Tactical Styles and Goal Scoring: Insights from the 2022 FIFA World Cup through Autoencoder Feature Selection Unveiling Tactical Patterns Using Machine Learning

Purpose: The present study aims to examine the technical indicators that have a significant impact on goal-scoring in the 2022 FIFA World Cup using neural network algorithms for feature extraction. Methods: This study analyzed a dataset including 96 match performance records from 48 games. Feature selection from 50 performance indicators was done using an Autoencoder; latent features were explained using Pearson correlation coefficients and Multivariate linear regression models were applied to examine the relationship between latent features and goal scoring. Results: The following significant correlations were obtained with latent feature: LF2, high-pressing play, was positively correlated with goals scored (β = 0.985, P = 0.042), and LF3, counter-attacking play, was negatively related to goals (β = -1.376, P = 0.004). Conclusion: From this study, it can be outlined that high-pressing play, when strategically combined with attacking transition efficiency, significantly enhances the opportunities for scoring goals. Conversely, counter-attacking strategies related to a low goal-scoring effectiveness. The study provides useful insights for coaches and analysts in top-level football by offering new perspectives in the analysis of playing styles.