Deakin University
Browse

Time-Series Analysis of Ball Carrier Open-Space (BCOS) in Association Football

journal contribution
posted on 2025-03-24, 05:04 authored by Ishara Bandara, Sergiy ShelyagSergiy Shelyag, Sutharshan RajasegararSutharshan Rajasegarar, Dan DwyerDan Dwyer, Eun-jin Kim, Maia Angelova
Abstract Assessing team performance in association football (commonly known as football or soccer) is challenging due to the sport’s low-scoring nature and inherent unpredictability. While evaluating strategies based on space control and the creation of open spaces has been explored in the literature, the temporal aspect of space availability for the ball carrier remains under-explored. This work introduces a novel time-series performance evaluation metric, Ball Carrier Open Space ( BCOS ), which focuses on the temporal dynamics of space available to the ball carrier to assess team performance. Additionally, it presents a novel approach to quantify open space for the ball carrier using player data extracted from television footage. This work discuss on BCOS in defensive third, central third and attacking third and a machine learning model is developed to evaluate their significance and temporal patterns. Trained model achieved 80.7% accuracy in classifying match-winning performances, underscoring the significance of BCOS . Correlation analysis between temporal features and match outcomes further reveals that BCOS in central third and attacking third are more important for match winning outcomes, while first-half performance plays a more critical role in determining match results than second-half performance. Based on the results of the correlation analysis, this study proposes a weighted ball carrier open space ( wBCOS ) metric to assess team performance, assigning weights to BCOS in attacking third, central third and defensive third based on their contributions to positive match outcomes. A machine learning model trained using wBCOS achieved an 82.5% accuracy in classifying match-winning performances, surpassing the performance of any previously published match-winner classification model.

History

Journal

SN Computer Science

Volume

6

Article number

302

eISSN

2661-8907

Language

en

Issue

4

Publisher

Springer Science and Business Media LLC

Usage metrics

    Research Publications

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC