Deep Siamese Metric Learning: A Highly Scalable Approach to Searching Unordered Sets of Trajectories

This work proposes metric learning for fast similarity-based scene retrieval of unstructured ensembles of trajectory data from large databases. We present a novel representation learning approach using Siamese Metric Learning that approximates a distance preserving low-dimensional representation and that learns to estimate reasonable solutions to the assignment problem. To this end, we employ a Temporal Convolutional Network architecture that we extend with a gating mechanism to enable learning from sparse data, leading to solutions to the assignment problem exhibiting varying degrees of sparsity. Our experimental results on professional soccer tracking data provides insights on learned features and embeddings, as well as on generalization, sensitivity, and network architectural considerations. Our low approximation errors for learned representations and the interactive performance with retrieval times several magnitudes smaller shows that we outperform previous state of the art.

Optimizing Field Body Fat Percentage Assessment in Professional Soccer Players

Body composition is a determinant of performance in soccer. To estimate the body fat percentage (%BF), dual energy X-ray absorptiometry (DXA) is effective though this method is expensive and not readily accessible. This study examines the validity of widely used field methods based on anthropometric data and bioelectrical impedance analysis (BIA). Participants were 21 male Spanish First Division soccer players aged between 22 and 35 years. In each participant, body fat mass was determined by BIA and using 18 anthropometric equations including skinfold (SKF) measurements. DXA was used as reference. Correlation with DXA measurements was excellent for all equations and separate SKF measurements yet only moderate for BIA. However, only the equation recently developed for use in soccer players based on iliac crest and triceps SKFs showed no significant or standardized differences with DXA-derived %BF and these measurements also had the lowest bias. Our findings suggest that when DXA is not available, the best field method for %BF assessment in footballers is the equation based on iliac crest and triceps SKF. As another good option, we propose the sum of triceps, subscapular, supraspinal, and abdominal SKFs, as this combination also showed good correlation with DXA.

Individual-based Creatine Kinase Reference Values in Response to Soccer Match-play

The aim of the present study was to determine the creatine kinase reference limits for professional soccer players based on their own normal post-match response. The creatine kinase concentration was analyzed in response to official matches in 25 players throughout a 3-year period. Samples were obtained between 36–43 hours following 70 professional soccer matches and corresponded to 19.1±12.1 [range: 6–49] samples per player. Absolute reference limits were calculated as 2.5th and 97.5th percentile of the samples collected. Creatine kinase values were also represented as a percentage change from the individual’s season mean and represented by 90th, 95th and 97.5th percentiles. The absolute reference limits for creatine kinase concentration calculated as 97.5th and 2.5th percentiles were 1480 U.L−1 and 115.8 U.L−1, respectively. The percentage change from the individual’s season mean was 97.45±35.92% and players were in the 90th, 95th and 97.5th percentiles when the percentages of these differences were 50.01, 66.7, and 71.34% higher than player’s season mean response, respectively. The data allowed us to determine whether the creatine kinase response is typical or if it is indicative of a higher than normal creatine kinase elevation and could be used as a practical guide for detection of muscle overload, following professional soccer match-play.

Knee Flexor Eccentric Strength, Hamstring Muscle Volume and Sprinting in Elite Professional Soccer Players with a Prior Strained Hamstring

The aim was to determine if players with a prior hamstring strain injury (HSI) exhibit bilateral deficits in knee flexor eccentric strength and hamstring muscle volume and differences in sprinting performance compared with players without a history of HSIs. Forty-six male professional soccer players participated in this study. Eccentric knee flexor strength, hamstring muscle volume (MRI), and a 20-m running sprint test (5- and 10-m split time) were assessed at the start of the preseason. Eccentric knee strength of the previously injured limbs of injured players was greater (ES: 1.18–1.36) than the uninjured limbs in uninjured players. Previously injured limbs showed possibly larger biceps femoris short heads (BFSh) and likely semitendinosus (ST) muscle volumes than the contralateral uninjured limbs among the injured players (ES: 0.36) and the limbs of the uninjured players (ES: 0.56), respectively. Players who had experienced a previous HSI were possibly slower in the 5-m (small ES: 0.46), while unclear differences were found in both the 10-m and 20-m times. Players with a prior HSI displayed greater eccentric knee flexor strength, possibly relatively hypertrophied ST and BFSh muscles, and possibly reduced 5-m sprinting performances than previously uninjured players. This can have implication for the design of secondary hamstring muscle injury prevention strategies.

Coaching Clinic Basic Movements and Dominant Physical Conditions for Athletes at Gemilang Raya FC

Basic movements and dominant physical conditions are often still a problem for soccer players in supporting the strategies and tactics of a coach in building attacks or in defense, especially at Football Club Gemilang Raya Kuningan. Some players still have shortcomings in carrying out correct and accurate basic movements and do not yet know the physical procedural requirements that a professional soccer player must possess. Based on the problems faced, the STKIP Muhammadiyah Kuningan Service Team carried out activities to overcome the problems faced by partners. Community service activities aim to improve the abilities and skills of Football Club Gemilang Raya Kuningan athletes in performing basic movements and knowing the dominant physical condition. The activity was carried out at the Gemilang Raya Kuningan Football Club and was attended by 20 players. The method used is a coaching clinic which consists of 2 sessions. The material provided consists of understanding basic movements and dominant physical conditions, practice on basic movements and dominant physical conditions. The results of the activity to the community showed an increase in the abilities and skills of Football Club Gemilang Raya Kuningan players regarding basic movements and dominant physical conditions.

Mandatory lockdown impairs performance in professional soccer players.

COVID-19 caused a total halt in sport competition during 2020. The purpose of this study was to analyze the changes between pre- and post-lockdown competitive periods in the players’ workload variables in competition. Seventeen professional football players were monitored using a WIMU PRO® inertial device throughout the 2019-20 season. Anthropometric and physical fitness were assessed with the aim to relate possible associations between these characteristics and the workload changes in the pre- and post-lockdown periods. During the lockdown, players carried out an 8-week guided self-training. There was a general decrement in the players’ physical workload demands in competition, and the parameters related to high-intensity actions as accelerations and decelerations (-8.96% [ES: 0.64] and -11.04% [ES: 0.77] respectively; p < .05), Ind HSR (-35.57% [ES: 0.92]; p = .002), HMLD (-8.58% [ES: 0.66]; p = .016), PLOAD (-7.03% [ES: 0.54]; p = .047) and Vmax (-3.80% [ES: 0.65]; p = .016) can be highlighted. The results showed high negative correlations between match workload variables prior to the lockdown and the percentage of change in these variables after the lockdown period. Individual percentages of change showed high variability in players’ changes. Individual self-training programs should be reviewed to minimize the impact of a “detraining” period in players’ physical performance during possible new lockdown periods.

Machine Learning on Soccer Players Position

During the last few years, sports analytics has been growing rapidly. The main usage of this discipline is the prediction of soccer match results, even if it can be applied with interesting results in different areas, such as analysis based on the player position information. In this paper, we propose an approach aimed to recognize the player position in a soccer match, predicting the specific zone in which the player is located in a specific moment. Similar objectives have never been considered yet with our best knowledge. We consider supervised machine learning techniques by considering a dataset obtained through video capturing and tracking system. The data analyzed refer to several professional soccer games captured at the Alfheim Stadium in Tromso, Norway. The approach can be used in real-time, in order to verify if a player is playing according to the guidelines of the coach. In the experimental analysis, three different types of classification have been performed, i.e., three different divisions of the field, reaching the best results with Random Tree Algorithm.