Football Without Crowds: Does Home Advantage Exist Anymore?

The people’s game without the people… a frequently discussed topic and now we’ve thrown the data at it. Let’s have a look at how teams in the English Premier League have fared pre-and-post the covid restart.

Before we start, there’s a few terms we should be familiar with. In football, we know that goals often don’t give us the best representation of performance in a football match and that can skew our analysis, particularly if we are working on smaller datasets. In recent years, we have seen the rise in popularity of the ‘Expected Goals’ (xG) statistic. The xG stat assesses the likelihood of a shot being a goal with better chances being given better odds. This helps to soften the impact of the odd and the spectacular.

We are going to look at a variant of the xG stat that excludes penalties (the non-penalty xG). Our proxy for performance in a match is going to be the difference between the non-penalty expected goals and the non-penalty expected goals allowed. Let’s take a match between Arsenal and Leicester for example. In the match, Arsenal might have a non-penalty xG of 1.2 and Leicester might have a non-penalty xG of 0.8. This makes Arsenals non-penalty xG allowed 0.8 and our performance proxy (for Arsenal) will be 1.2–0.8 (0.4).

So now that we have our proxy for performance we want to investigate if teams perform better at home (which we expect) or away from home, and more importantly if the lack of a crowd has changed this expected idea of home advantage.

To do this we will first look at the matches played in 2019/20 with the crowd present (pre-covid stoppage). We will compare, for each side, the average performance at home vs. the average performance away and plot the difference in these averages.

Figure 1: Comparing Home vs Away Performances for sides during the 2019/20 Season with Crowds Present (Pre-Covid)

It seems that, for the 17 sides present in both seasons, most sides performed significantly better at home. Southampton and Leicester (well versed on the counter attack) are the only sides outperforming on the road, with 11 of the 17 sides having home performance differences exceeding 0.5.

We can now isolate the post-covid restart data for the 2019/2020 season as well as the matches played without crowds this season and compare those results to this baseline.

Figure 2: Comparing Home vs Away Performances for sides during the 2019/20 and 2020/21 Seasons without Crowds

Figure 2 indicates that while most sides still prefer playing at home (lack of travel, familiar training facilities, etc.), there is much less in it. There are now 6 sides that actually perform better away from home and only 5 of the 17 sides have a home performance outperformance larger than 0.5 (compared to 11 with crowds).

While these graphics indicate that home (crowd) advantages seem to have faded, they don’t give us an indication of the underlying home performances. Are teams conceding more goals at home? Are they scoring fewer goals? Is it a similar effect across teams?

We can compare home performances with crowds against home performances without crowds for non-penalty expected goals and then non-penalty expected goals allowed.

Figure 3: The Change in Non-Penalty Expected Goals (Crowds Present vs. No Crowds) across Teams for Home Performances

The most striking observation from figure 3 is that every team except Newcastle United has seen a drop in their non-penalty expected goals (an ode to Callum Wilson). The hardest hit teams by the absence of crowds have been Burnley, Sheffield United, Brighton and Manchester United. While Manchester City’s xG has seen a large drop, they are still the most offensive side in the league.

Figure 4: The Change in Non-Penalty Expected Goals Allowed (Crowds Present vs. No Crowds) across Teams for Home Performances

A drop in attack at home is replaced by better defensive numbers for most sides. However for many of the sides who have seen the worst drop off in home attacking numbers, the defensive numbers make equally worrying reading. Manchester United and Brighton are conceding more expected goals at home while there is a lack of significant defensive improvement from Sheffield United.

There are certainly interesting differences when we look at pre-and-post covid matches. We have seen a massive drop in home over-performances from most teams as they look to grind out results rather than play crowd-pleasing, expansive football. We have also highlighted a few teams like Manchester United, Brighton, Burnley and Sheffield United that have been really hurt by the lack of crowd support.

I would like to highlight the understatr package in R which was used to source all the data for this analysis.