Measuring Changes In Attacking Style In The Premier League

Back in November we applied a clustering algorithm to find out which Premier League clubs had similar attacking styles. We wanted to see what we could find using match summary stats that anyone with an internet connection could get hold of. Our main rule was that we wanted to avoid using pure outcome stats, e.g. shots on target, completed passes, completed crosses, goals, assists etc. We thought we'd run the risk of just clustering teams together on how good/lucky they'd been so far. We didn't use anything too fancy, just per game stats based on the way teams attempt to attack; shots from outside the box, inside the box, open play, set pieces, short passes, long passes, dribbles, crosses and how much they use the wide areas when they attack.

We were happy with the results, for a start they made sense, 6 of the top 7 teams at the time were in the same meta-cluster and were generally teams you'd expect to spend matches on the front foot. So this was a good first step but we knew at the time the real test would be to use exactly the same method later on in the season to see whether a) we got similar, stable clusters and b) to see how attacking styles had changed over the season. You can see in the picture below how attacking styles looked in November (after 11 games) and now (after 22 games). We still get 6 main attacking clusters and most teams have kept a similar attacking style across the season but some have changed and we discuss this below. You can see the main diagram comparing attacking style clusters below and we discuss cluster changes in the rest of the article. A quick note is that the heatmap summaries now include xG per game but this wasn't entered into the clustering model.

Updated attacking clusters from November to January

The first thing to notice is that in both the left (November analysis) and right (January analysis) cluster trees we see at the start everything splits into 2 branches. The best way to describe this is the analysis immediately spots a ‘Top 6’ and ‘Rest of table’ style difference. This might provide a bit of an explanation for why we don’t see ‘clear’ mid table teams anymore (as discussed in this article.) It seems a lot of it could be explained by attacking approach.

(Most of) the top 6; Chelsea, Spurs, Liverpool, Arsenal and City

Clusters 1 (98% certainty) & 2 (91% certainty) -

(average xG per game= 2 , average points per game= 2.1)

Unsurprisingly the top teams here are pretty similar in that they all play attractive front foot football; a lot of short passes, dribbles, crosses and a lot of attempts on goal. They can be split into smaller clusters though on one main stat, City and Arsenal don't generate nearly as many shots from set pieces as Spurs, Chelsea and Liverpool.  Maybe a missed opportunity? (Not that it's affecting City too much.)

Out of these clusters City and Liverpool have probably changed the least, their figures are around the same as they were in November except they are having more shots from range. Perhaps that's due to teams moving ever further towards a strategy of setting up to be compact and deny space inside the box. This has worked somewhat as City's xG per game has dropped slightly. Liverpool haven't changed much either except they've also shown less focus on playing wide and their xG hasn't changed much either.

Over the last couple of months Spurs appear to have had slightly less of a focus on width and dribbling and played fewer long balls, they haven't been mixing up their passing as much as they did earlier this season. Arsenal have changed in the opposite way, they've slightly increased their number of long balls, crosses and dribbles per game implying that they have been slightly more direct, and while it has increased the number of shots per game that they are having it hasn't shifted their xG per game much.

Chelsea have been the standout changers within this cluster, after a slow start they've shown increases across pretty much every attacking stat. They have played more short passes, crosses and generated more shots in all situations and distances. Their increase in xG and rise in league position appears to well reflect this sharpened attack.

The Manchester United - Southampton (and Palace) Axis

Cluster 3 (96% certainty)

(average xG per game= 1.5, average points per game= 1.35)

Manchester United and Southampton belong to the same cluster together once again but now Crystal Palace join them. Although United and Southampton don't produce a huge number of attempts from open play they have both done quite well from set pieces, Apart from that their play is fairly balanced between the wings and central areas and they have a slight focus on short passes compared to the league average. Within the cluster Southampton have seen a drop in efforts on goal per game and Manchester United have also seen a small drop in close range efforts on goal per game, while both have shown a drop in xG. Palace on the other hand have shown an increase in xG.

