We would never analyse, say, weeks 18 to 21 in the Premier League in such detail, so why can’t we resist placing so much importance on weeks one through four?
The beginning of the 2023-24 season in England’s top flight has been predictably relentless, with not just a flurry of fresh faces and new ideas, but four opening rounds that have seen 39 games, 127 goals, more than 1,000 shots taken and just under 33,000 passes completed.
With so much information to take in, it is tempting to make broad judgements to help digest it all. It’s why we like tables, a neat way to categorise the 20-team chaos into the good, the bad and the OK.
But at such an early stage, wins, draws and defeats can be deceiving.
Only taking into account the final score, the early table offers just four data points for each side. At such a low sample size, those readings are unstable, disproportionately affected by moments of fortune, quality or randomness, separate from the underlying quality of the team.
While still not a perfect science, it helps to zoom out and look at some more stable metrics — things like passes, touches and shots — which will settle down at a quicker rate.
Because if we are going to make some early-season assessments, we might as well try to look at something slightly more statistically sound.
Two such metrics are expected goal difference (xGD) and field tilt.
On the vertical y axis, xGD per game takes into account the underlying numbers at both ends of the pitch. Based on the quality of opportunities a team both creates and concedes, it produces an average, projected goal difference, to measure if that side can be expected to outscore their opponents in any given game.
An xGD per game close to zero, for example, suggests a side producing a similar number of high-quality opportunities to those they concede, while a positive xGD per game profiles as a team who consistently make more dangerous chances than they allow, and who should be more likely to win games.
On the horizontal x axis, field tilt indicates the percentage of touches in the final third (the end of the pitch they are attacking) that a team take, relative to the total number of such touches in their games.
GO DEEPER
The Athletic’s football analytics glossary: explaining xG, PPDA, field tilt and how to use them
Teams who record a higher field tilt will generally dominate the ball close to the opponents’ goal, while conceding fewer touches in their defensive third — it is a metric that points to sustained, potentially threatening possession in advanced areas, rather than just possession itself.
Plotting the early-season Premier League data looks like this:
The top right of the above graphic throws up two familiar names, and one surprise.
Throughout the first four rounds of this campaign, Manchester City have begun their bid for a fourth title in a row by creating a high number of goalscoring opportunities, conceding few at the other end, and dominating the ball high up the pitch.
Arsenal are not too far behind, while recording a slightly higher field tilt percentage (73.6) than the defending champions (72.2), outlining their ability to set up camp in the attacking third, although their home games against Nottingham Forest and Fulham have to be taken into account here.
But a third team in the mix are Chelsea, who have been dominant with the ball so far, with the highest field tilt in the division (74.5 per cent), but have been unable to convert their high xGD into points quite as consistently as last season’s title winners and runners-up. With Chelsea’s five goals from an expected goals (xG) number of 8.4 so far, only Everton have underperformed their expected attacking output by more. Carry on as they are, and things should start to click for Chelsea.
At the other end of the spectrum, and just down the road in west London, it has been a difficult start to the season for Fulham, the only club with an xGD per 90 minutes of -2.0.
In collecting two red cards along the way, Marco Silva’s side have played 12.8 per cent of their season so far with 10 men, and already faced City and Arsenal. They can certainly point to fixture difficulty and game state for their poor underlying numbers, having conceded more shots on target (33) than any other Premier League side, and a bigger xG figure (11.7) than any side in Europe’s big five domestic leagues.
Even so, spectacular goalkeeping from Bernd Leno has seen Fulham pick up four points from their four matches.
This next graph outlines Leno’s sustained overperformance when compared to the number of goals he has been expected to concede over the last calendar year:
So far for Fulham, it has been a season of extremes.
Matters have been more cut and dried for the promoted sides, with all three struggling to impose themselves on games and create chances. In this instance, the early league table is actually reflective of their problems.
Burnley and Sheffield United are the bottom two for xG per 90 this season, while Luton have scored only once in open play. Of the three, Burnley have been the most front-footed with the ball, but both match results and the underlying numbers demonstrate how hard the step up to the elite division is proving for last season’s Championship winners.
But teams such as Forest and West Ham United, and particularly Brentford, have shown us that field tilt isn’t everything when it comes to generating chances.
Forest have averaged just 26.7 per cent of passes in the attacking third, having been largely penned in by high-quality opposition in the form of Chelsea, Arsenal and Manchester United, all of whom they have faced away from home.
But with the top direct speed of any Premier League side this season, and the highest share of shots created from set pieces (39.5 per cent), they have been incisive on the counter-attack and resourceful from dead-ball situations.
✅ Taiwo’s first goal of the season
✅ Anthony’s first assist as a Red pic.twitter.com/dudCGhMst6— Nottingham Forest (@NFFC) August 13, 2023
West Ham have been similarly quick on the counter, and Brentford have been exceptionally defensively organised while at the same time averaging the division’s highest xG per shot (0.16). Whether they can maintain that quality as the opposition ramps up remains to be seen.
One more side to have defied the early standings are Everton, who were goalless and pointless until a 2-2 draw away to Sheffield United in their final outing before this international break — three games in which they were expected to score 4.7 goals, based on the quality of their goal attempts.
While, again, it’s too early to say that things will turn around for them, the underlying numbers suggest Everton haven’t been quite as bad as their solitary point suggests.
Their next two matches against Arsenal and Brentford are likely to put that theory to the test, though.
Having said all of that, the original point stands: don’t overreact to the league table just yet.
Don’t even react to these underlying metrics — it is all far too early to make many meaningful predictions on how the 2023-24 Premier League will shake out.
Even a considered combination of the standings, advanced metrics and the good old eye test, at this stage, leaves us with more questions than answers.
(Top photos: Getty Images)