Building for the Future : A Data-Driven Decision-Support System for Football Performance Analysis

Sammanfattning

Football clubs increasingly rely on performance data to support sporting decisions, yet translating large volumes of analytical information into actionable insights remains a challenge. This thesis explores the use of large language models (LLMs) and retrieval-augmented generation (RAG) to automate the generation of strategically structured football analysis reports. A decision-support system was developed that combines football performance data with publicly available strategy documents to produce natural-language reports and provide traceable links between analytical claims and the underlying evidence. The system was evaluated using reports generated for four clubs in the Swedish top division, Allsvenskan. The results indicate that the proposed approach can produce reports that are largely grounded in the available data while maintaining traceability between conclusions and supporting visualisations. Feedback from a professional football practitioner further suggested that the generated reports were relevant, specific, and practically useful. The evaluation also identified limitations, particularly when the system moves beyond descriptive analysis and attempts to generate strategic recommendations, where human judgement remains important. The findings suggest that LLM-based decision-support systems have the potential to assist football organisations by reducing the effort required to transform complex performance data into interpretable insights. Rather than replacing analysts or decision-makers, such systems may serve as tools that support and augment human expertise.