Direct and Wide: Brighton, Huddersfield, Swansea, West Ham and West Brom

Cluster 4 (98% certainty)

(average xG per game= 0.9 , average points per game= 0.92)

Most of the teams here focus on playing wide, playing long passes and generate more shots than average from set pieces. Despite a wide focus, they don't put a lot of crosses in and generally don't attempt many shots per game whether from close or long range.  Although their attacks aren't generating a lot of shots, Brighton, Huddersfield and West Brom are having slightly more shots from open play and set pieces than they were in November so they are moving in the right direction. Brighton are showing more of a focus on playing wide than they were earlier this season. One thing that unifies these teams is that although they've all seen an increase in xG per game since November none has a prolific attacking threat, they are all at least a good finisher away from safety.

Flexible Passing and Balanced Width: Bournemouth, Newcastle, Leicester, Watford and Stoke

Cluster 5 (90% certainty)

(average xG per game= 1.1 , average points per game= 1.07)

The core of this cluster has stayed the same as it was in November: Leicester, Watford and Stoke but now Bournemouth and Newcastle have joined. They are similar to the teams in cluster 4 in that they focus on more direct passing but they generate slightly more shots per game and focus slightly less on playing wide than the teams in cluster 4. Most of these teams haven't changed a lot since November apart from Newcastle have been attempting more dribbles and long passes than they did earlier this season and Bournemouth who are generating a lot more shots in all situations since November. Perhaps that has something to do with the return of Callum Wilson?

Defensive teams: Burnley and Everton

Cluster 6 (83% certainty)

(average xG per game= 0.9 , average points per game= 1.39)

It's fair to say that the clustering of these two teams together possibly depends more on the defensive statistics that we haven't included in the analysis than the attacking stats that we have. In fact both (along with Leicester) are the only top half teams with negative expected goals per game differences. Both teams play a high proportion of long balls and don't generate a lot of shots, but the ones they do generate they seem to use well. Quite a big surprise is that both teams are below average on shots from set pieces and considering they look safely in the top half of the table already, if they added goals from set pieces to the mix then maybe we'd be able to enjoy watching Barcelona vs Burnley in the Champions League next year.

How much cluster change from November - January?

If the clustering method we used can find changes in attacking style then we'd expect it to see some changes in the flowchart below from November to January although if the clusters are reliable then most teams will stay together in clusters (unless every team undergoes extreme tactical changes). The diagram shows how teams have changed in similarity of attack since November. The main thing to notice is that generally the same teams that clustered together in November cluster together now. There have been a couple of big changers style wise, Palace are now in the same attacking style cluster as Man United and Southampton rather than more direct lower table teams. Everton now occupy the same attacking style cluster as Sean Dyche's Burnley matching both managers extreme pragmatism.

A focus on teams that changed manager

It's worth focusing on the teams that have changed manager so far his season because if anyone is likely to have drastically changed styles it would be them. We're not including Swansea or Stoke in this discussion as they only changed managers a couple of games ago but the others are summarised below.

Everton stand out because the Big Sam effect has meant that they've toned down their whole attack showing reductions in everything across the board apart from the fact that they now play more long balls and play wider. Looking at the cluster diagram they now are most similar in style to Burnley who have also been effective at converting the few chances they create in every game.

Crystal Palace have hit a decent run of form since Roy Hodgson has come in and they're now more similar in style to Man United and Southampton, this has mainly been based on the increased number of crosses they've been putting in and a slightly increased focus on playing wide, dribbling and putting in more crosses this has coincided nicely with Wilfred Zaha finding form again over the last 5 or 6 games and in general providing much more of a goal threat.

West Ham also haven't changed too much in attack since Slaven Bilic left although under Moyes they don't play as many long balls or have as many shots from long range. Interestingly this has coincided with an increase in xG per game, their increased trust in their creative players and playing through midfield seems to have paid of form wise and xG wise with the most noticeable difference being the improved performances of Marko Arnautovic.

Unsurprisingly since dumping Tony Pulis West Brom have shown an increase in attacking stats across the board. This is mainly reflected in the increase in the number of crosses they have and shots they generate per game but this has meant they've only had a slight increase in xG per game. Despite this they have yet to show much cutting edge in terms of finishing and as such they still belong in pretty much the same attacking cluster as they did earlier this season with Swansea, Brighton and West Ham.


