COACHING DECISION SUPPORT SYSTEM USING INFORMATION TECHNOLOGY

Among all sports, football stands out as the most popular one, characterized by intense competition where teams constantly seek ways to improve their performance.Football, being one of the oldest games, has begun to adapt to the rapid development of artificial intelligence.Increasingly, teams employ AI to analyze their opponents, study tactical behaviors on the field, detect open spaces that frequently appear during attacks or defense, and uncover other critical aspects of the game.Therefore, the relevance of this research is driven by the progress of artificial intelligence and the high level of competition in football, where new strategies for victory are sought every day.In modern football, decision-making increasingly relies not only on the intuition of coaches but also on objective data obtained from analytical systems.For this reason, this work presents the development of an intelligent analytical system that uses match video recordings to provide objective insights and decision support.A key aspect of evaluating player performance is the formation of a metric system that reflects both individual and team characteristics of the game.For example, for a forward, the system collects data such as average speed, maximum speed, and total distance covered during a match.Based on these data, a profile of the player's physical activity is generated, allowing assessment of his readiness for different tactical scenarios and overall on-field efficiency.Additionally, spatial aspects of the game must be considered.Using player coordinates extracted from video analysis, the system constructs two-dimensional maps of player positions, heatmaps, and influence zones.These data allow the coach to evaluate not only individual performance but also interactions between playersfor example, how often a forward finds himself in potentially dangerous areas or how effectively he utilizes open spaces created by teammates.To enable such functionality, the system must accurately detect in-game events, achieved through computer vision models.Among publicly available architectures, YOLO (You Only Look Once) was selected, as it has become a de facto standard in object detection tasks due to its speed and accuracy -key characteristics for the