# Predicting soccer matches with complex networks and machine learning

**Authors:** Eduardo Alves Baratela, Felipe Jordão Xavier, Thomas Peron, Paulino Ribeiro Villas-Boas, Francisco Aparecido Rodrigues  
**arXiv:** [2409.13098](https://arxiv.org/abs/2409.13098)  
**Date:** 2024-09-19  
**Summary page:** https://www.summarizepaper.com/en/paper/2409.13098v1/

## Key points

- Soccer research is increasingly focused on enhancing performance analysis and sports prediction
- Complex networks are being used as an innovative tool to predict soccer match outcomes
- Combining passing network structural analysis with match statistical data provides deeper insights into game patterns and strategies
- Machine learning models were developed to predict wins and losses of soccer teams across different leagues using complex network metrics and match statistics
- Models based on passing networks were found to be as effective as traditional models, and merging both approaches led to more accurate predictions
- The study emphasized the importance of integrating diverse methodologies for a comprehensive understanding of game patterns
- Network metrics and match statistics play significant roles in shaping predictive models
- Using networks with finer temporal granularity (e.g., separate networks for each half of a match) yielded superior results compared to using a single network for the entire game
- Led by Eduardo Alves Baratela, Felipe Jordão Xavier, Thomas Peron, Paulino Ribeiro Villas-Boas, and Francisco Aparecido Rodrigues, this research will be published in the Journal of Complex Networks

## Summary

Soccer has long been a focal point for researchers and professionals in the sports industry. With a growing emphasis on incorporating scientific methods, there is a strong focus on enhancing performance analysis and sports prediction. This study delves into the realm of complex networks as an innovative tool for predicting soccer match outcomes. By combining structural analysis of passing networks with match statistical data, this research aims to provide deeper insights into game patterns and strategies employed by teams. Utilizing complex network metrics alongside match statistics, machine learning models were developed to predict wins and losses of soccer teams across different leagues. Surprisingly, the results revealed that models based on passing networks proved just as effective as traditional models relying solely on general match statistics. Moreover, when both approaches were merged, more accurate predictions were achieved, underscoring the value of integrating diverse methodologies for a comprehensive understanding of game patterns. The study also highlighted the significance of both network metrics and match statistics in shaping predictive models. Interestingly, employing networks with a finer temporal granularity (such as creating separate networks for each half of a match) yielded superior results compared to using a single network for the entire game. Led by Eduardo Alves Baratela, Felipe Jordão Xavier, Thomas Peron, Paulino Ribeiro Villas-Boas, and Francisco Aparecido Rodrigues,this research is set to be published in the Journal of Complex Networks. By shedding light on the intricate relationships between players, their positions,and interactions during matches,this study underscores how the fusion of complex network analysis and machine learning can offer invaluable insights into soccer tactics and strategies.