Of the five teams who changed manager at least a month ago only Palace and Everton have massively changed their style of attack. There's probably a bit of a bias here because most of the managers that were bought in came in at teams near the foot of the table who were shipping goals and their priority was to stop that happening. Although Roy Hodgson gets lumped in with the ‘safe pair of hands’ managers, he seems to have focused on improving attack as much as defence the most.

This can be seen both in Crystal Palace's rise through the table and the fact that they have the fourth best improvement in xG per game after Chelsea, Spurs, and West Ham and are now most similar in attacking style to Man United and Southampton. At first when we saw this cluster we were skeptical but then we noticed that Crystal Palace and Southampton are the biggest under-performers when you compare goals scored to actual expected goals scored, and that they should have scored around 11 goals more than they have and Southampton are around 5 goals below par.

Overall the clustering method we used has shown that while styles are stable and can be identified at different points throughout the season the approach is also sensitive enough to pick up on slight changes.   ____________________

Written by Mladen Sormaz and Dan Nichol who together are Proform Analytics

La Liga At the Break: Valencia, Girona and Atlético

In just a scant few days the La Liga teams will return from their truncated winter rest. Before we dive head first into the home straight of the season, let's take a look back at what the first chunk of games has taught us, shall we? This isn't technically the halfway-point as Spain's domestic break comes after 17 games rather than 19. Don't worry though, we're not about to let that get in the way of a good ol' rummage through the numbers!


From 2013 to 2016 Villarreal had a pretty successful run of seasons under then manager Marcelino, culminating in achieving Champions League qualification and reaching the Europa League semifinals in 2015/16. That summer - before they had even played their CL qualification match - vague tensions between Marcelino and the club’s higher-ups lead to him getting the sack out of nowhere. He took a sabbatical for the 2016/17 season and now is back in La Liga, taking charge of Valencia. At this stage they sit 3rd in the table with the 2nd best goal difference. After chewing through five managers in two seasons, Los Che appear to have found their man.

Marcelino’s Villarreal built their success on a rock-solid defensive 4-4-2 shape, somewhat like a more reserved version of Diego Simeone’s Atlético Madrid. In his final season they ranked 3rd in terms of expected goals against and a pretty uninspiring 17th in terms of attacking xG. At Valencia things appear the same on the surface - the 4-4-2 is in use again - but the attack/defence split has been shaken up somewhat unexpectedly.



Functionally the attack is, again, closer to an Atléti than a Barca. Meaning that getting the ball forward with pace is the order of the day. They're just about middle of the pack in terms of possessions that are 10 or more passes in length. Conversely, they have the fourth highest percentage of their possessions leading to shots that are 15 seconds or shorter. This continues a trend from Mareclino's last season at Villarreal where they lead the league in this measure. However, in Valencia, the efficacy is greatly improved with them getting far more shots overall from these situations.


Also: Real Madrid *eye emoji*

While the attack is clearly humming, a word of caution: it's perhaps running a tad too hot. They're over-shooting their expected goals numbers right now and come season's end may fall back behind Real Madrid to about 3rd or 4th in terms of goals scored (unless this bizarre Real Madrid finishing slump just continues in perpetuity). Simone Zaza - who has genuinely been quite good this season after a couple of odd detours in his career - is the one to point to here. Regardless, things are encouraging.



Defensively there a couple of slight causes for concern. The relatively high amount of shots they're giving up isn't necessarily a problem as long as the quality of said shots is as low as it has been (45% of their shots conceded are from outside the box, the 2nd highest percentage in the league). However, it is a bit easier to move the ball against them than Marcelino would probably ideally like. Opposition possessions are reaching Valencia's defensive third at the 5th highest rate in the league. On top of that they're not doing especially well at keeping opponents out of the most dangerous areas. The xG says don't stress it but it's worth keeping an eye on.



Marcelino's boys likely aren't, as the early results have suggested, on a level with Barca and Real Madrid. However, the signs do point towards them being the league's third best, even beyond an Atléti (more on them later). That is plenty impressive on its own.



...are the surprise package in La Liga this season. Prior to this campaign most probably knew them as the destination for a lot of Manchester City's loan army (about a third of Girona's squad are loanees, five of which are from City). That doesn't define them though, they're a hoot in their own right. The newly promoted club sit 9th in the table at this point of the season, and look just as good - if not a bit better - in the underlying indicators.



