An extensive battery of physical assessments is typically employed to evaluate athletes' training progress or identify potential talent. However, many of the tests conducted may essentially measure the same aspect of performance. Thus, the efficiency of these processes could potentially be improved in some cases. Utilising a series of multivariate analyses, this study aimed to extract independent physical predictors of skeleton start performance from a testing battery. Multiple two day testing sessions were undertaken by 14 (nine male, five female) skeleton athletes across a training season. Sessions consisted of flexibility, dry land push track, sprint, countermovement jump, and leg press (strength-power) tests. The analysis resulted in over 30 physical variables. Principal component analysis was firstly conducted to reduce the large number of output variables to separate independent factors. This revealed three underlying factors, which represented sprint ability, lower limb power and strength-power characteristics. The three variables which best represented (most heavily loaded to) each of the factors (unresisted 15 m sprint time, 0 kg jump height, force at peak power during leg press, respectively) were then entered into a stepwise multiple regression analysis. All three variables significantly contributed (P < 0.01) to the prediction (R2 = 0.86, standard error of estimate was 1.52%) of skeleton start performance (sled velocity at 15 m). A regression equation was formulated from the unstandardised ß weights (-1.868, 0.015, and -0.011, for the aforementioned three variables, respectively; constant = 11.53). Finally, a K fold validation technique was adopted to assess model stability (R2 = 0.77 for predicted vs. actual, standard error of estimate = 1.97%). It was, therefore, concluded that only three physical test scores were needed to provide a valid and stable indication of skeleton start ability. These three variables were shown to be independent of one another, and each reflected truly important physical capabilities for skeleton athletes to possess. Thus, the inclusion of these tests in longitudinal monitoring protocols for skeleton athletes is vital. The regression model has clear practical implications when assessing both the progress of current athletes and identifying future talent. Importantly, this study has demonstrated a process through which to isolate the key physical variables underlying performance and thus, the wider application of this approach to other sports is encouraged. The methods adopted in this study can ultimately improve the validity and efficiency of athlete monitoring programmes, which can benefit sports scientists, coaches and athletes alike.
© Copyright 2015 BASES Student Conference - John Moores University, Liverpool 31 Mar 2015 - 1 Apr 2015. Veröffentlicht von Bases. Alle Rechte vorbehalten. / Übersetzt mit https://www.DeepL.com/Translator