Recent studies had shown the role of the central "governor" (Noakes 2000) and the importance of specific motor pattern training on performance (Termin 2000). Few studies dealt with performance modelling in order to lead to a better understanding of the kinematics factors which influence performance, and to propose a simulation in advance of the laps time and stroke frequencies corresponding to a predicted performance (Arellano et al. 2001). These researches based on multiple regression analysis failed in this purpose because they didn't integrate several complex aspects of the problem. 1-Several technical expertise clusters correspond to similar performances. 2-The relationships between performances and technical parameters are non-linear. 3-Races represent sequential events, so these different parts are inter-dependant (fast start race dragged slow race end). Methods One hundred female freestyle 227 performances analyses of the last 5 years international events were used. The fourth frequency and laps time were studied using a competition video-computer system analysis. For each race part stroke lengths were computed. 1/ Data were clustered using Kohonen selforganising map (Bauer and Schollhorn 1997). 2/ Instead of taking into account the sequential nature of the race as well as different race part interactions a model was constructed. Data were computed using a fuzzy inference system which is an expert knowledge model functioning using if-so type algorithms. A sub-fuzzy system is characterised by its belonging function. The first two thirds of data were used for learning and the last for prediction. Results Data collected underscore a large heterogeneity in performance and technique. Among the 227 subjects studied, 144 belonged to each of 9 clusters and 73 to intermediate classes. For predicted data, laps time were perfectly fitted (R²=0.87,EM=0.20s) c/mn and correctly for intermediate stroke rates (R²=0.59, EM=2.8c/mn). For a swimmer, while the actual best performance is 56s, table 2 offers a simulation of the way in which a 54.67s performance might be achieved. This model aims to help the coach building specific training contents.
© Copyright 2002 Expertise in Elite sport. 2nd International Days of Sport Sciences, 12.-15. November 2002, INSEP, Paris (France). Veröffentlicht von INSEP. Alle Rechte vorbehalten.