This is the third part of my “Fixing Chelsea” series. If you missed “Improving Chelsea – Part 1” or Part 2, I would recommend reading those first as they provide some context to this instalment, and what this series is trying to achieve more generally. Part 1 in particular sets out some assumptions about Chelsea’s current situation that I would recommend reading. I also post additional content that doesn’t make it into these posts on my Twitter.
In short, the solutions I suggest through this series are as much: parts of a realistic framework of actions that would improve Chelsea’s current first-team squad, as they are parts of a fun exercise answering an interesting question: If Sarri were to be successful at Chelsea, how could that be assisted through player-recruitment?
In our attempt to recreate the most effective and functional elements of Sarri’s Napoli side, we really couldn’t ask for a more similar pair of midfielders in style and quality to Allan and Jorginho, than Kanté and… Jorginho himself. It genuinely makes it a lot easier to isolate what the midfield is missing when you have the exact fit in one position, and about as close of a fit as you’ll find in the second.
What’s missing then?
In comparing the output of Marek Hamšík to that of most other midfielders in Europe, let alone the sterile Mateo Kovačić and Ross Barkley, it becomes very clear what Hamšík brought to Napoli’s midfield and, by extension, what Chelsea is missing in theirs.
As you can see, the excellent Bernardo Silva, pre-eminent David Silva, and “€150M” (Mateja-Kežman-represented) Sergej Milinković-Savić are the only central/defensive midfielders to have managed over 0.8 non-penalty box-shots90 and 1.8 open-play shot-assists90 this season. Marek Hamšík hit those marks in each of his three seasons under Maurizio Sarri.
I keep my data for wingers/“10s” separate from my data for “8s”/“6s”, which is why there aren’t more names hitting Hamšík’s benchmarks.
How have Kovačić and Barkley fared this season?
Now, finding someone who will have the same creative and offensive output as Hamšík did for Napoli will clearly be somewhere in the range of very difficult to impossible, but improving upon the performances of Mateo Kovačić and Ross Barkley is much more manageable. This season Barkley and Kovačić, like most midfielders, haven’t gotten near Hamšík’s ~3.5 box-shots + shot-assists per 90 minutes, but have still managed a respectable 2.3 and 2.1 box-shots + shot-assists respectively. That doesn’t tell the entire story though.
Kovačić’s shot-assists have largely propped up his figures, and Chelsea’s exceedingly average shot-quality means that a shot-assist for a Chelsea player generally isn’t particularly valuable. The Croatian’s 0.25 NPxG+xA90 is some distance off the 0.4 that Hamšík managed across those three seasons under Sarri’s tutelage. I have also made my thoughts on Kovačić’s general play pretty clear on my twitter. I’m less than impressed, to say the least.
Ross Barkley has managed an impressive 0.25 xG90 and 0.24 xA90 this season, but I would be amazed if he sustained that over a larger sample size than ~600 minutes as a starter, and against better opponents. Since Kovačić’s first start for Chelsea, the opponents Barkley has started against this season currently average just 22 points (level with Cardiff). Aside from that, Barkley offers even less ball progression than Kovačić, if that is possible, and loses the ball 4.9 times per 90 minutes. He just hasn’t been able to thrive in this possession-heavy style of football, compared to his time at Everton where he spearheaded their transitions with his ability to carry the ball.
Kovačić and Barkley aren’t bad players, but this system definitely doesn’t get the best out of either of them, and I am very confident that they could be improved upon.
What will be our Key Performance Indicators? (What else did Hamšík do for Napoli?)
(Did rather than does because he has a completely different role under Ancelotti, so it really isn’t worth analysing what he’s been up to this season.)
As we’ve already covered, two of Hamšík’s most valuable skills were his chance creation and shot volume/quality, relative to the position he played. We’ll use open-play shot-assists90 and non-penalty box-shots90 to evaluate this.
For anyone new to football stats: we remove set-pieces from shot-assists because set-piece taking is a different skill from creating chances in open play. We only look at shots from inside the box as they are much more valuable (more likely to result in goals) and as a result, we minimize some of the “noise” in generic shot-volume. We exclude penalties from shots as being the designated penalty-taker is a different skill to getting on the end of chances in the box.
