Quick out the gates: who has started the Premier League season off well?

How long into the season is long enough to start doing data analysis?

The answer to this question is, in the current data-driven age of football, both pertinent and incredibly difficult to answer. Teams and fans clamour for statistics from the very outset of the season to assess whether their players and teams have started well, whether they have cause for optimism, or even if they need to finish off some shopping before the transfer window slams shut.

The rule of thumb that I have always been taught in performing data analysis is that a sample size of 30 is the absolute minimum for statistical analysis. The reasoning behind this is rooted in the mathematics of statistics that I will avoid boring my readers with, looking at the assumptions underlying the Central Limit Theorem and when a t-distribution becomes effectively normal etcetera, etcetera. Those who are particularly mathematically inclined may be interested, but I will task them with the research themselves.

Of course, the larger the sample size the better. Thirty pieces of data is trumped by having 50, and 50 by 100, and so on. If we want to find the “truth” behind data, the only way to truly accomplish it is by having infinite data. Best of luck downloading that file.

So if we have a minimum of 30 for doing meaningful data analysis, what 30 do we select? Matches? If we waited until a team had played 30 games, we wouldn’t look at data until next Spring! Touches of the ball? While players have had hundreds of those already this season, they don’t lend themselves to interesting analysis at the team level very often. After 3 matches, however, there is one statistic that has crossed the 30 mark for most teams in the league: shots!

To see how teams have been performing both offensively and defensively so far this season, I have created two fairly simple charts based on shots and expected goals by teams so far this season. Each looks at the amount of shots taken per match (either that a team takes for an offensive approach or that a team concedes for a defensive approach) and then looking at the average xG of those shots. From there, we can get a rough estimation of how teams are performing compared to the league averages, to see if they’ve been peppering the goal, trying to walk it in, or have looked a bit like Swiss cheese at the back. Let’s start with what we can learn from the attacking side of the game so far.

The Attack

What can we see?

Steep Learning Curve – Nestled down in the bottom right of the chart are all 3 newly promoted teams. While none of them have had the easiest opening schedules with Burnley and Sheffield United having played Man City while Luton have travelled to Chelsea, there is still some cause for concern here. They will no doubt begin to sort things out, but they need to start creating chances at a better clip.

Reports of my death have been greatly exaggerated – Everton, who are yet to score a Premier League goal and have been predicted by many to face the drop this year, are performing fine? They are ever so slightly above average both in shots taken per match and their expected goals. Mean reversion says that the goals will start to go in soon, so don’t be surprised if they end up with a half decent run of attacking form soon. Similarly Wolves, who switched their manager a mere three days before the start of the season, have the same xG per shot as Chelsea, Newcastle, and Manchester United, and more shots per match than Chelsea, Newcastle, Liverpool, and Arsenal. These teams might need a second look.

The model’s working – Brentford are a team that is almost entirely built on xG. So it’s truly unsurprising to see that they have the highest xG per shot in the league so far this season. They craft their attacks to maximize that figure and it seems to be working. Not only that, but it seems Villa and West Ham have caught on.

Where is the best place to be? – On initial looks, Manchester United are doing best this season. High numbers of shots for relatively high xG is, on paper, where you want to be. But then again, the teams that have been receiving rave reviews to start the season all seem to be below average in xG per shot (Arsenal, Liverpool, Tottenham, City, and Brighton). My theory? The teams in the bottom right are facing low blocks more frequently, and so are forced to take their shots from worse positions. Maybe teams think that they can take the game to Chelsea and United more now.

The Defense

What can we see?

The Impenetrable City – Fewest shots faced, second lowest xG per shot. People talk about Haaland and De Bruyne, but this is the foundation on which the City empire has been made. No sign of slowing down.

United-Wolves makes sense now – People did say after the first Monday Night Football of the season that Manchester United against Wolves was a very fun match to watch. I now see why. Both teams high in both taking and conceding shots so far this season, and they tend to be for higher xG on both sides of the ball. Neutrals will love it, but United fans will be wanting this to change as soon as it can.

Brighton’s scouts should be watching Brentford – The Seagulls and the Bees couldn’t be farther apart in this chart. Brighton don’t seem to be conceding that many shots, but those that they do tend to be tremendously good chances. No wonder West Ham managed to pick them apart last week. On the other hand, Brentford show the same DNA as they did in the attacking chart, with the lowest xG per shot conceded in the league. That combined with below average shots conceded per match is the perfect recipe for a lot of clean sheets this year.

Merseyside blues? – Reasons to be hopeful coming out of Everton yet again. While it feels uncharacteristic for a Sean Dyche side to be conceding such high quality chances, their quantity is fairly good. While it isn’t perfect, it’s far from the disaster that many make it out to be.

---

So there we have it. The first glimpses of what this season might have in store for the twenty clubs vying for the Premier League title. We will no doubt be revisiting these charts as the season goes on to see who’s trending well and who drops off. Of course, it is far too early to read too much into this data and the charts could look different even by this time next week. But with every shot taken, the picture of this league gets that little bit clearer.