Faculty of Sports and Health Science, Fukuoka University, Fukuoka, Japan.
United Graduate School of Education, Tokyo Gakugei University, Tokyo, Japan.
PLoS One. 2022 Dec 2;17(12):e0278651. doi: 10.1371/journal.pone.0278651. eCollection 2022.
The present study aimed to establish a more robust, reliable statistical model of touchdown times based on the data of elite 110 m hurdlers to precisely predict performance based on touchdown times. We obtained 151 data (race time: 13.65 ± 0.33 s, range of race time: 12.91 s- 14.47 s) from several previous studies. Regression equations were developed to predict each touchdown time (times from the start signal to the instants of the leading leg landing after clearing 1st to 10th hurdles) from the race time. To avoid overtraining for each regression equation, data were split into training and testing data sets in accordance with a leave-one-out cross-validation. From the results of cross-validation, the agreement and generalization were compared between the present study model and the existing model. As a result, the proposed predictive equations showed a good agreement and generalization (R2 = 0.527-0.981, MSE = 0.0015-0.0028, MAE = 0.019-0.033) compared to that of existing equations (R2 = 0.481-0.979, MSE = 0.0017-0.0039, MAE = 0.034-0.063). Therefore, it can be assumed that the proposed predictive equations are a more robust, reliable model than the existing model. The touchdown times needed for coaches and elite hurdlers to set their target records will be accurately understood using the model of this study. Therefore, this study model would help to improve training interventions and race evaluations.
本研究旨在基于优秀 110 米栏运动员的数据建立一个更稳健、可靠的触地时间统计模型，以便根据触地时间精确预测成绩。我们从之前的几项研究中获得了 151 组数据（比赛时间：13.65±0.33 秒，比赛时间范围：12.91 秒-14.47 秒）。为了从比赛时间预测每个触地时间（从起跑信号到跨过第 1 到第 10 个栏架后领先腿着地的瞬间），我们建立了回归方程。为了避免每个回归方程过拟合，数据按照留一交叉验证的方式分为训练集和测试集。通过交叉验证的结果，比较了本研究模型和现有模型的一致性和泛化性。结果表明，与现有模型相比，提出的预测方程具有较好的一致性和泛化性（R2=0.527-0.981，MSE=0.0015-0.0028，MAE=0.019-0.033）（R2=0.481-0.979，MSE=0.0017-0.0039，MAE=0.034-0.063）。因此，可以认为与现有模型相比，提出的预测方程是一个更稳健、可靠的模型。本研究的模型可以帮助教练和优秀的跨栏运动员准确地了解实现目标记录所需的触地时间。因此，本研究模型将有助于提高训练干预和比赛评估。