Why not use xG/xA? I’m sure some people will wince at the sight of box-shots and shot-assists, rather than the xG and xA we’ve all become so accustomed to, but they do provide a useful place to start. Basically, of my totally unsustainable and utterly convoluted data-collection practices, manually compiling xG and xA from Understat for 339 players would probably take the cake (and my sanity), hence my willingness to exclude it here.
The box-shots will give us a pretty good approximation of xG, but shot-assists are much more variable. My compromise is: we will use the two counting stats just to put an initial list of players together before looking at their xG and xA as part of our final judgements.
I think this is also a good place to mention that we need to have a conversation about shot-assists and expected-assists. Being the player who’s making that final pass is great. Our ability to distinguish a perfect cutback for an open goal, from a simple two-yard pass to Suso/Townsend/Ziyech for them to shoot from 30-yards using xA is also great. Buuuut, Player A can be a better creative passer than player B, without necessarily setting up as many (quality) shots as Player B. This is important to keep in mind generally and is relevant when looking at some of the players who make our final list.
Alongside creating chances and shooting, one of Hamšík’s most important attributes was his ability to progress the ball. This is a different skill to creating chances, and we have to view it as such when compiling our list of midfielders. Metrics such as passes-above-expectation, SaturdayOnCouch’s yardage-gained, and deep-progressions (I’m not a massive fan of these but we’ll include them in the list) are great at giving you a better idea of how effective a player (or team) is at progressing the ball.
Seeing as I have neither the data, nor means to manipulate that data (…yet. I’m trying to learn), to create something to evaluate how progressive players are, we will have to rely more heavily on the eye-test initially than I would like. I also used some lovely pass maps from the Stats Zone app, to try to evaluate each player’s ball progression abilities. You can download the Stats Zone app here for IOS: http://t.co/ho58H8fogt. If that link doesn’t work the download link is also in their Twitter bio.
We don’t have to be too concerned about the defensive abilities of the players we look at, as Hamšík would only contribute ~1.5-2 tackles + interceptions90 for Sarri. This actually means that we can also look at players who play as “10s” but could be moved back to an “8” position.
Much like the current Chelsea team, Napoli racked up the passes under Sarri, with Hamšík contributing 75-80 passes per90. We want to find someone who is quite involved in their team’s possessions, particularly in advancing the ball for their team, rather than a second-striker type who is just lurking around the box and laying off passes to onrushing teammates.
Some more niche points regarding players identification:
Marcos Alonso, Emerson Palmieri, and Olivier Giroud. Those are the only left-footed players in Chelsea’s current first-team squad. Whilst I doubt there is much empirical evidence supporting this, it seems fairly obvious that having a balance of left and right-footed players throughout your team is at least somewhat beneficial. Manchester City with all of their left-footed attackers, and Liverpool with Salah and Shaqiri, always seem to be able to threaten teams sitting in low-blocks more than we do, thanks to their different range of passing/shooting/dribbling tendencies.
In contrast to Part 2, I will set a more lenient age-limit, because it would be somewhat remiss of us to block the pathway of the players in Chelsea’s pipeline who could potentially fit into that attacking “8” position, by signing a player of a similar age. I also feel that it will be easier to find an undervalued midfielder in their peak-years than it is finding an undervalued striker. That said, if we were to find a young player who looks like they will be cheaper, and eventually better than, the best peak-age player currently, that is probably a smarter business decision. Though there is always the risk of losing one of those players currently in the pipeline working that way.
Before we get into targets, let’s look at the players already under contract at Chelsea:
The Lewisham-born midfielder has been at the club longer than any other player in the first-team squad, having joined in 2003 as an under-8. He hasn’t really been given a fair shot at that third midfield position this season, having only played ~300 minutes in the Premier League (just one starting appearance) so we’ll have to judge him on his time at Crystal Palace last season.
I know he mostly played on the left of a midfield four last season, but given his role in the system, and the positions he would take up throughout a match, I thought it made more sense to show him on the CM/DM radar rather than the “10”/Winger radar.
As you can see, he excelled in the metrics that are the biggest priorities for the role we are trying to fill. Loftus-Cheek’s 2.4 shot-assists+box-shots90 is impressive, especially considering his expected-stats were right in line, with his 0.17 NPxG90 and 0.16 xA90 already exceeding the output we have seen from Kovačić in those areas so far this season. He also progressed the ball effectively, largely through his exceptional dribbling ability, an area in which he is an outlier amongst central-midfielders.
