After the first three matchdays of the 2025-26 Premier League season, we open* The xG Files *to break down the numbers behind the early performers and anomalies.
Football has changed.
The language and data literacy of the football world has transformed in the past 15 years or so. And yet, sceptics remain. For some, the power of age-old clichés and platitudes is steadfast, with the advancement of data and analytics merely the realm of “nerds” or sport’s equivalent of conspiracy theorists.
But xG isn’t going away.
The xG Files are for those who want to dig a little deeper, who want to believe in data that scratches beyond goals, assists and clean sheets, to get a better understanding of player impact.
Over the course of 2025-26, Opta Analyst will be football’s equivalent to Mulder & Scully as we bring you The xG Files, delivering key insights from Opta’s various xG-related metrics in the Premier League, helping you get a truer idea of who’s excelling in front of goal, creatively or in goal – and who isn’t.
The truth is out there, and The xG Files will provide it (well, at least provide the information for you to make your own conclusions…).
Close Encounters
The obvious caveat for the first edition of The xG Files is that the 2025-26 season is only three matchdays old, so the sample sizes we’re working with aren’t huge.
Nevertheless, many players have made multiple appearances, meaning their statistical outputs for the season are beginning to take shape.
But for all the desire to see The xG Files uncover surprising truths, it probably won’t come as a big shock to learn who’s leading the way for expected goals at this early stage.
Erling Haaland has recorded shots worth a total of 3.86 expected goals (xG), which is almost double that of any other player in the Premier League this term; Ismaïla Sarr is next with 1.98, followed by Bruno Fernandes (1.8).
Erling Haaland xG map
Fernandes drops out of the reckoning entirely when penalties are removed, though. Haaland and Sarr haven’t taken any spot-kicks and so remain first and second, while Antoine Semenyo (1.8) moves up to third after a lively start to the campaign.
Sarr’s output requires additional exploration, however. Crystal Palace’s Senegal international has had only three shots this term, meaning the average xG of his three attempts is 0.66, the highest of all Premier League players in 2025-26.
A penalty carries an average xG value of 0.79, so the average across his three shots isn’t far off that despite him not taking a single spot-kick.
All three of his attempts have been within nine metres of goal, scoring via a controlled volley on Matchday 2 and a close-range, back-post header on MD 3. He also nodded over from inside the six-yard box on MD 3.
Ismaila Sarr xG map
Either Sarr has a shrewd knack for finding dangerous shooting positions or something more mysterious is at work. We’re not saying aliens, but we’re not not saying it either.
Clubs in Focus
Of course, xG isn’t just relevant on a player level. We can use it to get a better understanding of how effective teams are in a more general sense.
Manchester United have actually generated the most xG this season (6.8), though they have had two penalties. So, it might be more pertinent to exclude spot-kicks to get a more accurate reading of how effective teams have been at creating good opportunities.
Ruben Amorim’s men are still relatively high, managing 5.2 non-penalty xG over the first three matchdays. That puts them behind only Chelsea (5.9) and Manchester City (5.8).
While it is worth saying United’s is boosted by 2.8 non-pen xG against Burnley, Chelsea (2.7 non-pen xG vs West Ham) and Man City (2.5 non-pen xG vs Wolves) are also helped by hefty single-game outputs. But that’s simply the reality of looking over just three matchdays.
Man Utd non-pen xG
But what about xG in specific phases?
Well, again Man City (4.6) and Man Utd (4.1) are up there for xG in open play, ranking first and second respectively. While Pep Guardiola’s men have scored five times from those chances, Amorim’s side have netted just once; no other team have underperformed in relation to their open-play xG to a greater extent than United (-3.1).
Considerable work to be done for Amorim, then. The next two episodes of United’s season won’t be strolls in the park either, as they face City (A) and Chelsea (H).
Speaking of the Blues, we’ll move on to threat from set-pieces.
Enzo Maresca’s men have been the most dangerous from dead-ball situations over the first three matchdays. Not only have they generated the most xG from set-pieces (2.6), they’ve also scored more than anyone else from such situations (4, excluding penalties).
Chelsea set-piece xG
Arsenal, the unofficial kings of set-pieces, aren’t too far behind though. They’ve scored three times from 2.3 xG at set-pieces, so the Gunners’ dead-ball threat certainly appears to be alive and well during these early weeks of 2025-26.
Mikel Arteta’s men have been rather miserly at the back at set-plays, too. Their 0.4 xG against from dead-ball situations is the third lowest in the Premier League this term after Newcastle (0.1) and Sunderland (0.3).
And at the other end of the spectrum? Man Utd, and that won’t surprise United fans.
They’ve looked incredibly shaky defending set-pieces this term, with goalkeeper Altay Bayindir routinely panicking like he’d just seen the Cigarette Smoking Man lurking at the back post (he was one of the main antagonists in The X-Files, keep up).
Their 2.4 xG conceded from set-pieces is comfortably the most in the Premier League this term, and although this is helped by Riccardo Calafiori’s point-blank-range goal on MD 1, it does feel like this tallies with the eye test and United’s generally unconvincing approach to defending set-pieces – corners in particular.
We can also use club-level xG data to take a hypothetical look at how the Premier League table might look if every match went the same way as the xG battle.
