Are Chelsea better without Romelu Lukaku?

Two games into the Premier League season, it all seemed pretty simple. Once Thomas Tuchel took the Chelsea job in January, they became the best defensive team in the world. Over his 19 Premier League games in charge, they allowed 13 goals — a 43% decrease on the 23 goals allowed in the season’s first 19 games under Frank Lampard.

Under Tuchel, they conceded 7.9 shots per game — fewer than all but two teams in Europe’s Big Five leagues allowed last season. And the average expected-goal value of each shot they allowed was just 0.08 — meaning the average shot Chelsea conceded had an 8% chance of turning into a goal, compared to the Europe-wide average of about 12%. Only one team limited their opponents to worse attempts, and all in all it added up to 0.6 expected goals allowed per game, which, over the course of a full season, would’ve been the lowest rate in Europe.

That impenetrable defense propelled Chelsea back up the Premier League table into the Champions League places come season’s end. Oh, and they won the Champions League, too. In the final, they held Pep Guardiola’s Manchester City to just seven shots and a meager 0.45 xG — City’s third- and second-lowest totals in those respective metrics since the Spaniard arrived in 2016.

Teams weren’t supposed to win titles like this anymore — and they rarely did. The previous six Champions League winners averaged 103 goals scored in domestic play. The peaks were Barcelona in 14-15 and Real Madrid the following year, both with 110, while the trough was the 17-18 iteration of Madrid that finished with a pathetic 94.

Chelsea wrapped up last season with … 58 goals in domestic play, and their scoring rate actually dropped off once Tuchel took over, as they scored 25 goals in his 19 Premier League matches, which would add up to 50 goals for a full season. Ten teams in the Premier League scored more than 50 goals last year. With their sights now set on a Premier League title, that just wasn’t gonna be enough. An elite defense could carry you across the seven-game volatility of the Champions League knockout rounds, but you simply just couldn’t win enough matches over a 38-game season unless you started significantly outscoring the likes of Leeds United and Aston Villa.

One easy way to fix that? Sign the best striker on the market and plop him on top of that impossible-to-score-against defense. But has it really worked out?

In August, Chelsea paid €115 million to re-acquire reigning Serie A MVP Romelu Lukaku from Inter Milan. He didn’t make the bench for Chelsea’s first match of the season, a 3-0 win over Crystal Palace, but he was out there from the jump in game two against Arsenal.

The appeal was clear right away. Chelsea dominated possession and kept the Gunners away from their goal, much in the same way they would’ve last season: 65% of the ball and just six shots allowed. But the presence of Lukaku, with his savvy final-third movement and his presence in the box, made it easy for Chelsea to break down the same kind of settled defense they seemed to struggle against the season prior.

Lukaku scored once and outshot Arsenal (eight) all by himself. Chelsea won 2-0, and it easily could’ve been more. Tuchel & Co. had proof of concept after just 90 minutes of play.

Of course, it’s never that simple. Through 13 Premier League games, Chelsea are sitting in first place — one point ahead of Manchester City, two beyond Liverpool. They’ve conceded just five goals and scored 31 — six more in six fewer games than in Tuchel’s first half-season. Except, only three of those goals have come from Lukaku.

So far this season, Chelsea have been better without their club-record signing.

With or without you

In basketball, teams and analysts use something called “WOWY” stats. Funny name, simple concept: With Or Without You. How does the team perform when this player is on the court, how does the team perform when this player is off the court, and how do those two things compare to each other? It’s a tidy concept that addresses a bunch of the shortcomings of individual player stats: they don’t always connect to winning, and they miss all kinds of contributions a player can make when he’s not touching the ball.

If you boil down the individual nebulous whole of player contributions down to a single concept, it would be: how this person increases a team’s ability to outscore its opponent. With its constant substitutions, a basketball game is something like an ongoing experiment, where different players get to play against different opponents with different teammates, lots of points are scored and the results are easily recordable. While it’s not perfect, since no two players play with the exact same teammates or the exact same opponents, over the course of a season you can look at a team’s scoring margin when a player is on the court and when he’s off of it to get a general sense of how he’s contributing.

Perhaps the most famous example of a player who was raised up by his WOWY stats was Shane Battier, the subject of Michael Lewis’s 2009 New York Times Magazine piece “The No Stats All-Star.” “We have been a championship team with him and a bubble playoff team without him”, said then-Houston Rockets general manager Daryl Morey.

