Shotput download professional#
We apply our model to an extensive real world data set on performance data of professional shot put athletes recorded at elite competitions.Ĭontemporary research has shown that only a small proportion of high achieving young athletes continue to become successful senior athletes. Further, the model allows for the prediction of athletes’ performance in future sport seasons. The resulting model provides an accurate description of the performance trajectories and helps specifying both the intra- and inter-seasonal variability of measurements. We rely both on a smooth functional contribution and on a linear mixed effect model with heteroskedastic errors to represent the athlete-specific trajectories.
In this work, we propose a hierarchical Bayesian model for describing and predicting the evolution of performance over time for shot put athletes. Moreover, fast evolving data gathering technologies require up to date modelling techniques that adapt to the distinctive features of sports data. In particular, the interest in understanding and predicting an athlete’s performance throughout his/her career is motivated by the need to evaluate the efficacy of training programs, anticipate fatigue to prevent injuries and detect unexpected of disproportionate increases in performance that might be indicative of doping. The use of statistical methods in sport analytics has gained a rapidly growing interest over the last decade, and nowadays is common practice. Thus, an athlete performance passport will potentially improve the effectiveness of in-competition anti-doping testing, where currently only placed athletes and a number of randomly selected others, are subjected to. The ultimate application of this type of modeling approach would be that when the projected credible level of performance is exceeded by an athlete (e.g., Figure 3-athlete 53 and 281 data points above the 95% credible interval), they would be identified for target testing via the ABP system. Therefore, this retrospective data modeling provides an indication that the effects of doping can be identified from longitudinal athlete performance profiles, and that it could potentially be used as a tool to estimate theoretical future credible performances for a given athlete. Nonetheless, our current model appears to be able to detect differences between the two populations in the early phase of the competitive year, with about 1 meter difference in performance between doped and non-doped athletes (see Figure 4). Moreover, factors such as individual athlete seasonal training and competition patterns will affect their individual competition results, independently of any doping related effect and so must be accounted for in any longitudinal model of performance. Further, the model does not currently accommodate longitudinal covariates potentially affecting performance, for example, as shown in Figure 5, athlete aging has a clear demonstrable effect on shot put throwing distance. Thus, our dataset shows more local, short-range variability, which the current version of the model cannot adequately represent. Instead, the shot put dataset has measurements often collected just a few days apart from each other on each athlete, and for a potentially long number of years. The Bayesian latent factor regression methodology was originally developed for very sparse longitudinal data ( Montagna et al., 2012) with the purpose of capturing a global trend in subject-specific trajectories.
of potential issues with the data, the model in its current formulation suffers from some limitations.
0 kommentar(er)
