It was hardly surprising that at least three of the Premier League’s leading clubs were interested and, ultimately, that one of them ended up taking the plunge.
The striker had scored freely in his domestic league, was a good age and, on the face of it, had all the attributes — physical in particular but technical too — to succeed in England.
Advertisement
Perhaps that should say “all the attributes that can be easily measured”. In a meeting with the manager of one of the three Premier League clubs, the player dropped into the conversation about how hard it was to play for his current team because of the pressure he felt to score goals.
Prolific or not, alarm bells were ringing. If a centre-forward was going to sign for one of the biggest clubs in the world, and for a huge transfer fee, in a league where it is even harder to score, how on earth would he cope with the spotlight and the level of expectation?
The last thing that anyone wants before completing a transfer deal is doubt. “Are you sure?” is one Premier League owner’s favourite line before a signing gets the green light, and nobody inside the club — coach, sporting director, head of recruitment — wants to give a woolly response to that sort of question.
In the end, the manager withdrew his interest and, with the benefit of hindsight, it turned out to be a wise decision because of what has happened to the striker since. He has, in football parlance, flopped.
Any transfer deal carries an element of risk, but there are more layers of complexity and uncertainty when it comes to players swapping countries and moving to a different league. How they adapt to the style of football is one thing, how they adapt to living in a new country is quite another. Arsene Wenger felt that overseas signings needed six months of grace. Some Premier League managers double that time frame.
Either way, it is interesting to cast your eye down the table below and wonder what the numbers will look like for Manchester City’s Erland Haaland and Liverpool’s Darwin Nunez — the Premier League’s two biggest imports this summer — at the end of their first season in England. Nunez came off the bench to score in Saturday’s Community Shield victory whereas Haaland missed a sitter late on, but how will their contribution be viewed in May next year?
To be clear, the players listed above are high-profile case studies rather than a large sample to draw any reliable conclusions from — besides, a look at goals and assists alone is skewed by the outcome and not necessarily the underlying performance of the player in question.
Nevertheless, there is some fascinating data to explore that looks beyond the “noise” of goals, taking in everything from analytical models that compare the relative strength of each league through to the use of artificial intelligence, and even ways of forecasting the impact of a new signing on a team’s results.
Advertisement
Real Analytics, which works with five Premier League clubs, predicted last year that Manchester United would finish fourth in 2022 with Harry Kane up front, yet still had only a 3.5 per cent chance of winning the league if they signed the Tottenham Hotspur striker. They predicted, correctly, that United would finish sixth without him.
Some definitive trends emerged through our research at The Athletic, whether that be the fact that Ligue 1 is the go-to overseas market for English clubs, or that the Bundesliga averages the most goals per game across Europe’s top five leagues (should we revise our expectations of Haaland accordingly?).
Using data from SkillCorner, we can even see that the Premier League is comfortably out in front when it comes to high-speed running… with Ligue 1 lagging behind (see below).
The metrics, of course, can only tell part of the story. Away from the pitch, clubs have to consider the language barrier, the emotional attachment that a player feels to their home country and the ease with which their partner and children can settle overseas, as well as cultural and climate differences.
“It’s not necessarily a quality to go to a different country, because the target in life is to be happy, and some people are happier when they are in their own environment,” Wenger told the Arsenal website in 2015 when interviewed at length on this topic. “Some people have that desire to discover other cultures, other places and people, and they will be happier, so you have to analyse that.”
Slaven Bilic has always been one of those people. The Croatian has worked for clubs in Russia, Turkey, China, Saudi Arabia and England. He knows what it is like to sign overseas players and, crucially, what it feels like to be one — he played in the Bundesliga and the Premier League after leaving Croatia.
Advertisement
“There are many things with the scouting, the analysing, with tracking the players for so long that the clubs are doing now, not only about their performances but also about their private life and character,” the former West Ham United and West Bromwich Albion manager tells The Athletic. “But then there’s also the element that you can’t predict, especially when it comes down to the players who are leaving their home city, home club or home country for the first time.
“It’s different when it’s someone like Ivan Perisic now. Perisic left Croatia when he was 17. He went to France, then to Belgium, then to Borussia Dortmund, then to Wolfsburg, then to Inter Milan, then to Bayern Munich… for him, there’s no danger that he’s going to take time to adapt (at Spurs). It’s his sixth or seventh time that he’s changing the city, the dressing room, the league. It’s not a problem.”
