Relegation for Brighton?

Sandipani Basu
5 min readMar 29, 2021

29 matches played in the Premier League so far and we have Manchester City leading the pack with 71 points. Defending champions Liverpool have managed to stay in the top 10 despite their injury crisis. West Ham, the surprise package this season sits at 5th, and The Arsenal under Mikel Arteta languishes at 9th. Leeds United(11th) seemed to have dropped points in the past few games, gaining only 4 points out of a possible 12.

The tail-end of the table sees Sheffield United and West Brom, struggling to keep pace with the rest. 18th place Fulham, although 8 points above West Brom, need to change their fortunes soon or go back to Championship in just one season. Brighton, the hero of this post are 16th, with 32 points and looking upwards.

Brighton, have completed 303 passes into the penalty area, averaging 9.34 key passes that lead directly to a shot every 90 minutes. They have completed 81 crosses into the penalty area, the highest in the league. They have an expected goals of 40.7, second highest xG in the bottom half of the table. They are still 16th in the league. Expectation however has not translated into reality for Brighton.

Expected vs Actual goals in PL 2020–21 till Gameweek 29

All the teams below the line have been consistently underperforming their xG this season, with Brighton(in red) and Fulham leading the underperformers. Brighton have scored 31 goals out of a possible 40 and conceded 36 when they should have conceded only 28(xGA = 28.5). That is not the case with Fulham though, their xGA and GA match each other, clearly showing their defensive frailties this season.

Brighton’s low count on the goals chart can be clearly explained with their low shot on target percentage of 28.3% this season. They have taken 364 shots with only 103 on target. This is the worst SoT% in the PL this season. Additionally their shot creating actions, defined as two offensive actions directly leading to a shot(passes, dribbles or drawing fouls) is a pretty impressive one at 20.21 per 90. However, up front their goal creation is sub par at 1.69 per 90. For reference, Leicester City has a SCA of 19.55 per 90 and 3.03 GCA per 90. They are creating chances but not finishing them.

The 50% unpredictable side of the game has been heavy on them. The match against West Brom, two of their penalties hit the post, a goal was disallowed in the most bizarre of circumstances. They had a massive xG of 2.6 that match, but scored none.

Expected goals are like the most important metric to determine a team’s performance nowadays. Is such a difference between what is expected and the actual, NORMAL ? Yes, it is actually, exactly that, a normal distribution.

Let me introduce the concept of correlation. The correlation coefficient between two measures which varies between -1 and 1, is a measure of the relative weight of the factors they share. For example the amount of electricity consumed per household in a locality and the number of accidents in that locality have a weak correlation. However, the average temperature of that locality and the electricity consumption per household might have a stronger correlation.

If we take a look closely at the data from this season, Brighton’s xG and their goals has a weak positive correlation. The correlation coefficient(r) is around 0.16 from the beginning of the season till now. But does this mean xG is not helping at all predicting the number of goals the team is scoring? Looking at this one would say xG does not work for Brighton and it can lead to unjustified speculations, like sacking Graham Potter.

(Left )— 2020–21. (Right)- 2017-Present

However if we take a look at the data across seasons from 2017 till the current season, the correlation is higher (0.42).

We would expect a team to score exactly or close to the number of goals as is expected of them. So that means, their goals minus expected goals or G-xG over time should be around 0. A positive G-xG would mean the team is scoring more than is expected of them and a negative value would mean that the team is underperforming. In other words ‘G-xG’ should be a Normal or Gaussian distribution. Below is Brighton’s G-xG distribution this season

KDE plot of Brighton’s G-xG this season

This clearly shows that the distribution is not normal(not a bell curve) and NOT centered around 0. It is centered around -1 to -0.5 instead, telling the tale of a misfiring Brighton this season. This however paves the way for us to estimate that their future games will not be as poor as the past ones have been. How can I say that? Look at the plot below of data from the 2017–18 season till the present.

KDE plot of Brighton’s G-xG from 2017-present

This means that when we take a larger sample of the population of data, G-xG tends to be a normally distributed metric. This tendency is known as regression to the mean and it is a statistical tendency. As Daniel Kahneman rightly said, “Whenever the correlation between two values is imperfect there will be regression to the mean”.

Now this does not mean, Brighton will win all the matches to come to balance out the skew. They are missing the end product and have been good in the middle with key passes and also defensively, I must say. Bissouma is attracting interest from Europe’s elites. Results will follow soon. Regression to the mean may take time. However, they will tend to score as many goals as are the chances they are creating, which was not happening earlier. A Welbeck “worldie” against Newcastle might just be the beginning of a team statistically tipped to stay up this season.

Their match up against Manchester United will be an interesting watch, since they are at the opposite ends of the spectrum. Man united have overperformed their xG by a lot this season.

Brighton and Hove Albion will stay up.

Data credits-



Sandipani Basu

An aspiring data geek with a keen interest in sports analytics, primarily soccer. I am on a gradual learning curve. But as they say good things take time.