PURPOSE: To apply a new method to identify, classify, and follow up young swimmers based on their performance and its determinant factors over a season and analyze the swimmers' stability over a competitive season with that method. METHODS: Fifteen boys and 18 girls (11.8±0.7 y) part of a national talent-identification scheme were evaluated at 3 different moments of a competitive season. Performance (ie, official 100-m freestyle race time), arm span, chest perimeter, stroke length, swimming velocity, speed fluctuation, coefficient of active drag, propelling efficiency, and stroke index were selected as variables. Hierarchical and k-means cluster analysis were computed. RESULTS: Data suggested a 3-cluster solution, splitting the swimmers according to their performance in all 3 moments. Cluster 1 was related to better performances (talented swimmers), cluster 2 to poor performances (nonproficient swimmers), and cluster 3 to average performance (proficient swimmers) in all moments. Stepwise discriminant analysis revealed that 100%, 94%, and 85% of original groups were correctly classified for the 1st, 2nd, and 3rd evaluation moments, respectively (0.11≤Λ≤0.80; 5.64≤χ2≤63.40; 0.001<P≤.68). Membership of clusters was moderately stable over the season (stability range 46.1-75% for the 2 clusters with most subjects). CONCLUSION: Cluster stability is a feasible, comprehensive, and informative method to gain insight into changes in performance and its determinant factors in young swimmers. Talented swimmers were characterized by anthropometrics and kinematic features.
PURPOSE: To apply a new method to identify, classify, and follow up young swimmers based on their performance and its determinant factors over a season and analyze the swimmers' stability over a competitive season with that method. METHODS: Fifteen boys and 18 girls (11.8±0.7 y) part of a national talent-identification scheme were evaluated at 3 different moments of a competitive season. Performance (ie, official 100-m freestyle race time), arm span, chest perimeter, stroke length, swimming velocity, speed fluctuation, coefficient of active drag, propelling efficiency, and stroke index were selected as variables. Hierarchical and k-means cluster analysis were computed. RESULTS: Data suggested a 3-cluster solution, splitting the swimmers according to their performance in all 3 moments. Cluster 1 was related to better performances (talented swimmers), cluster 2 to poor performances (nonproficient swimmers), and cluster 3 to average performance (proficient swimmers) in all moments. Stepwise discriminant analysis revealed that 100%, 94%, and 85% of original groups were correctly classified for the 1st, 2nd, and 3rd evaluation moments, respectively (0.11≤Λ≤0.80; 5.64≤χ2≤63.40; 0.001<P≤.68). Membership of clusters was moderately stable over the season (stability range 46.1-75% for the 2 clusters with most subjects). CONCLUSION: Cluster stability is a feasible, comprehensive, and informative method to gain insight into changes in performance and its determinant factors in young swimmers. Talented swimmers were characterized by anthropometrics and kinematic features.