PROFILING OF AGE-RELATED DEVELOPMENT OF METABOLIC PERFORMANCE DETERMINANTS IN ELITE YOUTH SOCCER PLAYERS: CONVENTIONAL STATISTICS AND CONTRIBUTIONS OF MACHINE LEARNING

INTRODUCTION:
Performance diagnostic is an indispensable process for optimizing individual training programs and monitoring an athlete’s development in all sports disciplines. Due to the high impact of biological maturation on performance development of young athletes, systematic and continuous physiological monitoring is even more important in this population than in senior athletes (Barker & Armstrong, 2010). However, data on physiological performance determinants and their time-course changes in young elite athletes is limited and rather complemented with data from untrained children and adolescents. Thus, this cross-sectional study aimed to explore physiological characteristics regarding endurance performance and the longitudinal development of elite youth soccer players based on chronological age and age groups.
METHODS:
221 male elite youth soccer players (17.0±2.9 yrs) performed an incremental step test till exhaustion (start at 2.8 m/s +0.4 m/s every 5 min), to determine physiological determinants of endurance performance such as running velocity at lactate thresholds (vLT1 & vLT2), maximal oxygen uptake (VO2max), oxygen cost of running (CR), and total distance covered (DTot). For each athlete, 1-10 tests were carried out in a period of 4 years (number of tests: U15 = 72, U16 = 136, U17 = 204, U19 = 201, U23 = 188). The data were organized in a “data warehouse” and used for conventional statistics and machine learning, such as clustering and decision trees.
RESULTS:
There was no significant difference between age groups (mean±SD) regarding VO2max (ml/kg/min; U15 = 55.5±3.8, U16 = 55.1±4.4, U17 = 54.6±4.0, U19 = 55.3±4.5, U23 = 53.7±4.1). In contrast, the key indicator of overall treadmill performance (i.e. DTot; m) increased with age: U15 = 4598±855, U16 = 5089±868, U17 = 5241±911, U19 = 5667±890, U23 = 5836±888. At the same time, improved CR (ml/kg/m) with increasing age was found: U15 = 0.241±0.019, U16 = 0.232±0.022, U17 = 0.224±0.017, U19 = 0.216±0.015, U23 = 0.208±0.014. A similar development pattern in these variables was also observed in the intra-individual analyses.
Due to the large number of experimentally available variables an important consideration is the strategy of “dimension reduction” to gain a better understanding of complex key indicators. For this purpose, factor-analyses with varimax-rotation have been used. Two major factors with an explained variance of 68% were extracted, which can be associated with dominant aerobic (52%) and anaerobic energy supply (16%).
CONCLUSION:
Our results indicated an age-related improvement in CR without a change in VO2max resulting in improved endurance performance. Although the positive impact of growth and maturation cannot be ruled out (Ariens et al., 1997), the decreased metabolic demand of running (CR) per se can be associated with performance development in elite youth soccer players. Metabolic profiling and player grouping using machine-learning can help to better understand and improve an athlete’s performance development.