Given the quality of the players around him, Crystal Palace’s style of play, and his few performances for Chelsea this season primarily in cup competitions, I have few reasons to doubt that Ruben could come in for the first-choice “8” without there being a significant drop in performance level.
Chelsea->Vitesse->*insert Championship club here* is a well-trodden and fairly ordinary path for many Chelsea academy graduates. Mason Mount’s last 18 months have been everything but that. Mason took the leap out of youth football at the earliest opportunity possible, and hasn’t looked back since, winning the U19 European Championships as the Player of the Tournament, claiming the Vitesse Player of the Year award, spearheading Derby to a famous FA Cup victory at Old Trafford, and being called up to the England Senior team despite playing in the Championship.
Much like Loftus-Cheek, Mount clearly excels in two of the areas we are paying particular attention to, chance-creation and shot-volume. Much like Ruben at Palace, Mount’s low defensive and passing numbers are more a consequence of the style of football played by Vitesse and Derby than his own ability. Mount is actually a better defender than his numbers would suggest as well, with his pressing being a standout feature of his game and one that suits the way Chelsea defend under Sarri, as James Socik detailed in his StatsBomb article earlier this season. Anyone who watched him man-mark Cesc Fabregas out of the game at Stamford Bridge (before it was cool to man-mark Chelsea’s deepest midfielder) would have picked up on this as well.
It is really impressive that Mount has been able to perform so well at such a young age, without being particularly physically imposing for his age, especially in two leagues/teams with such contrasting styles of play.
This comparison between Mount’s time in the Eredivisie and Championship also shows just how strange the Eredivisie can be in allowing midfielders and defenders to rack up the dribbles in a way that might not translate to more physical or defensively proactive/aggressive leagues. Essentially, I’m looking forward to all of the “What Happened to Frenkie de Jong?” pieces that will be written in November next season, after he tries to take players on in La Liga or the Champions League in the same way he does right now, only to get shouldered off the ball without a foul being given… 😉
The other young’uns
Tino Anjorin is a first-year scholar (16-17 y/o season) who was described by Chelsea U18-manager at the time, Jody Morris, as the best finisher in the group. He was still just a schoolboy (15-16 y/o season in this example) at the time, and this was a group that included Daishawn Redan and Charlie Brown who managed a combined 42 goals last season, and a certain Callum Hudson-Odoi. Pretty high praise.
Anjorin glides across the pitch with ease, has an exceptional weight of pass, and an uncanny habit of being in the right place at the right time with decisive actions. Physically, he is probably already prepared for Premier League football, with a similar build to Ruben Loftus-Cheek standing at 1.86 meters according to Transfermarkt, but he looks even taller than that when I’ve seen him in person. Don’t be surprised if you see him get opportunities sooner rather than later.
On the opposite end of the physical spectrum is Billy Gilmour, who is more of a Cesc carbon-copy than a Loftus-Cheek look-a-like. Billy does apparently model his game on Fabregas’ as well, and it seems to be working. He dictates the tempo of each and every match he plays with his metronomic passing, is impossible to dispossess, using Cruyff turns and stepovers to resist pressure, and has an eye for the home-run pass to send a teammate through on goal. I’m not aware if this is the case for Tino as well, but Billy has definitely been in first-team training so there is always the chance that he catches the manager’s eye. Billy even stood out for Scotland’s U21s when he was just a first-year scholar, competing against full-grown men without much of an issue physically.
I totally understand the opinion of those who would recommend against signing a player for that third midfield position, and would prefer to let Mount and Loftus-Cheek share the role and prioritize signings in other positions. However, my first instinct would be to bring in a world-class “8” to help make Chelsea a Champions League regular and Premier League title challenger, once again. I also think there would still be room for Mount and Ruben, as either could capably play as understudy to N’Golo Kanté in less-crucial matches. What is absolutely certain, is that there is no need for Chelsea to spend €50M on a Barella-type midfielder, or make the Kovačić deal permanent, when they have so much talent already under contract at the club.
The potential targets!
I will be using a similar process to find these players as I did for the strikers in “Part 2” but I won’t be able to rely on the stats quite as heavily, as I explained when we looked at what Hamšík did for Napoli.