Soccer is almost precisely designed to evade this kind of analysis. There are only three subs per game (in the Premier League), so the same players tend to play with each other, which makes it really hard to tease out the magnitude of individual contributions. Liverpool have a plus-29 goal differential this season, with Virgil Van Dijk and Mohamed Salah both playing every single minute; how much of that plus-29 comes down to van Dijk, and how much do you give to Mo? There’s an answer out there somewhere, but there’s still no satisfying way to find it.

On top of that, there just aren’t many goals in soccer. Liverpool have scored 39 goals so far this season; the Golden State Warriors score 114 points per game. The massively larger sample makes the on-off numbers for the NBA much more likely to be meaningful.

To work toward fixing that, you’d need a way to determine team performance that contains more information than just the handful of times the ball goes into the net. And you’d also need a coach who constantly messed around with his lineups. Well, thank you, Mr. Tuchel.

So far this season, no Chelsea player has played every minute in the Premier League, Edouard Mendy and Antonio Rudiger are the only ones to play at least 90% of the available minutes, while Thiago Silva and Jorginho are the only others to break 70%. Due to an ankle injury he picked up against Malmo in late-October, Lukaku has appeared in just slightly more than half (53%) of Chelsea’s Premier League minutes.

As for the lack of information inherent in goals, a better way to judge team performance would be to look at expected goals (xG), which accounts for all of the shots conceded and attempted by a team. For Chelsea, that would increase the number of data points from 36 (31 goals for, five against) to 325 (217 shots for, 118 against). All of the shots are then weighted based on their historical likelihood of being converted into a goal.

Taken together, what does it all say?

According to the site FBref, Chelsea’s xG differential per 90 minutes is 1.22 goals worse when Lukaku has been on the field, compared to when he’s been off it. (The numbers are similar in Europe, too.) When he’s played in the Premier League, Chelsea’s xG differential per 90 minutes comes out to plus-0.35. (Among Chelsea players with at least 600 minutes, only Marcos Alonso and Cesar Azpilicueta are attached to worse numbers.) That would be the fourth-best mark in the league for a team, but it’d be nowhere near Liverpool’s plus-1.60 or Manchester City’s plus-1.55, and it’s way below Chelsea’s performance under Tuchel last season of plus-1.10.

On the other end of the spectrum, Chelsea have been on or right below Liverpool and City’s level (by xG) when Hakim Ziyech, Ben Chilwell and Reece James are on the field.

We’ll know soon enough just how well Lukaku fits

Now, there are some massive caveats that go along with this. It’s a really small sample of matches, and Lukaku didn’t play against the same opponents as Ziyech or Chilwell or most of his teammates. Lukaku’s spate of matches include the games against City and Liverpool, and it didn’t include the ones against Norwich and Newcastle. At the same time, we saw Chelsea struggle against a struggling Juventus with Lukaku in the lineup, and then we saw Chelsea absolutely obliterate Juventus without Lukaku in the lineup. And while the xG numbers from the Liverpool game were skewed by a first-half red card, Chelsea did get annihilated — at home — against City with Lukaku in the lineup.

Some added concern comes from the fact that at Inter last year, the team didn’t get any worse when he didn’t play in Serie A, while their xG differential was 0.25 worse in his first season with the club, and Manchester United experienced a similar uptick (plus-0.27) when Big Rom wasn’t out there. Again, however, these are small and skewed samples. And United were better with Lukaku on the field when they finished second in 2017-18.

I outlined the broad story of Chelsea’s season last week: they played like a good-not-great team for the first couple of months, but churned out great results thanks to some unsustainable finishing and some great goalkeeping from Edouard Mendy. Then, Lukaku got injured and Chelsea pretty much immediately started looking like a great team again.

Before Lukaku’s injury, Chelsea averaged just 53% possession in the Premier League, and that number has leapt up to 67% since. A rotating cast of players have filled in, but they’re all either much more secure in possession — attacking midfielder types like Kai Havertz or Christian Pulisic — or not getting on the ball as much (Timo Werner), so the insecurity on the ball doesn’t affect Chelsea’s ability to keep it as much as Lukaku’s does. If Chelsea really are going to be better without Lukaku, it’s because a more fluid group of attackers allows Chelsea to keep more of the ball, which in turn allows them to both attempt more shots and prevent their opponents from doing the same.

Is this just a statistical quirk? Or is it a real issue?

Well, after Sunday’s 1-1 draw against Manchester United, Tuchel said, “If Romelu’s fully fit, he’s a starter for us, but he’s not in the moment.” So, soon enough, whenever Lukaku is back to 100 percent, we might actually get an answer.