But what about those who are moving for the first time? How can a manager know if it will work out on the pitch, let alone off the field? What are 25 goals in Portugal’s Primeira Liga, or the Eredivisie in the Netherlands, worth in the Premier League? “That is an art, that is a feeling,” Bilic says. “And it’s the biggest question. How can you ignore the levels? But you also can’t be a slave to the levels.”
Bilic smiles. “With your due diligence, you are reducing the element that you can’t control. Unfortunately, you can’t nullify that. If you could, there wouldn’t be mistakes. And there’s a big difference if you are managing top clubs, or managing medium clubs — I’m talking in the Premier League.
“Every manager would like to get a centre-forward who is a good age, who has got a private situation that is OK, that doesn’t drink, that is very stable, that is under 24, that will score 20-plus goals, and also will keep and hold the ball to link up play. But unfortunately, there are only a few of them. That’s why Guardiola can get Haaland. As a mid-table manager, not to mention lower clubs, you have to take risks. That’s the market you are in. Then the question mark is bigger, and that is a problem.”
So, how can we work out if one league is stronger than another?
Information from sources such as UEFA coefficient rankings or ClubElo can be used to add more or less weight to a player’s output depending on the strength of the league that they play in. But arguably, a stronger way of looking at league strength is FiveThirtyEight’s Soccer Power Index (SPI) ratings, which estimate a team’s overall strength between zero and 100, using difficulty-adjusted match results and underlying performance metrics to model a team’s — and subsequently a league’s — offensive and defensive strength.
Advertisement
This method is more reliable than the alternatives, as it looks beyond simply the result to inform the ratings. For example, if we take the recent Champions League final between Liverpool and Real Madrid, Jurgen Klopp’s team would show as stronger in their underlying performance despite losing 1-0.
Ranking the leagues in order using FiveThirtyEight’s method, the top 20 looks like this…
“The transfer market gives us a huge amount of information about the relative quality of different leagues,” Dan Altman, creator of analytics website smarterscout, tells The Athletic.
“When a player switches from League A to League B, we measure how their performance metrics change — that gives us one set of data points. We also look at players who make the reverse switch, from League B to League A — that’s another set of data. Then there are players who go from League A to League C, and others who go from League C to League B. We’re seeing the same journey in two steps, and the reverse journey is just as useful. Now you can see that there’s actually quite a lot of data to calibrate these league differences and we can compute them for each performance metric individually.
“In general, league adjustments have been very reliable, but there are always challenges. Some players are developing, others are declining. Players may go to a club that’s a poor fit, or they may need time to adapt to a new league. For example, Haaland needed half a season to adapt to the Bundesliga. After that, his attacking metrics were very similar to his metrics in Austria, adjusted to a Bundesliga standard. But with each new season, he was dribbling less and less, and his ball retention improved beyond where it had been in Austria.”
This objective measure of league strength could be used to calculate the percentage difference in quality between divisions and apply what you might call a blanket “tax” to players in their performance metrics. For example, if FiveThirtyEight rates the Premier League (SPI: 73) as 40 per cent more difficult than the Portuguese Primeira Liga (SPI: 52), then you need to update your expectation of a player’s output when they move leagues. For example, 10 goals for Nunez in Portugal might be worth closer to six in the Premier League.
So that solves all of that, then…? Well, not quite. Simply comparing League X with League Y ignores the context. What if a player moves to a more difficult league but is surrounded by better team-mates and his performance levels are elevated as a result?
That is exactly what happened to a high-profile overseas signing who had an instant impact in the Premier League. Although his physical metrics at his former club were impressive, there was a feeling that what he was doing with the ball wasn’t as effective as it should have been because of the standard of his team-mates. To put it another way, the theory was that he would fly higher in better company — and it proved to be spot on.
Advertisement
All of which brings to mind Bilic’s answer when he was asked whether Haaland is close to being a “sure thing” at Manchester City. In Bilic’s mind, it is easier to score goals in the Bundesliga than, say, Serie A. “I wouldn’t value 20 goals in Germany more than 12 or 14 in Italy. It’s more difficult to create chances in Serie A than in the Bundesliga,” he says.
Looking at the chart below, which shows the average goals per game in the top five leagues in Europe since the 2017-18 season, Bilic appears to have a valid point.
The flip side to all of this is that Haaland has moved to a harder league but also a better team. With all due respect to Dortmund, Haaland’s supply line has been significantly upgraded.
“Is the Premier League stronger in every aspect than the Bundesliga? I think it is,” Bilic adds. “But then Haaland is also surrounded by better players at Manchester City: Kevin De Bruyne, Bernardo Silva, Riyad Mahrez — oh my God.