Guys who have been good this season, but don’t quite make it to the final list
Florian Neuhaus – 21 – Borussia Mönchengladback – This has been his first season in a top-flight league and he has proven to be one of the better players in the Bundesliga at bringing the ball into dangerous areas with his passing. He just needs to contribute to more shots than his current 1.3 open-play shot-assists and 0.7 non-penalty box-shots per 90 minutes. He does some neat work around the box and is definitely one to keep an eye on.
Stefano Sensi – 23 – Sassuolo – Has only started putting impressive performances and a solid number of minutes together this season. The Italian does give you the benefits of being more of a two-way midfielder though, contributing 3.6 interceptions + tackles per 90. He’s a good player, but just doesn’t quite suit our needs.
Jonas Hofmann – 26 – Borussia Mönchengladback
Leonardo Bittencourt – 25 – Hoffenheim
Marcel Sabitzer – 24 – RB Leipzig
Hofmann, Bittencourt, and Sabitzer all have similar issues, in that they contribute plenty of end-product (between ~0.4 and ~0.6 NPxG+xA90), but are either involved later in possessions too often or not progressing the ball enough for our requirements. This is Sabitzer’s first season playing any minutes as a central midfielder, having spent the majority of the previous two seasons as a marauding winger/“10”. Both Hofmann and Bittencourt have only managed ~3000 starting minutes in the past 2.5 seasons.
Kevin Kampl – 28 – RasenBallsport Leipzig – Kampl has been one of the best midfielders in the Bundesliga this season. He actually does play in the right area of the pitch for what we’re looking for and is an excellent passer and his ability to progress the ball rivals any midfielder in the Bundesliga. His exclusion from the final list is mostly down to age. 28 is just too old, especially considering this is really his first standout season offensively.
Rade Krunić – 25 – Empoli – He’s been great this season at creating chances for teammates and dragging Empoli up the pitch with his excellent passing, but this is his first season showing that sort of output and he’s 25. I would be pretty sceptical as to whether you would end up with the 0.29 xA90 Krunić of this season, or the 0.07 xA90 of Krunić 16/17.
Pablo Sarabia – 26 – Sevilla – Sevilla have had the second-best attack in La Liga this season, and Sarabia has been the second-biggest contributor to that with his 0.7 NPxG + xA, only behind Wissam Ben Yedder. Unfortunately, he is definitely more of a winger/“10” type. You could make the argument that he could be transformed into an attacking “8”, but I would be hesitant about how much of his output would be retained.
Sergej Milinković-Savić – 23 – Lazio – The Spanish-born Serbian midfielder has the highest shot-contribution volume of any player on this list, at 1.6 NP box-shots and 2.0 OP shot-assists per 90 minutes. His 0.45 NPxG+xA90 is very healthy as well. There is no question that he produces a massive offensive output for a midfielder and causes lots of problems with the physical mismatches he creates at 1.91 meters tall. My only concern with Milinković-Savić is about his ability to advance the ball up the pitch, but the sole reason why he hasn’t made it to the final list is the reported €120-150M fee. If you’re wondering where those prices came from last season, and why his name hasn’t been mentioned at all in the past five months… Basically, he scored 12 non-penalty goals from midfield last season, catapulting his reputation into the stratosphere, when really, he outdid his xG by about 5.5 goals and this season he has regressed to the mean. That mean is still pretty impressive, but unless the asking price has dropped by at least €60M, the other players on the final list would provide a similar output, if not better in some areas, for at most two-thirds of the rumoured price for Milinković-Savić.
I have now rambled on for too long in this post, so the players who do make the cut will come in the next part of this series. I have actually finished the next part already, so the more people enjoy and share this post, the more encouraged I will be to publish the next one sooner.
I hope you did enjoy this post, even though I took it in a slightly different direction to the last one, which was largely forced on me because of the quality of the players already in this position at Chelsea and the nature of the data available to me. That said, I also chose to focus more on the requirements of the role, and the players who didn’t make the cut so that I could analyze each player that I do like more granularly, as I do in the next post.
If you do have any criticisms, suggestions, requests, feel free to comment on this post or tell me on my twitter, where I’m more likely to respond. Thanks for reading. 🙂