Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/152100
Title: Analyzing Football Tactics through Finishing Sequences Classification
Author: Llinàs Martínez, Alexandre
Tutor: Divorra Vallhonrat, Teresa
Abstract: This study introduces a robust framework for analyzing and comparing tactical behaviors in football through the clustering of possession sequences and the use of tactical data. Leveraging unsupervised machine learning techniques, possession sequences ending in shots were classified into meaningful clusters, capturing distinct tactical contexts such as counterattacks, long build-ups, and high-pressure recoveries. Principal Component Analysis (PCA) was applied to identify key trends within these clusters, enabling the characterization of team playing styles and facilitating comparisons between teams in various contexts. The analysis demonstrated the capacity of clustering to highlight nuanced tactical differences, revealing that teams tend to adapt their behaviors within specific contexts rather than adhering to rigid tactical profiles. Furthermore, a neural network was trained using tactical data to predict these tactical contexts, achieving moderate success despite data limitations. Results showed that tactical data, while not perfect, are representative enough to discern tactical behaviors, especially when clusters are well-defined and grounded in football semantics. The framework provides practical applications for coaching staff, allowing them to evaluate and compare team strategies, not only between opponents but also within their own team across different periods. This enables insights into tactical evolution and adjustments over time. However, the study also highlights several limitations, including the small dataset size, restricted to one season, and the focus solely on sequences ending in shots. Future research should address these limitations by incorporating more diverse data, refining tactical features, and exploring alternative methodologies.
Keywords: Football analytics, Machine learning, play styles
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: Jan-2025
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
TF_GCDA_allinasm.pdf2,47 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.