Pep Guardiola’s bid for an undefeated Premier League campaign ended Sunday at Anfield, and the football world sounded off with opinions on what worked, who the key individuals were, and why Jürgen Klopp has succeeded against Manchester City in ways others haven’t. STATS Playing Styles and Tier 6+ data take it beyond unquantified analysis to substantiate and challenge those claims while giving credit to Liverpool players less obvious than Mohamed Salah, Roberto Firmino and Sadio Mane. The main takeaway? Press on.
It’s something of a cycle. When Liverpool win, Jurgen Klopp’s system is lauded. There’s of course been plenty of that lately. When the Reds relinquish a lead or implode in their own half, the same system tends to fall under defensive scrutiny. The commentary is often reactionary and oversimplified, and it’s occasionally unsubstantiated. For better or worse, the term gegenpressing follows the Liverpool boss like a role follows a typecast actor even if he’s capable of much more.
Whichever side you’re on from week to week, you have to accept one simple fact: After Sunday’s 4-3 victory, in four Premier League matches, Klopp has seven points against Guardiola and Guardiola has four against Klopp. Something is working. Rather, somethings are working. At the centre of those somethings lies efficacy in individual player possessions and how that fluctuates based on pitch position. No football eye, as keen as it may be, can track that on its own. With STATS Playing Styles, we’re going to show what actually went down at Anfield as Liverpool stretched their Premier League unbeaten streak to 14 while ending City’s at 30 dating to last season.
Let’s first address the match’s stylistic deviation for each club in comparison to their season form, and we’ll work into more complex and individualised touch data as we go. The Playing Styles webs below give us some surface-level insight for this specific match and how each team compared to their 2017-18 season selves:
What this tells us isn’t in itself groundbreaking: City were disrupted in possession-based attacking styles such as build up, sustained threat and fast tempo. These are styles they display more frequently than any club in the Premier League. Liverpool, meanwhile, employed more counter attack, high press and direct play than they typically do. This also isn’t a radical consideration given the expectations football minds have for this specific fixture.
As much as we’d like to compare Liverpool’s win to City’s 5-0 victory at the Etihad back in September, it’s illogical to do so because Sadio Mane was sent off in the 37th minute, so Liverpool played down a man for the majority of the match. Instead, we’ll start by comparing the teams to their season norms in a few key categories.
Liverpool had 10 possessions on which their high-press membership accounted for 50 percent or more of the possession’s value, which ranks third among their 23 matches behind a 4-0 win at Bournemouth on Dec. 17 (15) and that frenetic 3-3 draw with Arsenal on Dec. 22 (13).
What we should notice first about City is an incredible dip in build-up play, which is defined as periods of play in which a team is looking for opportunities to attack between midfield and the edge of the 18. Manchester City operate at 136 percent above Premier League average in build up. Last weekend at Anfield, they were at +6 percent. And because build-up play can feed sustained threat and fast tempo, those percentages fell off as well. City’s season average for sustained threat is +71 percent. It fell to -14 at the weekend. City’s average for fast tempo is +192 percent of the Premier League average. It fell to -70 against Liverpool.
These are all league-leading season marks, as is their +52 percent maintenance, which rose to +106 against Liverpool. City had the ball plenty, but it didn’t nearly as often progress beyond maintenance, which captures possessions in which a team looks to maintain and secure possession within their defensive area. At Anfield, they had 55 maintenance possessions for their highest total of the season. This begins to hint at the position of City’s possessions or possibly an inability to progress the ball into more dangerous attacking areas.
Let’s now look into where they lost the ball and how frequently. Given Liverpool’s success Sunday, it might follow that we should expect Klopp’s side to have dispossessed City more frequently higher up the pitch than other sides have.
City, one way or another, were dispossessed 36 times in positions spanning from their own goal line to five metres beyond midfield – the zone STATS Playing Styles defines for an opposition’s high press opportunity. But that’s actually rather average for City. They lost the ball in this zone on average 37.8 times per game in their other 22 matches, so what made the widely lauded Liverpool press effective?
Here’s where those analysing the match might not be finishing the job. It wasn’t necessarily that City were being dispossessed at some incredible rate by Liverpool. It’s that they were being displaced. The Reds’ success almost certainly had something to do with pushing City deeper than they typically play. With Tier 6+ event data, we can average the XY positions of each player on his touches. The initial insight here lies in the average position of each City player up the pitch. City’s possession leaders were defenders, not midfielders, but that’s not an oddity in itself, particularly in a match such as this when so much of City’s possession was in maintenance.
