Is Stephen Curry really a guard? A new perspective on player typologies using functional data analysis
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914231843028992 |
|---|---|
| author | Golovkine, Steven Gunning, Edward |
| author_facet | Golovkine, Steven Gunning, Edward |
| contents | We present a novel representation of NBA players' shooting patterns based on Functional Data Analysis (FDA). Each player's charts of made and missed shots are treated as smooth functional data defined over a two-dimensional domain corresponding to the offensive half-court. This continuous representation enables a parsimonious multivariate functional principal components analysis (MFPCA) decomposition, producing a set of common principal component functions that capture the primary modes of variability in shooting patterns, along with player-specific scores that quantify individual deviations from the average behavior. We first interpret the principal component functions to characterize the main sources of variation in shooting tendencies. We then apply $k$-medoids clustering to the principal component scores to construct a data-driven taxonomy of players. Comparing our empirical clusters to conventional NBA position labels reveals low agreement, suggesting that our shooting-pattern representation might capture aspects of playing style not fully reflected in official designations. The proposed methodology provides a flexible, interpretable, and continuous framework for analyzing player tendencies, with potential applications in coaching, scouting, and historical player or match comparisons. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_21761 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Is Stephen Curry really a guard? A new perspective on player typologies using functional data analysis Golovkine, Steven Gunning, Edward Applications We present a novel representation of NBA players' shooting patterns based on Functional Data Analysis (FDA). Each player's charts of made and missed shots are treated as smooth functional data defined over a two-dimensional domain corresponding to the offensive half-court. This continuous representation enables a parsimonious multivariate functional principal components analysis (MFPCA) decomposition, producing a set of common principal component functions that capture the primary modes of variability in shooting patterns, along with player-specific scores that quantify individual deviations from the average behavior. We first interpret the principal component functions to characterize the main sources of variation in shooting tendencies. We then apply $k$-medoids clustering to the principal component scores to construct a data-driven taxonomy of players. Comparing our empirical clusters to conventional NBA position labels reveals low agreement, suggesting that our shooting-pattern representation might capture aspects of playing style not fully reflected in official designations. The proposed methodology provides a flexible, interpretable, and continuous framework for analyzing player tendencies, with potential applications in coaching, scouting, and historical player or match comparisons. |
| title | Is Stephen Curry really a guard? A new perspective on player typologies using functional data analysis |
| topic | Applications |
| url | https://arxiv.org/abs/2504.21761 |