FIBA 농구 월드컵 팀 실력별 승률 결정 요인에 대한 분위수 회귀 분석

초록

This study aims to analyze the performance characteristics of teams with different win rates and to explore offensive and defensive factors across various tiers using the Quantile Regression (QR) method, based on the official statistics provided by the 2023 FIBA Basketball World Cup. The analysis was differentiated by performance level. The dataset comprises a total of 184 games played by 32 national teams, extracted from the official match records of the 2023 FIBA Basketball World Cup. For data processing, the Python programming language and its pandas library were utilized. All statistical tests were conducted at a 95% confidence level, and the main findings are as follows: First, the QR analysis revealed that paint scores and three-point scores were consistently significant across all quantiles. Free throw scores showed a statistically significant effect in the Weighted Least Squares (WLS) model and were significant in QR at the 25th, 50th, and 75th percentiles. Second, defensive rebounds had a significant effect in the WLS model, and in the QR model, they were significant at the 50th, 75th, and 90th percentiles. In contrast, fouls were significant in both WLS and QR, particularly at the 25th, 50th, and 75th percentiles in the latter. Third, the adjusted R² of the WLS model was 0.477, indicating a relatively low explanatory power. Meanwhile, the QR model exhibited higher explanatory power and statistical significance, with Pseudo R² values ranging from 0.625 to 0.7059 across the quantiles (Q10-Q90). Based on the study’s findings, compared to traditional linear regression models used in prior research, quantile regression provides a more detailed understanding of how independent variables affect the dependent variable. Moreover, it enables a more nuanced analysis of how different teams, located at various percentiles in performance distribution, are influenced by key performance factors.

키워드