[Paper Review] Home advantage of European major football leagues under COVID-19 pandemic

Eiji Konaka|arXiv (Cornell University)|Jan 2, 2021

Sports Analytics and Performance11 references4 citations

TL;DR

This study quantitatively evaluates the impact of spectator absence on home advantage in European football leagues during the COVID-19 pandemic using a dynamic rating model with a shared home advantage parameter. It finds that home advantage significantly decreased in closed-door matches across four major leagues (England, Germany, Italy, Spain), though a residual advantage remained, providing strong evidence for the crowd effect as a key driver of home advantage.

ABSTRACT

Since March 2020, the environment surrounding football has changed dramatically because of the COVID-19 pandemic. After a few months' break, re-scheduled matches were held behind closed doors without spectators. The main objective of this study is a quantitative evaluation of ``crowd effects'' on home advantage, using the results of these closed matches. The proposed analysis uses pairwise comparison method to reduce the effects caused by the unbalanced schedule. The following conclusions were drawn from the statistical hypothesis tests conducted in this study: In four major European leagues, the home advantage is reduced in closed matches compared to than in the normal situation, i.e., with spectators. The reduction amounts among leagues were different. For example, in Germany, the home advantage was negative during the closed-match period. On the other hand,in England,statistically significant differences in home advantage were not observed between closed matches and normal situation.

Motivation & Objective

• To assess the quantitative impact of spectator absence on home advantage in European football leagues during the pandemic.
• To isolate the effect of crowd presence from other confounding factors such as unbalanced schedules and team strength fluctuations.
• To provide robust evidence on whether the 'crowd effect' is a primary driver of home advantage in football.
• To compare home advantage levels across normal, mixed, and closed-door match conditions in major European leagues.
• To develop and apply a statistical model that estimates home advantage independently of schedule imbalance and team strength changes.

Proposed method

• A dynamic logistic regression model estimates match outcomes based on team strength ratings and a shared home advantage parameter.
• Team strength is represented by a time-varying rating $ r_i $, with home advantage modeled as an additive parameter $ r_{\text{homeAdv}} $.
• The probability of a home team winning is modeled as $ \frac{1}{1 + \exp(-\Delta r)} $, where $ \Delta r = r_i + r_{\text{homeAdv}} - r_j $.
• Parameters are re-estimated every five matchweeks using the most recent results to track changes in home advantage over time.
• Home advantage is compared across three match types: past normal matches, 2019/20 normal matches, and 2019/20 closed-door matches.
• Non-parametric Wilcoxon rank sum tests are used to compare the medians of estimated $ \bar{r}_{\text{homeAdv}} $ across conditions, ensuring robustness to distributional assumptions.

Experimental results

Research questions

• RQ1 Does the absence of spectators during the COVID-19 pandemic significantly reduce home advantage in European football leagues?
• RQ2 How does the magnitude of home advantage in closed-door matches compare to that in normal matches across different leagues?
• RQ3 Is there a residual home advantage even when stadiums are empty, indicating non-crowd-related factors?
• RQ4 To what extent do schedule imbalances and team strength variations during the pandemic period affect the estimation of home advantage?
• RQ5 Are the effects of spectator absence on home advantage consistent across different top European leagues?

Key findings

• Home advantage was significantly reduced in closed-door matches compared to normal matches in all four leagues (England, Germany, Italy, Spain), with p-values < 10⁻⁴ for all comparisons.
• The reduction in home advantage was most pronounced in Germany (p = 3.98×10⁻⁵) and Spain (p = 4.10×10⁻⁴), indicating strong statistical evidence for the crowd effect.
• Despite the reduction, a residual home advantage remained in closed matches, with median $ \bar{r}_{\text{homeAdv}} $ values still above zero across all leagues.
• No significant difference was found between past normal matches and 2019/20 normal matches (p = 0.272), validating the stability of the model and the reliability of the comparison.
• In England, the lack of significant difference between normal and closed matches was attributed to schedule imbalance, as weak teams played more home games in the closed period.
• The correlation between final league standing and number of home matches in the closed period was highest in England (0.4052), suggesting stronger teams were disproportionately assigned home games, which may have masked the true reduction in home advantage.