Workload Ratio Assessment in Football: Evaluating Simple and Exponential Moving Averages

Abstract

Introduction: To identify the optimal technique for examining time series data related to the Acute Chronic Workload Ratio (ACWR), correlations between the Simple Moving Average (SMA) and the Exponentially Weighted Moving Average (EWMA) were investigated in this study utilising a decay factor (λ) over a period of 7/28 days. Five GPS metrics were included in our analysis: Total Distance, Accelerations, Decelerations, High Metabolic Load Distance, and Distance in Speed Zones 3+4+5 (>19,9 km/h). These data points were collected from 22 players across 47 days, excluding the first 28 days, for a total of 596 data points per pair [SMA/EWMA]. Methods: Shapiro-Wilk and Kolmogorov-Smirnov normality tests were performed on the SMA and EWMA datasets prior to using the Spearman, Kendall Tau, and Distance Correlation techniques to assess correlations and dependencies between pairings. Using Python and libraries including Pandas, NumPy, Matplotlib, SciPy, Scikit-Learn, Statsmodels, OpenPyXL, Dcor, and IPython.display, the analysis was carried out in Anaconda's Jupyter Notebook. Results and Discussion: Significant departures from the normal distribution were shown by normality tests (p<0.05 for most of the variables). With p-values of 0.00, Spearman analysis showed significant correlations for every pair of variables, ranging from moderate (0.46) to somewhat weak (0.23). Additionally, Kendall's Tau revealed statistically significant correlations (p=0.00) across strengths, ranging from moderate (0.32) to weak (0.16). With values ranging from 0.25 to 0.44, Distance Correlation showed significant connections (p<0.00), while Energy Distance values displayed a range of discrepancies. Interestingly, EWMA frequently displayed values that were marginally lower than SMA, highlighting a significance level of p=0.00. Conclusion: The results show continuous trends and modest to moderate positive correlations between the variables under study. Both SMA and EWMA can be used with the help of distance correlation. EWMA is typically chosen for responsive trend analysis and offering a realistic representation of current conditions in ACWR monitoring due to its emphasis on recent data. The decision between SMA and EWMA, however, may change depending on the coaching needs; in this study, EWMA approaches produced somewhat lower scores than SMA.