運動賽事期間隊伍能力的變化分析

Abstract

在成對比較類型的運動數據分析中，實務工作者和研究者們認知球隊在連續比賽中，由於受傷、團隊心理和團隊進步會導致團隊能力變化。描述分數差或比賽結果最常用的框架主要是對主場隊伍和客場隊伍的能力差做適當的轉換。在這樣的考慮下， 隊伍能力的變化可以在頻率派學者或貝氏的觀點下進一步建模。通過整合這些特點到模型的構造中，我們對團隊能力提出更通用的動態模型。此外， 我們還制定了一些準則來從競爭模型中選出擁有較好季後賽預測力的模型。我們透過美國國家籃球協會 2009 2010 賽季到 2018 2019 賽季的數據來調查提案的實用性。

In paired-comparison sports data analysis, practitioners and researchers have identified the varying abilities of teams due to injuries, team psychology, and team improvement in the course of sequential competitions. The most commonly used framework to describe the score difference or the match outcome is mainly based on an appropriate transformation of the difference in abilities of the home team and the visiting team. Under such consideration, the abilities of teams can be further modelled with dynamic effects in the frequentist or Bayesian perspective. By integrating these features into a model formulation, we propose more general dynamic models for the abilities of teams. In addition, some criteria are developed to select a better predictive model for playoffs among competing models. The practicality of our proposal is also investigated by the data from the 2009-2010 season to the 2018-2019 season of the National Basketball Association.