Statistical Study Sheds Light on Variation in Basketball Players' Performance
Two researchers, Martí Casals of the UIC's Biostatistics Department and the Biomedical Research Centre Network for Epidemiology and Public Health (CIBERESP) and José A. Martínez of the Universidad Politécnica de Cartagena (UPCT), have published a study in the International Journal of Performance Analysis in Sport identifying the variables that affect player performance in basketball.
Their research is amongst the first studies on basketball to use real, quantifiable data to get to grips with variation in player performance in matches.
By employing mixed models and taking into account the effect of the number of minutes played and the offensive plays each player is involved in, the study concludes that the main factor determining performance is the differing potential of opposing teams, whereas the role of other variables is insignificant or largely irrelevant. The study’s results also demonstrate that, despite the existence of extremely hard-to-measure player-specific variables, player performance is broadly stable on a game-to-game basis.
In the study, Casals and Martínez analysed the drivers that affect the performance of NBA basketball players. They did so by focusing on players who were involved and played at least five minutes in every match in the regular season, i.e. players that coaches rely on and who should therefore have the physical and mental attributes required to compete at the top level of the sport.
The pair’s analysis deployed two variables to measure player performance: points scored and “win score”, an indicator calculated by tallying a player’s positive and negative contributions, in a similar way to the valoración metric used in Spain's ACB Endesa basketball league. The following factors were also taken into account: a player’s team, which division and conference it plays in, the period of the season in question, home-court advantage, the quality of the two teams facing off, days off, whether or not a player is a starter, hot streaks, the idea of getting “revenge” on certain teams, a player’s salary compared to the rest of his team, whether or not the team is in contention for the playoffs, a player’s position and age, contractual conditions, minutes played, the number of offensive plays a player makes, and other player-specific factors that cannot be observed directly.
Once the main variables were accounted for, the study found that random factors play the biggest role in determining the varying effectiveness of players such as Luol Deng and Shane Battier, the former of whom substantially outscores the latter. This highlights players’ extraordinary physical and mental resilience, as they are able to perform at a more or less consistent level (within the scope of their relative ability) throughout the games in the regular season, irrespective of the different situations they face in each match.
Most basketball experts and fans typically measure player contribution based on the number of points scored. As for win score, the other metric considered in this study – which offers a more in-depth breakdown of a player’s contribution or attributes – it was found that, besides the relevant variables outlined above, another important factor is age, which particularly affects centres: performance tends to drop amongst veterans in this position. However, as the authors note, their models offer a more reliable explanation for points scored than matches won, thereby making the ability to devise a more comprehensive player performance metric somewhat more challenging.