足球比賽的空間分析：使用代理人模型探索不同戰略情境下足球隊的進攻模式

Spatial Analysis of Football Match: Using agent-based model to explore the attacking patterns of football teams under different strategic scenarios

倪煒傑(Wei-Jye Goy)

指導教授 ： 溫在弘

國立臺灣大學/理學院/地理環境資源學系/碩士(2023年)

https://doi.org/10.6342/NTU202301488

摘要

本研究旨在探索足球隊伍在不同策略情境下的進攻模式，使用代理人模型（Agent-Based Model）開發一個融入真實軌跡數據和運動模式的足球比賽模擬器。本研究利用英超聯賽（EPL）2017/2018賽季的軌跡數據，通過分析頂級球隊的運動模式和策略決策，將這些數據應用於研究的代理人的決策過程中。接著，建立了基於足球不同事件的運動模式的概率矩陣。通過將這些運動頻率和幾率納入模型，將觀察到更準確的球員運動和場上不同區域之間的傳球模式。在模擬器的設置階段中建立了各種規則以模擬策略選擇，並展示了模擬結果的策略特徵與真實足球比賽的策略相同。與以往的研究相比，本研究展示了此足球比賽模擬器的靈活性，透過調整模擬器的參數可以探索不同策略下的球隊進攻模式。本研究的模擬結果為足球模擬模型的發展做出了貢獻，融入了現實世界的數據，捕捉了運動模式，提高了模擬的逼真度和效果。這些研究成果可以進一步增強足球遊戲模擬的真實性，並為足球教練在部署新策略或分析足球比賽時提供寶貴的見解。未來對我們模型的擴展和改進具有重大潛力，可進一步推動足球模擬和分析領域的發展。

關鍵字

空間分析 ； 體育分析 ； 足球策略分析 ； 比賽事件資料 ； 代理人基模擬 ； 循序資料樣式探勘

並列摘要

This study aims to explore the attacking patterns of football teams under different strategic scenarios by developing an Agent-Based Model that incorporates real trajectory data and movement patterns. The study utilizes trajectory data from the 2017/2018 season of the English Premier League (EPL) and analyzes the movement patterns and strategic decision-making of top-tier teams. Subsequently, a probability matrix based on the movement patterns for various football events is established. By incorporating these movement frequencies and probabilities into the model, more accurate player movements and passing patterns between different areas of the field are observed. Various rules are established during the configuration stage of the simulator to simulate strategy selection, and the simulation results demonstrate that the strategic features are consistent with real football matches. Compared to previous research, this study showcases the flexibility of the football simulator, allowing for the exploration of team attacking patterns under different strategies by adjusting the parameters of the simulator. The simulation results of this study contribute to the development of football simulation models by incorporating real-world data, capturing movement patterns, and enhancing the realism and effectiveness of the simulations. These research findings can further enhance the authenticity of football game simulations and provide valuable insights for football coaches in deploying new strategies or analyzing football matches. The future expansion and improvement of our model hold significant potential for advancing the field of football simulation and analysis.

並列關鍵字

Spatial Analysis ； Sport Analysis ； Football Strategies Analysis ； Match Event Data ； Sequential Pattern Mining ； Agent-based Modeling

參考文獻

Achten, P. (2011). The soccer-fun project. Journal of Functional Programming, 21(1), 1-19.

Barreira D, Garganta J, Guimarães P, Machado J, Anguera MT (2014a). Ball recovery patterns as a performance indicator in elite soccer. Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and Technology. 2014;228(1):61-72.

Bialkowski, A., Lucey, P., Carr, P., Yue, Y., Sridharan, S.,\& Matthews, I. (2014b). "Large-scale analysis of soccer matches using spatio-temporal tracking data". In 2014 IEEE International Conference on Data Mining (pp. 725-730). IEEE.

Brasen, L.,\& Van Haaren, J. (2018). "Football analytics: predicting and analysing the outcome of soccer matches". In Sports Data Mining (pp. 94-111). CRC Press.

Bornn, L., Cervone, D.,\& Fernandez, J. (2018). Soccer analytics: Unraveling the complexity of soccer through on-ball and off-ball analysis. In Handbook of statistical methods and analyses in sports. Chapman and Hall/CRC.