Where to go: Computational and visual what-if analyses in soccer
Where to go: Computational and visual what-if analyses in soccer
复制标题
DOI：
10.1080/02640414.2019.1652541
复制
发表时间：
2019-08-12
影响因子：
3.4
通讯作者：
Janetzko, Halldor
中科院分区：
文献类型：
--
作者：
Stein, Manuel;Seebacher, Daniel;Janetzko, Halldor
为了让球队为即将到来的比赛做好准备，专业足球分析师每天观看并手动注释多达三场比赛。当注释匹配时，领域专家试图根据直觉和专业经验识别和改进次优动作。需要手动分析的大量匹配导致繁琐且耗时的过程，并且结果可能是主观的。我们提出了一个自动的方法，实现有效的区域为基础的假设分析足球。我们的系统涵盖了基于区域的错误运动行为的自动检测，以及可能改进的替代运动的自动建议。正如我们所展示的，我们的方法通过加快以前耗时的工作，有效地支持分析师和教练调查比赛。我们使领域专家，包括他们的领域知识的分析过程中，允许互动调整建议改进的运动，以及其对区域控制的影响。我们证明了我们所提出的方法的实用性，通过专家研究与三个领域的专家，一个是主教练从第一个奥地利足球联赛。由于我们的研究结果表明，专家最经常同意建议的球员运动（83%），我们提出的方法提高了足球的分析能力，并支持更有效的分析。
To prepare their teams for upcoming matches, analysts in professional soccer watch and manually annotate up to three matches a day. When annotating matches, domain experts try to identify and improve suboptimal movements based on intuition and professional experience. The high amount of matches needing to be analysed manually result in a tedious and time-consuming process, and results may be subjective. We propose an automatic approach for the realisation of effective region-based what-if analyses in soccer. Our system covers the automatic detection of region-based faulty movement behaviour, as well as the automatic suggestion of possible improved alternative movements. As we show, our approach effectively supports analysts and coaches investigating matches by speeding up previously time-consuming work. We enable domain experts to include their domain knowledge in the analysis process by allowing to interactively adjust suggested improved movement, as well as its implications on region control. We demonstrate the usefulness of our proposed approach via an expert study with three invited domain experts, one being head coach from the first Austrian soccer league. As our results show that experts most often agree with the suggested player movement (83%), our proposed approach enhances the analytical capabilities in soccer and supports a more efficient analysis.