Method of ranking athletes in a decision support system in the process of preliminary selection of players

The topic of the article is the method of ranking athletes in a decision support system in the process of preliminary selection of players. The purpose of the is to develop an athlete ranking method that enables the integration of all available information types, facilitating informed decision-making, as selection accuracy directly impacts the quality of the resulting team. Research objectives: investigate the information types utilized in athlete ranking and identify appropriate methods for data representation and processing; to develop a method based on specific metrics, production rules, weighting coefficients, and an artificial neural network (ANN) model to generate player ratings within a DSS; to conduct an experiment to validate the effectiveness of the proposed method. Research results: The article proposes a hybrid ranking method that integrates a deterministic component (IF-THEN production rules) with an adaptive ANN model. This architecture combines expert heuristic knowledge with automated data-driven learning, ensuring both decision explainability and the model's robustness in complex conditions. At the first level, the system ranks candidates using predefined logical rules derived from expert assessments by coaches and analysts. These rules process formalized metrics such as physical development, fitness, and competition history, enabling initial screening based on basic functional suitability criteria for a specific position. At the second level, an ANN module is integrated to perform a multidimensional analysis of the players' numerical characteristics. By identifying latent non-linear dependencies between tactical and technical action indicators and anthropometric variables, the model provides a predictive assessment of overall potential. This approach is particularly effective in scenarios where manually formulated rules cannot capture the complex interdependencies between variables. The final output integrates a binary classification (fit/unfit) with a quantitative rating generated by the neural network. These results can be mapped to a formalized scale for further ranking or utilized as primary input for final coaching decisions. This structure ensures an optimal balance between interpretability, accuracy, and adaptability within the volleyball player selection process. Conclusions: The scientific novelty lies in combining a structured system of TTA indicators with role-specific weighting obtained through expert consensus, allowing for contextual adaptation. Additionally, the integration of a neural network component to identify latent dependencies between performance indicators dynamically represents a novel approach for short-term selection in volleyball. Further research will focus on developing adaptive player models and optimizing selection processes to enhance the effectiveness of decision-making within the DSS.