | 2009 | 0.137 | CLE (61 - 21) | NJN (12 - 70) | |
| 2002 | 0.115 | SAS (60 - 22) | DEN (17 - 65) | 2019 | 0.134 | MIL (56 - 17) | GSW (15 - 50) | |
| 2003 | 0.108 | IND (61 - 21) | ORL (21 - 61) | 2013 | 0.134 | SAS (62 - 20) | MIL (15 - 67) | |
| 2004 | 0.124 | PHX (62 - 20) | ATL (13 - 69) | 2014 | 0.133 | GSW (67 - 15) | MIN (16 - 66) | |
| 2005 | 0.106 | DET (64 - 18) | POR (21 - 61) | 2010 | 0.132 | CHI (62 - 20) | MIN (17 - 65) | |
| 2006 | 0.104 | DAL (67 - 15) | MEM (22 - 60) | 2000 | 0.132 | SAS (58 - 24) | CHI (15 - 67) | |
| 2007 | 0.141 | BOS (66 - 16) | MIA (15 - 67) | 2012 | 0.13 | MIA (66 - 16) | ORL (20 - 62) | |
| 2008 | 0.143 | CLE (66 - 16) | SAC (17 - 65) | 1999 | 0.13 | LAL (67 - 15) | LAC (15 - 67) | |
| 2009 | 0.137 | CLE (61 - 21) | NJN (12 - 70) | 2024 | 0.129 | OKC (68 - 14) | UTA (17 - 65) | |
| 2010 | 0.132 | CHI (62 - 20) | MIN (17 - 65) | 1998 | 0.128 | SAS (37 - 13) | VAN (8 - 42) | |
| 2011 | 0.128 | CHI (50 - 16) | CHA (7 - 59) | 2011 | 0.128 | CHI (50 - 16) | CHA (7 - 59) | |
| 2012 | 0.13 | MIA (66 - 16) | ORL (20 - 62) | 2015 | 0.127 | GSW (73 - 9) | PHI (10 - 72) | |
| 2013 | 0.134 | SAS (62 - 20) | MIL (15 - 67) | 2017 | 0.126 | HOU (65 - 17) | PHX (21 - 61) | |
| 2014 | 0.133 | GSW (67 - 15) | MIN (16 - 66) | 2004 | 0.124 | PHX (62 - 20) | ATL (13 - 69) | |
| 2015 | 0.127 | GSW (73 - 9) | PHI (10 - 72) | 2018 | 0.12 | MIL (60 - 22) | NYK (17 - 65) | |
| 2016 | 0.107 | GSW (67 - 15) | BKN (20 - 62) | 2021 | 0.119 | PHX (64 - 18) | HOU (20 - 62) | |
| 2017 | 0.126 | HOU (65 - 17) | PHX (21 - 61) | 2020 | 0.116 | UTA (52 - 20) | HOU (17 - 55) | |
| 2018 | 0.12 | MIL (60 - 22) | NYK (17 - 65) | 2002 | 0.115 | SAS (60 - 22) | DEN (17 - 65) | |
| 2019 | 0.134 | MIL (56 - 17) | GSW (15 - 50) | 2001 | 0.111 | SAC (61 - 21) | GSW (21 - 61) | |
| 2020 | 0.116 | UTA (52 - 20) | HOU (17 - 55) | 2003 | 0.108 | IND (61 - 21) | ORL (21 - 61) | |
| 2021 | 0.119 | PHX (64 - 18) | HOU (20 - 62) | 2016 | 0.107 | GSW (67 - 15) | BKN (20 - 62) | |
| 2022 | 0.091 | MIL (58 - 24) | DET (17 - 65) | 2005 | 0.106 | DET (64 - 18) | POR (21 - 61) | |
| 2023 | 0.137 | BOS (64 - 18) | DET (14 - 68) | 2006 | 0.104 | DAL (67 - 15) | MEM (22 - 60) | |
| 2024 | 0.129 | OKC (68 - 14) | UTA (17 - 65) | 2022 | 0.091 | MIL (58 - 24) | DET (17 - 65) | |
| 2024 | 0.129 | OKC (68 - 14) | UTA (17 - 65) | 2022 | 0.091 | MIL (58 - 24) | DET (17 - 65) |
Additionally, we can compare the closest and least close seasons in terms of standings in our data:
In the 2022-23 season, the team standings are extremely close, with only 6 teams eclipsing 50 wins, and 14 teams between 40 and 50 wins. The Giannis led Bucks placed first with a record of 58-24 while the Pistons finished last at 17-65.
Meanwhile, the 1997-97 was the least close season, as Michael Jordan and his Bulls led the league with a record of 69-13. 3 teams had over 60 wins, and 7 had less than 30 wins (compared to only 3 in 2022-23). Adding to the gap in competitiveness, the Celtics and Grizzlies finished at 15-67 and 14-68, respectively.
Parity and Excitement/Tension
Plotting a regression of each season’s Gini index with median daily excitement and tension, we can see a negative relationship between Gini and median daily excitement, meaning that excitement is lower in seasons where team standings are further separated. Under this model, we can expect that the difference in median daily excitement between the 2022-23 and the 1996-97 season to be approximately 1, although the actual difference was almost 2.5. 2019 does stand out as an outlier, most likely due to the bubble season. Only top teams were invited to compete, with every game being meaningful for playoff hopes and seeding. As a result, daily median excitement was higher, and the standings were also skewed, which affects the Gini coefficient.
Removing the bubble shows a slightly stronger relationship between Gini coefficient and median daily excitement:
For tension, there is a lack of a clear relationship. In fact, the 1996-97 season, which had less parity according to the Gini coefficient, actually surpasses the 2022-23 season in median daily tension.
Here, we arrange the seasosns in increasing order of parity and further examine the relationship between daily median excitement and parity. There is no clear and straightforward increasing trend, but we can see that median daily excitement, as well as the upper quartiles and maximums tend to increase.
Conclusion
I theorized initially that more parity might improve median daily excitement and daily tensions, using the heuristic that as viewers, we tend to enjoy watching more evenly matched teams, and enjoying the uncertainty of the outcome. By using the Gini coefficient as a measure for parity, we can see that as teams are further apart in standings, median daily excitement decreases, whereas the relationship between parity and tension is unclear (or rather, undetectable given our current data).
While we did show a negative correlation between less balance and excitement, the quantitative difference is hard to intepret. The regression shows a difference of 1 (1.5 if you remove the outlier bubble season) in median daily excitement, but we are unable to claim that is a meaningful difference in our perceptions of how exciting the season is. Additionally, previous discussed limitations of our metrics still apply. In particular, even if the median daily excitement has decreased, our tendency to remember the more exciting games may alter our perceptions, minimizing the relationship with parity with the daily median excitement.