Romann Michael, Javet Marie, Hernandez Julia, Heyer Louis, Trösch Severin, Cobley Stephen, Born Dennis-Peter
Department of Elite Sport, Swiss Federal Institute of Sport Magglingen, Magglingen, Switzerland.
Faculty of Science and Medicine, University of Fribourg, Fribourg, Switzerland.
Front Sports Act Living. 2024 Dec 23;6:1491064. doi: 10.3389/fspor.2024.1491064. eCollection 2024.
Longitudinal performance tracking in sports science is crucial for accurate talent identification and prognostic prediction of future performance. However, traditional methods often struggle with the complexities of unbalanced datasets and inconsistent repeated measures.
This study aimed to analyze the longitudinal performance development of female 60 m sprint runners using linear mixed effects models (LMM). We sought to generate a practical tool for coaches and researchers to establish benchmarks and predict performance development.
We analyzed 41,123 race results from 8,732 female 60 m track sprinters aged 6-15 years, collected from the Swiss Athletics online database between 2006 and 2021. Only season-best times per athlete and only athletes with at least 3 season-best times in their career were included. LMM was used to generate performance trajectories, benchmarks, and individual predictions. A practical software tool was developed and made available to allow individual performance prediction based on race times from previous seasons. In addition, classic empirical percentile curves were constructed using the Lambda-Mu-Sigma (LMS) method.
LMM handled the dataset's complexities, producing robust longitudinal performance trajectories. Compared to empirical percentiles generated using the LMS method, which provided a retrospective view of performance development, the mixed model approach identified individualized longitudinal performance developments and estimated predictions of future performance. The best-fitting model included log-transformed chronological age (CA) as a fixed effect and random intercepts and slopes for each athlete. This model explained 59% of the variance through fixed effects (marginal R) and 93% through combined fixed and random effects (conditional R).
LMM provided longitudinal sport performance data, enabling the establishment of performance benchmarking and prediction of future performance. The software tool can assist coaches in setting realistic training goals and identifying promising athletes.
在体育科学中，纵向表现追踪对于准确识别人才和预测未来表现的预后至关重要。然而，传统方法常常难以应对不平衡数据集和不一致重复测量的复杂性。
本研究旨在使用线性混合效应模型（LMM）分析60米短跑女运动员的纵向表现发展。我们试图为教练和研究人员生成一个实用工具，以建立基准并预测表现发展。
我们分析了2006年至2021年间从瑞士田径在线数据库收集的8732名6至15岁的60米径赛女短跑运动员的41123次比赛成绩。仅纳入每位运动员的赛季最佳成绩，且仅纳入职业生涯中至少有3次赛季最佳成绩的运动员。使用LMM生成表现轨迹、基准和个体预测。开发了一个实用的软件工具，可根据前几个赛季的比赛成绩进行个体表现预测。此外，使用Lambda-Mu-Sigma（LMS）方法构建了经典的经验百分位数曲线。
LMM处理了数据集的复杂性，生成了稳健的纵向表现轨迹。与使用LMS方法生成的经验百分位数相比，后者提供了表现发展的回顾性视图，混合模型方法确定了个性化的纵向表现发展并估计了未来表现的预测。最佳拟合模型包括对数转换的实足年龄（CA）作为固定效应以及每位运动员的随机截距和斜率。该模型通过固定效应解释了59%的方差（边际R），通过固定和随机效应的组合解释了93%的方差（条件R）。
LMM提供了纵向运动表现数据，能够建立表现基准并预测未来表现。该软件工具可协助教练设定现实的训练目标并识别有潜力的运动员。
