Changes in landscapes of passing opportunities along a set of competitive football matches

As a football match unfolds, the ball carrier aims to perform passes that outplay as many opponents as possible. This means that a potential receiver will have less opposing players between him/her and the opposing goal than the current ball carrier. This outplay principle served to characterize passing opportunities into: i) penetrative passes, ii) support passes, and iii) backwards passes. Data from five competitive football matches were used to create heatmaps that illustrate a landscape model of passing opportunities for each type of pass. Due to the specificity of player’s interactive behavior, results showed heatmaps with a variety of patterns. Using the Earth Mover’s Distance is was possible to identify difference between the heatmaps, for instance the fifth match was dissimilar to the other four. Moreover, the mean time of passing opportunities was calculated and compared across matches and different types of passes. Results display that penetrative passes were available over shorter periods of time than backward passes that were available shorter than support passes. Also passing opportunities were available for longer in the first match compared to the other four. This model could provide teams’ technical staff with a tool to study the dynamics of passing opportunities that occur in football matches.

The data presented corresponds to positional data of the 22 players at all moment in the field as well as the events data.

随着足球赛事的推进，持球球员的目标在于送出能够尽可能突破对方多人防守的传球。这意味着，相较于当前持球球员，潜在接球球员与对方球门之间的对方防守球员数量应更少。基于这一突破原则，传球机会可被划分为三类：i) 穿透传球（penetrative passes）、ii) 支援传球（support passes）以及 iii) 回传球（backwards passes）。本研究选取五场竞技足球比赛的数据，绘制热力图以呈现各类传球的传球机会全景分布模型。鉴于球员互动行为具有特异性，最终得到的热力图呈现出多样化的模式。借助推土机距离（Earth Mover’s Distance），可有效识别不同热力图间的差异——例如第五场比赛与其余四场比赛存在显著区别。此外，本研究还计算了各类传球机会的平均持续时长，并在不同赛事及不同传球类型间开展对比分析。结果表明：穿透传球的可用时长短于回传球，而回传球的可用时长又短于支援传球；同时首场比赛的传球机会可用时长，相较于其余四场更为持久。该模型可为球队技术团队提供分析足球比赛中传球机会动态变化的专业工具。

本次呈现的数据包含全场22名球员在每一刻的位置数据，以及赛事事件数据。