Manager Pablo Machín has them at a high level of defensive organisation all over the pitch. As you may have seen above, they allow one of the lower proportions of opposition completions in the final 18 yards, pointing towards a sturdy shape near to their own goal. They engage pretty high up the pitch too, forcing the 6th lowest completion percentage on passes originating in the opponent's defensive third along with the 4th lowest proportion of opposition possessions that are 'long' (10 or more passes). Put this all together and you see them keeping opponents out of their third at a super impressive rate.



On the other side of the ball there's two big defining stylistic points: 1) they've gotten the joint most goals from set pieces in the league (9) - a route of value you'll hear reiterated time and time again on this here website, especially for promoted teams - and 2) they love a headed shot, racking up the 2nd most of those, with the 2nd highest proportion of their completed passes into the opposition box coming via crosses. Other than that, there's no one regard in which their attack sticks out. They just do a little bit of everything. Long possessions per game? Middle of the pack. Counter-attacking situations? Solid there too. Shots from turnovers in opposition's third? They do them quite well. You get the gist.

Lots of players deserve praise here but the most eye-catching has been 20-year-old right-back Pablo Maffeo. The City loanee has shown real creativity going forward as a wing-back and generally just a nice all-round game, handling himself very well in marking assignments on bloody Lionel Messi and Cristiano Ronaldo. Opponents complete their passes in his zone (the right side of Girona's defensive third) at the 3rd lowest rate in the league. He looks ready to slot in back at City, should Guardiola see fit to do so.



Someone might want to talk to him about the shots though. Oy vey.



Atlético Madrid

On the face of it there's little wrong with Atlético Madrid this season. They currently sit 2nd in the table, only having just suffered their first league loss of the campaign in the final match before the break. Everything hunky dory then? Well, peer below the surface and things start to look a tad iffy.



The defence is still looking about par but the attack has dropped off to a quite alarming extent. It's like Marcelino has sapped all the offensive capability out of Madrid and directly into his Valencia team.

41% of their shots came from outside the box last season, that’s up to a league-leading 47% this time around. The root of this appears to be in their transition game. Atléti's deep and compact defensive shape is the yin that require's a counter-attacking yang. Otherwise you're just defending all the time. Right now, they're struggling to get out.



On top of that, the shots they're getting from turnovers in the opposition's third is down by about half a shot per game. Possibly most troubling of all is that they've only managed 3 goals all season from set pieces, one of the worst tallies in the league. Something that was previously their bread and butter. All in it's a dual issue of both shot volume and quality. Even Antoine Griezmann, whose individual volume is about the same as last season on a per 90 basis, isn't producing a great deal.



They've certainly improved their passing game over the years, to the point where now they're about average in how many shots they squeeze out of possessions that last 10 passes or longer. That'd be A-OK if everything else were working normally but it just isn't. With the points banked already maybe it won't be such a big deal. Simeone's obviously a fantastic manager and he might potentially cure what ails them. And of course the impending return of Dastardly Diego Costa will be a boon. If it doesn't turn around though, and the results don't keep going their way, it could be an underwhelming first season in the new stadium. __________________

Thanks for reading. You can find me @EuanDewar for any questions or the like. Enjoy the rest of the season!

How Good Is Malcom?

Ah yes, we've approached the silly season that is the January transfer window, where the rumors are hot and heavy. One of the most talked about names this time around has been Malcom, the audacious winger who has been scoring golazos left and right this season. I talked a little bit about Malcom in late September when Bordeaux started the league strong and looked like a potential surprise package in Ligue 1 this season. Since then, both he and his team have slowed up somewhat.

Despite those issues, Malcom has had a solid season on a flawed team. His non-penalty goals + assists rate is 0.61 per 90, and his shot contribution (shots + key passes) per 90 rate of 5.14 is good too, especially for a 20 year old wide player in his 2nd full season in European football. Compare that to other wide players not from PSG (cause you know, playing on a super team can distort your numbers if you’re good enough), and he stacks up relatively well.

