A few weeks ago, we did a project looking at hockey and identifying if it was a strong-link or whole chain sport. Basically, is it better to have a single all-star or a group of consistent, slightly better-than-average skaters on the rink?
With hockey, we identified it as a whole-link sport; yet, since we’re in March Madness season, we thought it would be fun to apply the same methodology to NBA basketball.
Methodology
- Source and structure a “player-data” and a “team-data” data frame.
- Identify the success metric (Wins)
- Identify input metrics (Strongest, Weakest, Average (minus strongest)
- Run a correlation and significance test
Data Notes
The primary data source can be found here and includes a handy game-score metric. I didn’t run a regression to identify the important metrics (e.g., field goals, 3-pointers, etc.). Instead, I’m going to do a quick check against the strongest player’s individual game score, the average (without the strongest), and the weakest player’s game score.
I was tempted to be lazier and just do an average or total game score, weighed against the input metrics, but instead used season wins as the success metric.
Results
It’s a strong link sport! Basically, the game score of the best player matters more than the remainder of the team’s scores and more than the weakest player’s score. There is a .45 correlation between the game score of the strongest player (P-value < 0.0130), indicating a significant, moderate positive correlation.
Neither the average team (without its superstar) nor the weakest link was significant (P-value <.05).
Sources
NBA – Player Stats – Season 24/25. n.d. Retrieved March 24, 2026. https://www.kaggle.com/datasets/eduardopalmieri/nba-player-stats-season-2425.
Leave a Reply