ECSS Paris 2023: CP-AP09
INTRODUCTION: In recent years, the integration of Artificial Intelligence (AI), the Internet of Things (IoT), and sensor technologies has brought about a paradigm shift in sports analytics. In team sports such as soccer, the use of wearable devices equipped with GPS and accelerometers has enabled the precise measurement of performance metrics, including distance covered and movement intensity. The analysis of such tracking data has contributed to multifaceted insights, particularly in injury prevention and prediction. Against this backdrop, the present study utilizes network analysis to visualize the offensive and defensive tactics of Japanese professional soccer team. This approach aims to highlight the characteristics of team strategy while quantitatively elucidating the interactions and individual contributions of the players. METHODS: In this study, top-tier Japanese professional soccer league matches were analyzed using wearable accelerometers. A unique algorithm was developed to evaluate player coordination networks by processing acceleration data and deriving relevant random variables. By calculating the entropy of player movements through statistical analysis, we established an index for inter-player synchronization. This metric elucidates the characteristics of team performance in both offensive and defensive phases during a match. Specifically, this research focuses on quantifying three-player coordination, a fundamental unit of tactical teamwork in soccer. RESULTS: Analyses of official matches between closely ranked teams revealed that player synchronization, measured via phase difference entropy, was significantly higher in matches where team tactics were effectively executed. Furthermore, a marked difference in the degree of interplay was observed between periods of successful organized attacks compared to less effective periods. Specifically, player synchronization during successful organized attacks was approximately 30% higher than the overall match average (p < .05). Notably, diminished coordination was associated with erratic, high-intensity compensatory movements, indicating that network evaluation could serve as a vital metric for monitoring mechanical stress and managing injury risk within team structures. CONCLUSION: This research demonstrates that quantifying inter-player coordination allows for objective performance evaluation, the visualization of tactical quality, and the assessment of injury risks. Integrating these network metrics with individual-level data will facilitate a more holistic understanding of team performance. While detailed analysis of all tactical variations remains a challenge, future research should focus on collaborating with professional teams to identify and visualize key performance indicators (KPIs) in offensive and defensive phases. By optimizing high-quality collective movements through real-time feedback, this approach has the potential to provide transformative value not only to soccer but to the broader field of sports science.
Read CV takuya magomeECSS Paris 2023: CP-AP09
INTRODUCTION: The integration of data analytics in elite sports has become crucial for performance analysis. However, the data engineering process—the transformation of raw data into actionable insights—remains underexplored. This study proposes an engineering-driven framework for converting open-source football data into decision support. The contribution lies not in introducing new performance metrics, but in developing a methodologically rigorous pipeline for data processing. METHODS: The pipeline consists of several stages: Data Acquisition: Open-access football data are gathered, ensuring compliance with legal and ethical standards. Data Cleaning: Raw data are cleaned to address issues such as missing values and noise, ensuring reliability. Feature Engineering: Relevant performance indicators are extracted for analysis. The framework incorporates Descriptive, Diagnostic, Predictive, and Prescriptive Analytics. Each layer builds on the previous one, progressively providing more detailed and actionable insights into team performance. Descriptive Analytics: Summarizes key metrics such as possession (54.3% ± 4.2), shots per match (15.8 ± 6.1), and goals scored (1.6 ± 0.5). Diagnostic Analytics: Uses effect size estimation and Welch’s t-test to identify significant differences between teams using possession-based and direct play styles. Predictive Analytics: Models match outcomes using logistic regression and ROC curves, yielding an AUC of 0.78 for predicting success. Prescriptive Analytics: Proposes strategies to optimize future performance, helping coaches and analysts refine decision-making. RESULTS: Results from the 2022–2023 Serie A season demonstrate the pipeline's effectiveness in providing actionable insights from raw data. For example, direct play teams scored fewer goals (1.1 ± 0.7) compared to possession-based teams (1.7 ± 0.6), with significant differences observed in goal difference (14.9 ± 19.0 vs. -14.9 ± 20.4, p < 0.05). Odds Ratios (OR) indicate that possession teams are 3.5 times more likely to have a higher goal difference than direct play teams (OR = 3.5, 95% CI: 1.2–10.2). Shot-Creating Actions were significantly higher for possession teams (16.2 ± 7.4) compared to direct play teams (12.1 ± 5.1), with a p-value < 0.05. Using open-source data offers a transparent, accessible, and cost-effective approach to performance analysis. The pipeline's application to the Serie A dataset allows for a more reproducible and robust evaluation compared to traditional methods. CONCLUSION: This study contributes to Sport Intelligence by shifting from tactical interpretation to a structured, engineering-driven approach for transforming raw data into decision support. The framework addresses the engineering gap in sports analytics, ensuring that data processing is transparent, replicable, and systematic. In conclusion, the study offers a novel pipeline for transforming open-source football data into decision support, enhancing decision-making through methodological rigor.
Read CV Bruno RuscelloECSS Paris 2023: CP-AP09
INTRODUCTION: Eye-tracking technology has become a widely used method for objectively assessing visual search strategies and information processing in athletes. However, the evidence in this area is limited, and further research is needed. This study aims to compare the eye-tracking metrics exhibited by soccer players when observing successful and unsuccessful offensive and defensive game scenarios. METHODS: The study included 18 healthy male soccer players (age: 18.18±0.29 years; body height: 176.44±7.67 cm; body weight: 70.22±7.39 kg). Inspired by the 2025 UEFA Champions League final between Paris Saint-Germain FC and Inter Milan, two 45-second animated game simulations were created on the Tactical Board platform. Athletes watched each video while imagining they were a player on the attacking or defending team. Eye-tracking data were collected under four experimental conditions for each participant: successful offensive, unsuccessful offensive, successful defensive, and unsuccessful defensive. Fixation (duration, count, pupil diameter, eye openness) and saccade (count, velocity, amplitude, direction) metric data were extracted using Tobii Pro Lab software (Tobii Pro AB, Stockholm, Sweden). The data were analyzed using a 2 (Play Result: goal vs. no goal) × 2 (Tactical Role: offensive vs. defensive) repeated-measures ANOVA in SPSS version 27 (IBM Corp., Armonk, NY, USA). RESULTS: A significant main effect of play result was observed for average whole-fixation eye openness, number of saccades, maximum peak velocity of saccades, maximum amplitude of saccades, total amplitude of saccades, and direction of first saccade (Fs(1,17) = 5.89–21.57, ps = <.001–.027, partial η² = .307–.559). A significant main effect of tactical role was found for average whole-fixation pupil diameter, average peak velocity of saccades, average amplitude of saccades, and maximum amplitude of saccades (Fs(1,17) = 4.48–9.85, ps = .006–.049, partial η² = .209–.367). No significant interaction effects between play result and tactical role were detected for any of the analyzed parameters (ps > .05). CONCLUSION: The findings indicate that both the play result and tactical role have distinct effects on oculomotor behavior. These results offer significant insights for coaches and clinicians in the design of training programs intended to enhance visual attention and decision-making processes. It is recommended that subsequent research examine these behaviors under real-field conditions.
Read CV Murat EmirzeoğluECSS Paris 2023: CP-AP09