  • Memphis Depay: Minutes = 1295, NPG+A per 90= 0.83, Shot Contribution per 90= 5.07
  • Lucas Ocampos: Minutes = 960, NPG+A per 90 = 0.66, Shot Contribution per 90 = 3.94
  • Florian Thauvin: Minutes = 1606, NPG+A per 90 = 0.79, Shot Contribution per 90 = 6.61
  • Thomas Lemar: Minutes = 1020, NPG+A per 90 = 0.53, Shot Contribution per 90 = 4.50
  • Rony Lopes: Minutes = 1236, NPG+A per 90 = 0.44, Shot Contribution per 90 = 3.57
  • Malcom: Minutes = 1472, NPG+A per 90 = 0.61, Shot Contribution per 90 = 5.14

Considering that every other player mentioned here is at least a year older, it makes sense not to talk about Malcom on the level of Kylian Mbappe or Ousmane Dembele, but perhaps the level below alongside the likes of Yannick Carrasco during his 14–15 season or Bernardo Silva. That still equates to a very tantalizing prospect.

So we know that Malcom has been quantitatively good this year, but what about qualitatively? What are his strengths and weaknesses as a player?


If I were a team that was scouting Malcom, I would look at this sequence and get a little excited.


The ability to effortlessly beat his opponent in a 1v1 situation, and having the coordination to hit a fairly difficult pass while at top speed is not commonly seen. And this isn’t just a one-off, there's evidence to suggest that Malcom does a pretty decent job at finding teammates with these type of passes, whether it be in transition or against a set defense.

Obviously, you’d have to construct a plan where this type of talent would be used for the good for the team, but having such a dynamic playmaker in transition with players making runs off the ball would be a fun sight. He also profiles as a potentially good crosser, with his ability to hit passes both to the back post and the near side. Considering how dominant he is with his left foot when it comes to shooting, supplementing that with crosses from either foot adds value too.

One other thing that's fun about Malcom is how he gets the ball from one zone on the pitch to another



It’s fair to wonder how much value this would hold on a better team because there would be less opportunities for this type of sequence to occur, but even if so, there’s enough evidence to suggest that Malcom is good at beating his opponent in 1v1 situations in higher positions.


It’s easy to see why so many people are enamored with Malcom. Speed demons who flash the ability to be high level playmakers don’t exactly grow on trees, and when harnessed in the correct manner, you can see how he would be a genuine asset.


With inverted wingers, you’re always running the chance that the added freedom to cut inside will result in shots being taken from ineffective areas. There will be times where taking a 3–4% shot is the best option because there’s not even an option to recycle the ball to a nearby teammate and keep probing for another shot. As much as one can loathe the idea of settling for bad shots, there is a level of sympathy to be had for situations where you got nothing else but to take a bad shot and hope for the best.

But when your xG/shot is in the single digits, that excuse runs thin. To put it frankly, Malcom has bad shot discipline. The only reason why it’s not been talked about more often is because he’s converted 15% of his shots from outside the penalty area (4/27). And while people remember the goals, the misses stack up too:

Again, this is a little bit of a two way street. Bordeaux aren’t great in terms of presenting options for a player in possession, and with the burden that’s placed on him in attack, It's understandable that he’s going to probably be on the low end of shot quality on average. However, he could stand to cut around ~0.5 shots per 90 and perhaps redistribute it to someone else or even just improve his own shot selection. He faces the same problem that plagued Adam Ounas during his days at Bordeaux, where you felt that he was leaving stuff on the table by settling for bad shots.

With being fairly left footed dominant when it comes to his shooting, there might be situations where he can be shaded into less advantageous areas on the pitch. That’s easier said than done because his speed is overwhelming, and even if that does happen, he’s shown to be able to deliver passes on his right foot. Also, as much fun as it is to see Malcom drop deep and tries to advance the ball by himself, one could question how much he’ll get to do it on a team that has something resembling structure. That level of freelancing isn’t all that prevalent, so it’d be curious to see how a soloist would adjust to those situation. Perhaps leveraging his speed into more off-ball runs into the penalty box.

And then there’s obviously the question of whether Malcom would be able to do his defensive duties at a more demanding level if given a lesser role in attack. Nobody knows if that’s the case, but it’s something to keep an eye out on.

