Ranking the greatest NBA players : an analytics analysis

Loading...
Thumbnail Image
Authors
Mertz, Jeremy
Advisor
Judge, Larry
Issue Date
2015-07
Keyword
Degree
Thesis (B.?)
Department
Honors College
Other Identifiers
Abstract

The purpose of this investigation was to present a statistical model to help rank top National Basketball Association (NBA) players of all time. As the sport of basketball evolves, the debate on who is the greatest player of all-time in the NBA never seems to reach consensus. This ongoing debate can sometimes become emotional and personal, leading to arguments and in extreme cases resulting in violence and subsequent arrest. Creating a statistical model to rank players may also help coaches determine important variables for player development and aid in future approaches to the game via key data-driven performance indicators. However, computing this type of model is extremely difficult due to the many individual player statistics and achievements to consider, as well as the impact of changes to the game over time on individual player performance analysis. This study used linear regression to create an accurate model for the top 150 player rankings. The variables computed included: points per game, rebounds per game, assists per game, win shares per 48 minutes, and number of NBA championships won. The results revealed that points per game, rebounds per game, assists per game, and NBA championships were all necessary for an accurate model and win shares per 48 minutes were not significant. Any attempts to simplify the model proved ineffective; therefore the four significant variables explained 53% of the variation in ranking. The results of the present study indicate that winning championships is not necessary to be ranked as one of the all-time greats. This model may be appropriate to comparable basketball leagues and may be modified to use in other sports.