EPL 2020/21 Season Analysis and Prediction

Published in

Towards Data Science

8 min readOct 7, 2020

Going into the international break after Game-week 4, 38 matches have been played which is exactly 10% of the total 380 matches to be played during the season. The current season has been by far been unpredictable with last year’s top teams dropping points and some mid-table teams and minnows performing remarkable well. In this post, I try to analyze the performance of teams and try to predict the result of upcoming fixtures.

Expected Goals(xG) is the major factor used for analysis and prediction, if you are not familiar with xG, it is recommended that you check out this post were xG is explained before proceeding further.

Home Advantage, Does that even exist in Post COVID Era?

Due to the pandemic, matches are currently happening in empty stadiums. Home advantage is more than the familiarity with the playing turf, it is the spirit and encouragement by tens of thousands of die-hard fans rooting for the victory of the home team.

That’s why even the thought of visiting Anfield or Old Trafford sends shivers down the spine of away teams. Generally, teams perform better in front of their home crowd compared to away fixtures.

In the current season, so far there is no evidence of home advantage. Out of the 38 matches played so far, 19 resulted in the away team winning, 3 were draws and the Home team only managed to win 16 matches which is around 42% of the total matches played.

The curious case of Home Disadvantage!

Home teams have been outscored by away teams this season. Moreover, home teams have underperformed in both xG(Expected Goals Scored) and xGA(Expected Goals Conceded).

The defending champions Liverpool is leading in expected goals per game with around 2.5 xG per game including penalties. Spurs, Aston Villa and Everton have also managed xG greater than 2.
West Brom which is currently the lowest ranking team is struggling to create chances with very low xG.
Teams with low xG lack creativity, they should reinforce the team with creative players.
xG can be low for teams that employ a defensive style of play that relies on counter attacks or mistakes from the opposition to score goals.
Everton, Southampton and West Ham seem to have impenetrable defense registering xGA below 1. Sheffield, Aston Villa and Brighton have also displayed solid defense.
3 out of last season’s top four teams-Liverpool, Manchester City and Manchester United features in the worst 5 defensive teams as per xG conceded. Manchester United recorded the highest expected goals conceded averaging around 2.5 goals per game. Leeds and West Brom also have more than 2 xGA.
Teams with high xGA should immediately address their problems in defense.

Net expected Goals Scored

Delta xG is the difference between expected goals scored and expected goals conceded which can be used to evaluate the form of a team. Everton seems to dominate the league with delta xG per game almost 1.5. Tottenham and Aston Villa also have delta xG more than 1. Despite 7–2 humiliation at the hands of Aston Villa, the Reds are close behind in the 4th position. On the other end, Manchester United and West Brom have delta xG of more than 1.5.

The teams above the red line have better xG compared to xGA. Last season’s top scorers Manchester City who usually dominates the league, has failed to produce more chances than their opponents. Their city rivals Manchester United had the worst start in decades.

Expectation Vs Reality

A team could outperform xG either when the team has prolific strikers who can convert half-chances or luckily managed to score some goals.
Aston Villa is outperforming xG by 1.5 goals per match which they are not likely to carry forward.
Leicester City also seems to outperform the xG by a huge margin, this can be attributed to 3 penalties scored by Vardy. Unless Leicester is awarded penalties every weekend(which is highly unlikely even with the new lenient handball policy), their margin would come down.
2 out of Manchester United’s 5 goals are also from penalties, xG from open play for the red devils also would be much lower.

Aston Villa, New Castle, Arsenal and Leeds are the only 4 teams out of 20 who managed to outperform the expected goals conceded. It seems that goal scoring has been very easy this season.
Liverpool, Manchester United, Fulham, Burnley and Brighton were unlucky to concede more than one goal per match than expected.

Quadrant Analysis

The best scenario is over-performing xG and xGA by a healthy margin.
Over-performing xG by a high margin is generally not sustainable in the long run.
Falling behind xG is not a big problem if the team manages to create more chances(xG).
The size of the bubble indicates the expected net goals scored(xG-xGA) per match. The first aim of the teams should be to improve the net goals scored.
1st Quadrant: Over-performed xG and Over-performed xGA
Teams in Q1 are doing well in over performing expected goals they have to carry on the momentum teams like Leeds and New Castle should try to improve the net Goals scored.
2nd Quadrant: Over-performed xG and Under-performed xGA
Teams in Q2 should try to move towards Q1, which is to improve the defense.
3rd Quadrant: Under-performed xG and Under-performed xGA
Teams in Q3 should improve both defense and attack.
4th Quadrant: Under-performed xG and Over-performed xGA
There are no teams in Q4.

Predicting Results

The logic used to predict the scores of the match is based on xG per game and xGA per game for the previous period.

The home team’s score is calculated as the average of xG of the home team and xGA of the away team.
G_home=(xG_home+xGA_away)/2

The away team’s score is calculated as the average of xG of the away team and xGA of the home team.
G_away=(xG_away+xGA_home)/2

Goal Difference(GD) is calculated as the difference between the home team’s score and the away team’s score.
GD=G_home-G_away

Total goals scored(GS) is calculated as the sum of the home team’s score and away team’s score.
GD=G_home+G_away

The result of Gameweek 5 matches predicted is as below.

The higher the absolute value of Goal Difference, the more one-sided the match would be and the value of Goal Difference closer to zero means an evenly contested match can be expected.
The negative value of GD implies that the Away team is likely to win and the positive value favors the home team to win.
A higher value of Goals Scored implies that we can expect a goal fest from the fixture and a lower value indicates a low scoring match.
Based on the model, Burnley has the best fixture away against West Brom with a goal difference of 0.77 in favor of the Clarets.
Brighton’s game vs Crystal Palace is also expected to be a walk in the (Selhurst) park for the seagulls.
Man United’s visit to St. James’ Park is estimated to result in yet another heartbreak for the Red Devils.
Manchester City vs Arsenal match is the most evenly contested match of the gameweek with xG slightly in favor of the Gunners.
Chelsea vs Saints and Leeds vs Wolves are also going to be closely contested.
Mersey side Derby is expected to be the goal shower with the hosts having an edge over the defending champions.

Update: Prediction Vs Actual Result

Finally, it’s the moment of truth, it’s time to compare the predictions with the actual game-week 5 results.

Teams which over performed the predicted xG are above the line whereas, teams with underperformed with respect to the predicted xG are below the line.
Manchester United has over performed their predicted xG by very high margin. Their xG predicted was low as the sample size was just 3 games where they played Crystal Palace(one of the best defensive side), Brighton and Spurs (with 10 men). Their xG would improve as the season progress.

The predictions were in the right direction in only 3 games out of the total 10 games played which was expected as the sample size is very small, other factors not considered and of course due to the inherent uncertainty of the game.
Prediction was most accurate in the Crystal Palace Vs Brighton Game. The away team was predicted to dominate xG, dominate they did with 20 shots against 1 shot from Crystal Palace.
The predictions were highly fallacious in the case of Newcastle vs Man United, Everton vs Liverpool, Spurs vs Hammers and Leeds Vs Wolves.