Valuation Discussion

We’ve already gone through what he can and can’t do at his current state. He’s an exciting player but has flaws too, and it could very well be that his solid results are masking the actual problems in the process. Saying that though, it wasn't totally surprising to see that Bordeaux are asking for as much as £50M considering that he's their crown jewel and his contract doesn't end until 2021.

To some extent, there’s a comparison to be made with Sofiane Boufal when he was at Lille. Both were pass first wingers, with an excellent first step + dribbling abilities to go along with their questionable shot selection. They both had to do the bulk of creation on the team because no one else was good enough to help out in that department. And the statistical comparison sort of makes sense.

  • Sofiane Boufal 2015–16: Minutes = 2351, NPG+A per 90 = 0.50, Shot Contribution per 90 = 4.67
  • Malcom 2017–18: Minutes = 1472, NPG+A per 90 = 0.61, Shot Contribution per 90 = 5.14

The comparison isn't foolproof. Boufal was compared to Eden Hazard when he broke through because of the vague similarities in playing style and the way they moved on the pitch, while Malcom’s pure speed and athleticism is on a higher level to go along with his ability to get the ball from one zone to another. Also, as bad as Malcom's shot selection has been, Boufal’s was even worse back in the day. Add to it that Boufal was 22–23 when he had his full season at Lille, whereas Malcom isn’t turning 21 until February. Perhaps by this time in 2019, there have been incremental growths in his game and he's a better rounded player. Malcom’s season so far has been better than Boufal, and Boufal went for a base fee around £16M. That was also in a climate that was far less silly than what we have currently. If 2015–16 Boufal was in today’s market, even with Lille’s awful financial situation at the time, he could’ve fetched more than he did.

Given all the information already provided, along with Bordeaux not being the strongest financially as a club, I would feel somewhat comfortable paying in the region of £25M. I would probably begrudgingly accept paying up to £30M if that included add ons or an agreement to keep him in Ligue 1 until the end of the season. Anything appreciably above that, and I would start looking at other options. Malcom has some special gifts with the effortlessness he has in getting by his marker, the fact that he is a left footed shooter in attack makes him something of a rarity in European football. and he’s still young enough to teach the value of shot discipline before it’s too late (Memphis at Lyon is the poster child for shot selection reclamation). While I'd be hesitant to say that he’s a potential world beater, there’s reason to think that there's high upside and he's young enough to learn and improve.

Where would the best place be for him? The likes of Arsenal, Manchester United, and Tottenham have been linked over the past week or so. Arsenal do have potential openings with Sanchez/Ozil being 6 months away from leaving on frees, but he’s nowhere near the playmaker that Ozil is, nor the high usage inside forward that Sanchez has been. Manchester United need a creative type on the wing and just as you get yourself excited at the idea of Malcom joining the attacking talent already in place, you remind yourself that Jose Mourinho is their manager and all those good vibes drift away.

Tottenham would probably be the best place out of the three clubs. It’s not the cleanest fit, and a fully functional Tottenham means that he would be a rotation member rather than a starter over the next year or so, but he’s someone who would represent one of their better chances at finding a high upside attacking player without having to guarantee him starter minutes or wages that would unbalance their current structure. There’ll be some who look at the likes of Georges-Kévin Nkoudou and Clinton N’Jie and do a reflexive gag at the idea of buying another Ligue 1 attacker, but that train of thought doesn't hold much water in this case. There wasn’t anything to suggest that Nkoudou was more than a tricky dribbler at Marseille, and N’Jie was an inside forward/second striker having to masquerade as a wide player at the time. N’Jie was in some way’s a poor man’s Son Heung-Min, but that role wasn’t available. Even with the flaws in his game, there's more evidence to suggest that Malcom would fit in better than either of those two did at the time, with his profile as a dynamic playmaker.

Malcom has been a productive wide player, and there are certain things he can do that get you excited about his potential. He is incredibly quick and at best can leverage his gravity as a threat into finding teammates with dangerous passes. He also probably needs to be coached even harder about the value of shooting in better areas, and being more efficient as a player. While not the absolute top tier prospect that Ligue 1 has seen over the past few years, he’s in the discussion at the next level, and is definitely one of the more intriguing ones.