Deporte y Entrenamiento Research Group, Departamento de Deportes, Facultad de Ciencias de la Actividad Física y del Deporte (INEF), Universidad Politécnica de Madrid, C/Martín Fierro 7, 28040 Madrid, Spain.
Health Sciences Faculty, Universidad San Jorge, Autov A23 km 299, Villanueva de Gállego, 50830 Zaragoza, Spain.
Sensors (Basel). 2023 Jul 27;23(15):6714. doi: 10.3390/s23156714.
Performance analysis based on artificial intelligence together with game-related statistical models aims to provide relevant information before, during and after a competition. Due to the evaluation of handball performance focusing mainly on the result and not on the analysis of the dynamics of the game pace through artificial intelligence, the aim of this study was to design and validate a specific handball instrument based on real-time observational methodology capable of identifying, quantifying, classifying and relating individual and collective tactical behaviours during the game. First, an instrument validation by an expert panel was performed. Ten experts answered a questionnaire regarding the relevance and appropriateness of each variable presented. Subsequently, data were validated by two observers (1.5 and 2 years of handball observational analysis experience) recruited to analyse a Champions League match. Instrument validity showed a high accordance degree among experts (Cohen's kappa index (k) = 0.889). For both automatic and manual variables, a very good intra- ((automatic: Cronbach's alpha (α) = 0.984; intra-class correlation coefficient (ICC) = 0.970; k = 0.917) (manual: α = 0.959; ICC = 0.923; k = 0.858)) and inter-observer ((automatic: α = 0.976; ICC = 0.961; k = 0.874) (manual: α = 0.959; ICC = 0.923; k = 0.831) consistency and reliability was found. These results show a high degree of instrument validity, reliability and accuracy providing handball coaches, analysts, and researchers a novel tool to improve handball performance.
基于人工智能与游戏相关统计模型的性能分析旨在为比赛前、比赛中和比赛后提供相关信息。由于手球表现的评估主要关注结果，而不是通过人工智能对手球比赛节奏的动态进行分析，因此本研究旨在设计和验证一种基于实时观察方法的特定手球仪器，该仪器能够识别、量化、分类和关联比赛中的个人和集体战术行为。首先，通过专家小组进行仪器验证。十位专家回答了一份关于每个变量的相关性和适当性的问卷。随后，由两名观察员（具有 1.5 年和 2 年手球观察分析经验）对一项冠军联赛比赛进行数据分析，对数据进行了验证。仪器验证显示专家之间具有高度一致性（科恩氏kappa 指数（k）= 0.889）。对于自动和手动变量，均具有非常高的内部一致性（自动：Cronbach's alpha（α）= 0.984；组内相关系数（ICC）= 0.970；k = 0.917）（手动：α= 0.959；ICC = 0.923；k = 0.858）和观察者间一致性（自动：α= 0.976；ICC = 0.961；k = 0.874）（手动：α= 0.959；ICC = 0.923；k = 0.831）。这些结果表明仪器具有高度的有效性、可靠性和准确性，为手球教练、分析师和研究人员提供了一种改进手球表现的新工具。
