This study addresses a central tactical dilemma for football coaches: whether to employ a defensive strategy, colloquially known as "parking the bus", or a more offensive one. Using an advanced Double Machine Learning (DML) framework, this project provides a robust and interpretable tool to estimate the causal impact of different formations on key match outcomes such as goal difference, possession, corners, and disciplinary actions. Leveraging a dataset of over 22,000 matches from top European leagues, formations were categorized into six representative types based on tactical structure and expert consultation. A major methodological contribution lies in the adaptation of DML to handle categorical treatments, specifically formation combinations, through a novel matrix-based residualization process, allowing for a detailed estimation of formation-versus-formation effects that can inform a coach's tactical decision-making. Results show that while offensive formations like 4-3-3 and 4-2-3-1 offer modest statistical advantages in possession and corners, their impact on goals is limited. Furthermore, no evidence supports the idea that defensive formations, commonly associated with parking the bus, increase a team's winning potential. Additionally, red cards appear unaffected by formation choice, suggesting other behavioral factors dominate. Although this approach does not fully capture all aspects of playing style or team strength, it provides a valuable framework for coaches to analyze tactical efficiency and sets a precedent for future research in sports analytics.
Football coaches face a critical decision before each match: which formation to implement to maximize their team's performance. The organization of the eleven starting players into specific structural patterns is a foundational element of tactical strategy. This project aims to provide data-driven insights for this decision by analyzing whether certain defensive formations, commonly referred to as "parking the bus" (1), are truly effective in specific scenarios. José Mourinho, former Chelsea FC manager, popularized this term to describe a rival's excessively defensive approach. These defensive strategies are often employed against stronger opponents (2,3); yet their actual effectiveness remains a subject of debate. This study addresses this debate directly for the benefit of coaches and tactical analysts. We examine the direct relationship between formations and several key match outcomes, moving beyond anecdotal evidence to provide a robust statistical analysis. After generalizing and grouping formations based on tactical similarity and expert criteria, a Double Machine Learning framework is applied to estimate the isolated impact of each formation combination on various variables, such as goal difference. While formations may not be the sole predictor of match statistics, understanding whether specific systems consistently enhance or hinder key performance outcomes is invaluable for strategic planning.
Ultimately, the objective is to assess the isolated impact of formations themselves, providing coaches with a clear, unbiased look at how tactical systems, independent of individual team quality, influence match results.
Match data for this study were obtained from the Sportmonks API (4), a commercial provider of comprehensive football data. The dataset comprises over 22,000 professional league fixtures and includes a wide range of variables critical for our analysis, such as match statistics, team formations, and contextual information.
To ensure the dataset was robust and representative of modern football tactics, specific criteria were applied for data inclusion. The selection of seasons and leagues was guided by two primary considerations: regulatory consistency and tactical diversity.
The analysis includes seasons from 2018-2019 to 2024-2025. This broad temporal scope was chosen to ensure a sufficient data volume for robust statistical modeling. While the majority of this period reflects the modern tactical environment shaped by the “five substitutions” rule, seasons prior to this regulatory change (2018-2019, 2019-2020, and 2020-2021) were also included to significantly expand the dataset’s size and statistical power. The effects of this regulatory change are implicitly captured by the inclusion of both league and season within the confounder set, which accounts for temporal and league-specific variations that may be correlated with the rule change.
To create a geographically and tactically diverse dataset, matches from the seven highest-ranked European men’s football leagues according to UEFA coefficients were included: the first divisions of England, Italy, Spain, Germany, France, the Netherlands, and Portugal. To further enhance the statistical power of the models, additional leagues were selected based on two criteria: a minimum of 32 regular-season matches and sufficient data availability from the API. This led to the inclusion of the top divisions of Turkey, Belgium, and Poland, as well as the second divisions of Spain, Italy, and England.
The primary objective of this analysis is to isolate the intrinsic relationship between a team’s formation and a rival’s formation with respect to various match outcomes. To achieve this, a set of confounding variables has been selected. These variables are chosen to control for factors that influence both team strength and match outcomes but are not a direct result of the chosen formation. Consequently, mediator variables such as accumulated goals or possession, which are themselves outcomes of a team’s playing style, will be intentionally excluded from the model.
The following variables are used as confounders:
• Match Context: The season, league, or day of the week of the fixture.
• Team Side: A binary indicator for whether the main team played at home or away.
• Team Strength Metrics:
-The rate of accumulated points for both teams prior to the match.
-A binary flag indicating whether each team was simultaneously competing in the UEFA Champions League.
-The league ranking of each team at the time of the fixture.
-The winning streak from last matches.
• Environmental Factors: Weather data for the match venue.
The final dataset included additional key variables for each fixture which are used as mediators, treatment variables or target variables at some point of the analysis:
• Performance Metrics: Goals, corners, possession, and disciplinary actions (red and yellow cards) for both teams.
• Tactical Information: The formation system used by
This content is AI-processed based on open access ArXiv data.