Data-Driven Approach to Optimize A Football Team Within A Budget

Abstract

"The football player transfer market poses a challenge for low-budgeted clubs due to its unpredictability, leading to a rapid rise in value for players with average key performance metrics. This project aims to implement a data-driven solution to optimize team composition while adhering to budget constraints, addressing matters such as the lack of data-driven tactics, overlooking budget constraints, and focusing on short-term benefits. During the designing phase of the project, a thorough requirements analysis was carried out. By adopting an iterative and incremental approach, the project aims to continuously refine the solution based on new insights, promoting progress transparency, flexibility, and a deep understanding of project requirements. The PRINCE2 project management methodology provides a consistent framework for development, promoting planning, implementation, monitoring, and control. To ensure a thorough evaluation of the finished work, tasks are divided into logical phases with specified timeframes. In addition, the Rational Unified Process (RUP) software development approach was applied aligning with the project's iterative nature, dividing solution development into phases such as elaboration, construction, and transition. These methodologies facilitate the systematic revision of project objectives and ensure that a solution is effectively developed. The effectiveness of the solution was demonstrated through the comparative analysis of three machine learning models: K-Means Clustering, Support Vector Machines (SVM), and Random Forest, across different team selection modes. The results highlighted the superior performance of SVM and Random Forest in key scenarios, particularly with SVM achieving an accuracy of 0.8461 in Possession Mode, significantly aiding strategic team building within budget constraints. This approach not only bridges the identified gap in data-driven football strategies but also sets a foundation for extending similar methodologies to other sports domains, leveraging machine learning to optimize decision-making processes. "