“And I think Haaland is everything. He can use space when there is space. Some people will say — which I agree with — that there’s no space with the way that Man City play, but they’re still capable of creating that three or four yards behind opponents. And Haaland’s also a poacher. So that’s why he’s almost a guarantee.”
Timo Werner and Romelu Lukaku. “They’re two interesting cases,” says Ian McHale, professor of sports analytics at Liverpool University and the owner of Real Analytics. “Before Werner moved to England, we thought he would do quite well.
“We’re not generally looking at scoring goals, because there is so much noise in that. But that’s how some clubs are still going about judging players. I want to strip the signal out from the noise to find out how good the player actually is. Whereas goals depend on who he’s playing with, where his chances are coming from, and the strength of the opposition. So I’m trying to strip all of that out and say, ‘What is his contribution to the team?’.”
McHale types Werner’s name into the Real Analytics platform and shares his screen (as shown below). The graph at the bottom is coded with coloured dots according to whether Werner’s performances for Chelsea have been good or bad.
“Green dots are when he’s been good, red dots are when he’s been bad,” McHale says. “There are still loads of green dots. What’s interesting is the perception that he’s been bad because he misses goalscoring chances. In his first season (when Werner registered six goals and eight assists in the Premier League), he wasn’t as bad as people think.”
The obvious question to ask is how these performances are measured. “We use something called deep reinforcement learning,” McHale says. “It’s an artificial intelligence (AI) method to learn the value added when a player does something on the ball. So it looks forwards and backwards during the possession and tries to pick up patterns that coincide with increasing your team’s chances of scoring and decreasing the opposition’s chances of scoring.
Advertisement
“So if I intercept the ball, that’s just decreased the opposition’s chances of scoring, and it’s increased my team’s chances of scoring. So you’re measuring these changes in what is called an expected value of possession, for before and after each player does something.”
For the analytics folks among you, this acts in a similar way to expected threat (xT), as covered by The Athletic previously.
It is a fascinating concept that comes to life (see below) when McHale shows a video of Everton scoring from a counter-attack against Manchester United last season. “This graph along the bottom, this is what the AI determines the value of the possession to be in terms of Everton’s scoring chance — it’s going up because they hit Man United on the break — and this shows you how the green and red dots work.
“Allen has just cleared it for Everton, and then Fred gets his foot on the ball but he makes a mistake. For Everton, the AI likes that, but for Man United, that’s a negative. So Fred gets awarded minus — 0.274.”
McHale lets the footage run on to the point where Fred tries and fails to make a recovery tackle. “Then we’ve got Fred trying to make up for it here,” he adds. “He attempts a challenge and bumps into Demarai Gray. So he fails but the AI has learned that Everton’s chances of scoring — if you look at the graph at the bottom — goes down a little bit after that challenge because Fred has disrupted it a little.
“There’s no opinion in here, there’s no, ‘Ian says an attempted tackle is worth a little bit’, it’s just worked it out. Then Demarai Gray goes on a little run and it (the AI) likes it, the possession value is going up all the time now. What’s amazing about this is that Andros Townsend doesn’t just get ‘1’ because he scored a goal and everybody else gets ‘0’ for not doing anything. The contribution to Everton is spread out among the team, depending on what their contribution was (to the goal). So this is how you judge players’ contributions to teams properly and it strips out the noise.”
Real Analytics is often asked by clubs to measure the impact that a particular player would have on their team if they signed them. “Our background is actually in forecasting, so we were helping out betting syndicates before we were helping out clubs,” McHale adds. “We would gain an edge in the market that was based on the players on the pitch, not the team.
Advertisement
“The advantage of that approach is I can then pick up Nunez and put him in Liverpool’s team, or squad, and then simulate the results for the league a million times. And then you try to work out what, on average, happens with Liverpool’s old squad and Liverpool’s new squad, and the difference is the impact that you are expecting Nunez to have.”
McHale pulls up an example on screen. It was produced in April last year and although the transfer is domestic, the model and the principle stand. In this case, Real Analytics looked into the impact Kane would have on Manchester United in the 2021-22 season if they signed him.
“As of April 7, 2021, United were second in the league. But our predictions for the 2021-22 season were that their probability of winning the league was 0.5 per cent, the probability of top four was 26 per cent, and the average league position was 5.9, so sixth,” explains McHale.