Nicolas Otamendi led City with 111 possessions, and his average touch was 16.8 metres behind midfield. His season average? 10.0 metres into his own half. Fellow centre back John Stones only had a variation of -2.4 metres, but that dropped him back to an average touch point of 18.6 metres behind the centre line for the deepest position of any outfield player in the match other than Dejan Lovren (-21.1). To the right, Kyle Walker fell off from 3.2 metres in front of midfield to 3.7 behind.
Granted, these distances can be made up with the right passes, but there’s something to be said for the psychological impact of consistently possessing deeper in one’s end, and that’s compounded when you’ve got burners like Mane and Mohamed Salah running at you. Possibly the most striking deviation for City was that of Danilo, who came on at 31 minutes for Fabian Delph. Danilo’s season average touch happens 3.5 metres into the attacking half. Against Liverpool, it occurred 11.7 metres behind half, and few will be reluctant to give Salah some credit for that.
It clearly happened in the attack, too. Raheem Sterling’s typical touch occurs a remarkably advanced 21.7 metres beyond midfield. It fell back to 13.7 against Liverpool.
This may sound interesting enough on its own, but none of it means much if we can’t assign efficacy to what occurs in a given position. Let’s go to the middle of the pitch, where value has traditionally been difficult to measure for players who don’t score much but also aren’t the last line of defence.
Fernandinho is considered a more-than-capable holding midfielder, and for good reason. There’s just a limit to how deep that comfort – and value – goes. He’s not used to consistently possessing closer to his own penalty area, and it showed against Liverpool. His average touch occurred 5.9 metres behind half. His season average is 3.9 metres advanced from that. Fernandinho had 93 possessions against Liverpool, so that amounts to considerable ground for one player to compensate for in one match.
Here’s where STATS Ball Movement Points comes in. We’ve used BMP quite a bit in past posts, but here’s the rundown: BMP considers ball movement made by an individual player from a start zone to an end zone and assigns value based on past results from massive amounts of league data. These scores accumulate during a match or across a season to indicate the value of a player’s ball distribution. BMP considers every involvement a player has to credit or discredit decisions with the ball and reward creativity. It’s what football minds could always see but never calculate. It goes beyond expected assists by looking at the full chain of passes, weighing the probability of that pass leading to a shot later in the play. Passing points generate expected shot points, so if a player generates one BMP, he has generated passes to lead to or defend one shot. It expresses the level of threat or wastefulness that can be attributed to a player. It’s broken down into categories of offensive and defensive as well as positive and negative (oBMP+, oBMP-, dBMP+, dBMP-) with net values telling the more conclusive story.
For the season, Fernandinho’s 3.93 net oBMP ranks seventh with a truly elite group of Premier League midfielders, and that’s particularly impressive because it’s among players who have far more offensive opportunity and responsibility than he. But his oBMP for the Liverpool loss was 0.08. You know what’s coming: Among the 21 matches in which he possessed the ball at least 60 times, it was his lowest mark. His dBMP- – measuring a player’s liability in possession for giving the ball away in dangerous areas – of minus-0.11 is his third-worst mark of the season behind City’s Nov. 5 match with Arsenal and Dec. 10 match with Manchester United.
It wasn’t just Pep Guardiola’s go-to holding man falling short. It was also his potential player of the year. You can see above that Kevin De Bruyne wasn’t nearly as withdrawn as some of his teammates, but his 0.06 oBMP was his second-worst mark of the season, ranking narrowly ahead of an outlier against Swansea last month. City had a 3-0 lead on 52 minutes, so there wasn’t exactly a need to get ambitious for much of the game. And he made up for any creative deficiency in that match by finishing a chance himself.
That list goes on for City’s players. It’s time to assign direct credit to the Reds. We’ll start on the team level and work down to individuals.
We already discussed the Reds’ high press. Liverpool also achieved their second-most possessions with a counter-attacking membership of at least 50 percent. Those 12 trail only the 13 they had in that December match with Arsenal. When their counter and high press overlapped, good things happened, particularly in the 62nd minute. Salah’s press of Otamendi and regain turned into transition and a beautiful finish from Mane that gave the Reds a 3-1 lead. Without it, we’d potentially be talking about another Liverpool defensive collapse rather than City’s first loss.
While City players fall in across the board against Liverpool, the Reds’ blazers played more of their game against City. Salah’s average position per touch for the season is 20.4 metres beyond half. Against City, it was 17.7. Mane: 15.4 for the season and 12.6 against City. Roberto Firmino fell in from 16.4 to 14.0, but it’s far from the deviation for City’s front.