Premier League expected points table 2025-26
Don’t take this too seriously. Again, we are working with a limited sample size and xG data doesn’t take things like game state into consideration, but it gives some indication as to how well (or otherwise) teams are performing.
Creationism
From the chaos of set-pieces to the calculated brilliance of creators, we shift focus to creativity now – talk about signs of intelligent life…
xG-related metrics don’t just tell us about shots. We can also use variants to measure chance creation through expected assists (xA) and xG assisted.
While the two metrics are quite similar, there’s a subtle difference between them. xA measures the likelihood that a given pass will become a goal assist. So, the model rewards players who pass into dangerous areas, regardless of whether the receiver takes a shot or not.
xG assisted, on the other hand, is the total xG of the shots taken following chance-creating passes, therefore rewarding players who find teammates in positions to get high-value shots away.
Much like with xG in the previous sections, there are the usual pitfalls that come with a small sample size, but there’s still enough to have fun with and draw insights from.
For instance, it’s been a hugely promising start to 2025-26 for Jack Grealish. For many, there was a perception that life at Manchester City sapped the fun out of the England international’s play, but he’s looking re-energised at Everton, even winning the Premier League Player of the Month award for August.
The headline fact for him has been that he’s already got more assists (4) than he managed across the previous two Premier League campaigns combined (2). Three of those have come in open play, and he leads the way this term for open-play xA with 1.3.
Jack Grealish xA
Of course, that does mean he’s overperforming in relation to his xA, which can be partly explained by his assist for James Garner’s long-range strike not being particularly threatening. But even so, no Premier League player can match his 1.3 open-play xA despite plenty managing more than his 196 minutes on the pitch.
Bournemouth’s David Brooks (1.2) is the closest to Grealish, while his Everton teammate Kiernan Dewsbury-Hall (1.01), Bryan Mbeumo (1.02) and Estêvão (1.03) are the only others to reach 1.0 open-play xA.
Grealish also leads the way for xG assisted (1.6), undoubtedly partly influenced by the header he played across goal to tee up Beto for a close-range goal against Wolves.
Brooks (1.3) is up there again as well, but between him and Grealish is Palace’s Daniel Muñoz (1.5). The Colombia international has only created two chances this season, but they were both played into very dangerous positions; the first was a low cross for the aforementioned Sarr volley against Nottingham Forest, the second was a slightly higher delivery right across goal that – again, as mentioned before – Sarr headed over vs Aston Villa.
Daniel Munoz chances created
Muñoz was a revelation for Palace last season, with his relentless running, offensive mentality and final-third effectiveness making him a huge asset. The early signs are he could be a menace once again in 2025-26.
Stopping the Unstoppable
And finally, we come to the goalkeepers. While xG might ordinarily be used in relation to the taking of shots, it can be adapted to measure the effectiveness of keepers as well by using expected goals on target (xGOT).
xGOT measures the likelihood of an on-target shot resulting in a goal, based on the combination of the underlying chance quality (xG) and the end location of the shot within the goal. It gives more credit to shots that end in the corners than those that go straight down the middle.
The fact xGOT only measures on-target shots is crucial, because it means the sole factor preventing these attempts being scored – beyond paranormal activity – is the goalkeeper. So, a shot worth 0.3 xGOT has a 30% likelihood of being scored and a 70% chance of being prevented by the keeper.
As such, it allows us to more accurately credit goalkeepers for their ability to prevent goals with their saves instead of forcing us to rely on vague metrics like clean sheets, which obviously favours keepers at the better teams.
The goalkeeper leading the way after the first three matchdays of the season is Tottenham’s Guglielmo Vicario. According to xGOT data, his saves have prevented 2.2 more goals than would’ve been expected based on the quality of shots on target he’s faced.
He faced 14 shots on target across the first three matchdays, conceding just one goal.
Guglielmo Vicario xGOT
Manchester City’s James Trafford is the next highest on the list, with his saves deemed to have prevented 1.9 goals more than the average goalkeeper would have been expected to save based on the quality of shots on target faced.
And yet, he’s probably about to lose his place in the team to Gianluigi Donnarumma following the Italy international’s arrival from Paris Saint-Germain. However, we probably ought to mention that Trafford did also make a big mistake that led to a goal in the defeat to Tottenham, so any praise should be tempered.
Neither Vicario nor Trafford have had their figures boosted by penalty saves either; they’ve just simply made difficult saves, with the former impressing particularly against Bournemouth and the latter making a string of stops at Brighton.
On the flip side we have José Sá. According to xGOT data, the Wolves goalkeeper should have conceded 5.8 goals on average, but he’s actually let in eight. So, that’s an underperformance of -2.2.
This will be partly influenced by a couple of goals against Man City that he might feel he could’ve done better with, though we should also provide a little balance here. Sá has conceded three goals from shots that were within eight metres, which are obviously going to be tricky to stop regardless of whether they’re heading for the corners or not.
However, there’s plenty of time for variance to work its magic in all areas of the game before we start suggesting anything more sinister.
It’s early days, but The xG Files remain open.
Premier League Stats Opta
Subscribe to our football newsletter to receive exclusive weekly content. You should also follow our social accounts over onX, Instagram, TikTok and Facebook.******