“If they’d have bought Harry Kane, their probability of winning the league went up to 3.5 per cent, still really low because they’re miles behind Liverpool and Man City. Their probability of top four rockets, though — it goes up to 62 per cent. That’s because you get the double whammy of getting a really good player and massively weakening a rival (Spurs) for the top four. They gain six and a half points on average, and they gain nearly two league positions. That’s the kind of information that when you’re signing a cheque for £100million, the board and the chief executive need to know.
“So, as best you can, going all the way back to talking about how strong leagues are, you try to account for how strong the team the player is coming from is, and how strong the league they’re going into is. But it’s not a perfect science. Annoyingly, it’s not random who’s moved in the past — there is a bias because the Premier League clubs are choosing the best players they can — so you can’t fully measure it.”
Bilic is thinking back to his time in charge of West Ham and Dimitri Payet’s first season at the club, in 2015-16. “People say, ‘What have you done with Payet?’. What have I done with Payet? Nothing. Not nothing. But basically nothing spectacular or revolutionary. I put him in the team in a good position and he was feeling good. Sometimes it’s to do with the things that you don’t influence.
“Sometimes the player comes to you, the numbers are good, and he plays terribly. Maybe — because I found this in my career and from my colleagues — just a couple of months before or after he signed the contract, he found out that his wife was cheating on him, or that his mother got cancer, or that his parents are divorcing, and that ruined his concentration and ruined his season. And you don’t have a clue about what is going on because he keeps it private. He plays terribly and then everyone is saying, ‘Why did you sign him?’, or, ‘How can’t you get the best out of him?’.”
Advertisement
It is easy to imagine those sorts of questions being asked about Sebastien Haller, who joined West Ham in a club-record deal worth up to £45million ($55.2m) in 2019, a couple of years after Bilic had departed. Prolific in the Eredivisie and the Bundesliga, Haller scored only 10 goals in 48 appearances in the Premier League.
Reflecting on that experience earlier this year, when he was leading the goalscorer charts in the Champions League, Haller told the Guardian that as well as playing in a system that he didn’t enjoy at West Ham, he struggled with how he was feeling off the field. His wife was seven months pregnant with their second child when he joined West Ham and the baby was ill when it was born.
“He didn’t sleep at night, was crying all the time, in constant pain,” Haller said. “I was worrying about him, about my wife, who couldn’t sleep and was feeling like a zombie. Then you have to adapt to your new environment, a new competition, then there was COVID, pfffff… that was maybe a lot to deal with.”
Adapting to the football can be tough, especially when the physical demands are so much greater. Wenger tells a story about Robert Pires sitting next to him on the bench at Sunderland, shortly after signing for Arsenal, and asking him after half an hour, “Is it always like that?” Wenger replied, “Yes, and it can get worse than that!”Another player, who went on to have a fantastic career, told Wenger, “This game isn’t for me” and asked to “go back”.
That was then, and this is now. “The world of football has changed,” Bilic says. “Before we could talk about the English style, the German style, the Spanish style. Now the clubs like Manchester City, or even Brighton, are playing a more Spanish style than Atletico Madrid.”
Looking through the transfer data, Ligue 1 is the preferred market for Premier League clubs in recent years, which is reflective of a feeling that there is value for money in France as well as — and this is interesting given our earlier data on high-speed running — a sense that the physical transition will be easier.
Boubacar Kamara (Aston Villa), Nayef Aguerd (West Ham), Cheick Doucoure (Crystal Palace), Sven Botman (Newcastle), and Giulian Biancone (Nottingham Forest) have moved to the Premier League from Ligue 1 this summer.
Leicester believed that Boubakary Soumare, who joined them from Lille for £17million last summer, would be a perfect fit for the Premier League. That was the theory, but in reality, the midfielder has struggled. “He’s used to playing at a slower tempo,” Leicester manager Brendan Rodgers said in April. “He knows he needs to get up to speed.”
Advertisement
Soumare has the physical attributes (as Rodgers has acknowledged), which can be easily measured these days, whether that’s speed, distance covered or anything else. “But it still doesn’t really reveal how quick your decision-making has got to be and how quickly you are under pressure,” adds McHale, highlighting a point that Wenger observed in the past about how the intensity of the Premier League “forces you to make quicker decisions”.
Breaking down those overseas transfers by position reveals that Premier League clubs are more likely to sign defenders and attackers — and that the Bundesliga is the most popular market for the latter as a share of the total transfers — in keeping with the high-profile signings in recent years (Werner, Haaland, Havertz, Leon Bailey and Haller) and the league’s high goal average.
Some have been hits and some have been misses.
As Wenger said, “Not everybody is exportable” in the world of football.
(Photos: Getty Images/Design: Sam Richardson)