So what exactly pushed City into those deeper positions, and who else accounted for the regains? We don’t need high-level event data to tell us it wasn’t Philippe Coutinho. We do need STATS Playing Styles and Tier 6+ to quantify the value of a Liverpool midfield that’s often overlooked in favour of the flair provided by Salah and Mane.
Yet if we look at Liverpool’s team-wide deviation against City, it’s considerable. In two matches, Liverpool fell in 3.7 metres behind half against City, and interestingly enough, it was more drastic with 11 men (-4.5) than at the beginning of the season with 10 (-2.9) for the majority of the match. For the season against all opponents, their average touch position is 1.3 metres advanced of half.
So if Mane, Firmino and Salah aren’t accounting for much of that dip, it follows the farther we go back in the 4-3-3 formation, the more ground Liverpool cede against City. Emre Can, one of the most centralised players in the Premier League (X: 0.0, Y: -0.7) is getting plenty of love for his central presence in the match, but is anyone providing empirical evidence as to why? Where Fernandinho might have faltered, we can objectively state Can flourished. His average touch’s position went from -0.2 last time out against City to -10.0 on Sunday, yet his oBMP actually increased. That’s difficult to do, particularly so against a Manchester City side that holds the ball as much as they do. And Can’s dBMP- was his fourth-best single-match total of the season, despite playing in easily his most withdrawn position of the season. If Juventus are as interested in Can as certain reports suggest, Liverpool would be wise to show tape of his efforts against City – and STATS Playing Styles can objectively support it.
Onto Can’s midfield teammates. Arsenal and Newcastle fans, it’s time to look away. Alex Oxlade-Chamberlain and Georginio Wijnaldum played substantial roles in putting City in those uncomfortable scenarios, and it goes well beyond Oxlade-Chamberlain’s ninth-minute goal or 59th-minute assist to Firmino.
The £40 million man contributed four counter attack regains, a high-press regain and a truly impressive 192.7 metres of counter-attack distance covered (126.2 passed and 66.5 carried) for the seventh-highest single-match total in the Premier League this season.
Oxlade-Chamberlain’s counter distance accounts for 25.1 percent of his season total (766.8 metres), so this might have been an outlier of a performance. Or, as some in the media have suggested, it could be an indication of an emerging role he’ll take on with Coutinho gone. But that’s an odd suggestion even if you went no further into it than watching the match. And it’s certainly not supported by his average position on the pitch against City, particularly the east-west axis. Oxlade-Chamberlain’s season average touch occurs 13.1 metres beyond midfield and 2.7 right of centre. Against City, his touches occurred on average 9.3 metres beyond midfield and 6.1 right of centre. Coutinho’s average horizontal position with Liverpool was 13.0 metres forward and 7.9 left, so well across the pitch from where Oxlade-Chamberlain operated. It seems someone else was slotting in behind Mane on the left.
That someone, at least for one thrilling match, was Wijnaldum, who’s average position changed from his season marks of 3.6 metres beyond midfield and 2.9 left of centre to 0.3 forward and 10.0 left. The result was three counter regains and a substantial 120.1 metres of counter distance (37.2 carried and 82.8 passed).
In all, Liverpool’s counter attack covered an impressive 642.1 metres for their second-highest single-game mark this season behind only the Arsenal match. City, Arsenal and Watford have each topped that mark once this season while many Premier League clubs would need multiple matches to get up to that kind of total.
Of course, not all of this distance covered contributed to scoring or even to shots. But it changed the way City went about their match. It changed the positions of their players on the ball. It took them out of the comfort zone that, though 22 matches, it seemed they could patent.
Liverpool had their deepest average starting point of any match this season at 4.5 metres short of midfield, and their 527 possessions rank ahead of only that 10-man match at the Etihad, yet they found a way to get City flustered – all the way back to surefooted distributing goalkeeper Ederson for the fourth goal. That comfort zone typically means Guardiola’s side on average touches 3.3 metres into the attacking half. City’s average touch came 2.8 metres behind midfield against Liverpool, which for the season is ahead of only that first match with Klopp’s side (-2.9). Leicester City is the only other club to knock City’s average back into the defensive half, and they did it by a mere 0.2 metres.
Liverpool aren’t doing this to everyone, and it’s not the only way they succeed. Leicester back on Sept. 23 began their possessions 4.8 metres beyond midfield against Liverpool, which resulted in a 3-2 win for Klopp’s side. They are quite frequently a possession-based attacking side, so it’s probably oversimplifying to pin that gegenpressing term to him so dutifully.
But against City over two seasons, it’s resulted in two wins, a draw and a loss. The next meeting should be even more interesting. A phrase that of course follows Guardiola around is total football. That means he knows how to press back.