Search

Association football (commonly known as football or soccer) in the modern era places a greater emphasis on collaborating and working together as a team instead of relying solely on individual skills to strategize winning performances. The low-scoring and unpredictable nature of association football makes evaluating team performances challenging. Space creation and space utilization have been discussed in the football world lately. Existing literature evaluates this with on and off-ball runs by players for deceiving defenders to create open spaces. However, the contribution of these team ball movements’ enhanced randomness or chaotic nature to winning performances has yet to be explored. This work proposes a novel entropy-based time-series performance evaluation metric, EDRan, for quantifying this enhanced random nature by analyzing the spatial distribution of game events at regular intervals. Additionally, an unexplored cumulative ball possession matrix is used to quantify randomness. The correlation between the match winner and spatial event distribution randomness at regular intervals was analyzed. The significance of the proposed metric was demonstrated using a generalized linear model (GLM), which achieved an average accuracy of 80% for match-winning performance classification. The GLM p-values and coefficients revealed statistically significant relationships between the extracted temporal features and match-winning performances. Findings further revealed dispersed, highly random event distribution by winning teams during the early phases of the game, implying attacking behavior, followed by a compact, cautious playing style toward the end, suggesting that the game’s first-half performances are more pivotal. Despite the unpredictability of actual scores in association football, the proposed approach effectively captured the differences in performances between stronger and weaker teams with temporal relationships, highlighting its significance as a time-series metric for performance evaluation.

Team sports can be viewed as dynamical systems unfolding in time and thus require tools and approaches congruent to the analysis of dynamical systems. The analysis of the pattern-forming dynamics of player interactions can uncover the clues to underlying tactical behaviour. This study aims to propose quantitative measures of a team's performance derived only using player interactions. Concretely, we segment the data into events ending with a goal attempt, that is, “$Shot$”. Using the acquired sequences of events, we develop a coarse-grain activity model representing a player-to-player interaction network. We derive measures based on information theory and total interaction activity, to demonstrate an association with an attempt to score. In addition, we developed a novel machine learning approach to predict the likelihood of a team making an attempt to score during a segment of the match. Our developed prediction models showed an overall accuracy of 75.2% in predicting the correct segmental outcome from 13 matches in our dataset. The overall predicted winner of a match correlated with the true match outcome in 66.6% of the matches that ended in a result. Furthermore, the algorithm was evaluated on the largest available open collection of soccer logs. The algorithm showed an accuracy of 0.84 in the classification of the 42, 860 segments from 1, 941 matches and correctly predicted the match outcome in 81.9% of matches that ended in a result. The proposed measures of performance offer an insight into the underlying performance characteristics.

Understanding reliability of performance analysis tools is important to ensure match to match comparisons can be undertaken with the knowledge of consistency between coding situations. There are few published studies examining the reliability of commonly used performance analysis tools. The aim of this project was to assess the inter- and intra-rater reliability of the NetballStats application and to make comparisons between live and video-based coding situations. Two ‘coders’ coded eight netball matches using the NetballStats application, coding each match live, then twice from video. Level of agreement was assessed for frequency counts across the variables coded. Results showed that intra-rater agreement was higher than inter-rater agreement and that reliability from video coding is better than from live coding. High frequency events automatically coded by the application and events that are well defined had greater levels of agreement than lower frequency events and subjectively judged events. Live coding situations underrepresent occurrence of events, particularly for high frequency events such as ‘possession’. To ensure reliability between coders, clubs should provide an extensive training program to coders with clear instructions on coding subjective events. Coaches should be aware that live coding underestimates some event types and factor this into their decision making processes.

Studies of training and competition load in sport are usually based on data that represents a sample of a league and or annual training program. These studies sometimes explore important factors that are affected by load, such as training adaptations and injury risk. The generalisability of the conclusions of these studies, can depend on how much load varies between seasons, training phases and teams. The interpretation of previous load studies and the design of future load studies should be influenced by an understanding of how load can vary across seasons, training phases and between teams. The current study compared training loads (session rating of perceived exertion x session duration) between all (8) teams in an elite Netball competition for multiple (2) season phases and (2) seasons. A total of 29,545 records of athlete session training loads were included in the analysis. Linear mixed models identified differences between seasons and training phases (p < .05). There were also differences between teams and a complex set of interactions between these three factors (season, phase, and team) (p < .05). While the absolute value of the training loads reported here are only relevant to elite netball, these results illustrate that when data is sampled from a broader context, the range and variation in load may increase. This highlights the importance of cautiously interpreting and generalisation of findings from load studies that use limited data sets.