How to use passing networks in soccer
/One of the fondest memories I have from my childhood was when my brother and I would sit in our basement, load up FIFA 11 on the Nintendo Wii, and play a kick-off match between our two favorite teams: Arsenal and Manchester United. 4-3-3 versus the classic 4-4-2. I wasn’t much of a tactico as a six year old, but I saw the game of soccer through the lens of formations; that’s how I learned to watch the sport, and that’s how I learned to play it. That’s how everyone learns to play it.
It wasn’t until I reached college that I began to really care about not just watching a game of soccer, but also trying to analyze it at the same time. This time, I found myself trying to focus on particular players and describe what I was seeing to better understand the match. It was in that search, spurred by my newfound interest in the game that I had loved and watched for my entire life, that I came across the passing network. The extent of my soccer stats knowledge until then was mlssoccer.com’s ‘Stats’ page and maybe the occasional FBref visit. The passing network drew me in precisely because it showed me something new about a game of soccer in the exact medium through which I learned to understand the sport: formations. It was the first data visualization I learned how to make, and it was the reason I began to think about making MLSStats what it is today. So today, I want to take a look at what passing networks are (for those that are unfamiliar), why they are useful, why they might have some drawbacks we need to keep in mind, and what they can tell us about the Major League Soccer season-to-date.
What It Is
The passing network is simply a graphic that aims to describe how the players on a team were actually positioned during a match. Using event data (a documentation of every pass, shot, defensive action, etc. that took place during a game), the location of each player on the field is found by looking at the average x- and y-coordinates of the passes that person played during the match. Then, lines are drawn between players, where the thickness — and sometimes color — of each line signifies various attributes about the passes that took place between those players.
The most common and basic style of passing network simply shows these average player locations and lines between them, where the thickness of the line denotes the amount of passes completed between each set of players. For example:
For the purposes of this article, I’ll just be focusing on the most simple of passing networks. Mine include the following two variables affecting nodes and lines: the number of successful passes played determine the size of each node and the number of successful passes between players determine the thickness and color of the lines.
Usefulness of Passing Networks
Passing networks can be useful in telling a story; they show where the players played on the field, which players played a lot of passes, which ones weren’t as involved, and can even show a team’s patterns of play. For example, a team might start the match in a 4-2-3-1, but play with one full back that drifted higher up the field while the other played more defensively in possession. The passing network is useful for visualizing trends like this. If a team plays more direct, their passing network might be stretched out, and if all a team can do during a match is rotate the ball across their backline, you could end up with a “horseshoe” style passing network:
A passing network is also fairly straightforward and not too difficult to understand at first glance. You can see the player numbers, take a look at the lines, and probably have a pretty good guess as to what the graphic signifies. However, there are a few issues with passing networks that are worth exploring, namely, that they’re built on averages.
Critiques of passing networks
The number one fault that can be pointed out when critiquing passing networks is their lack of nuance. Soccer is a constantly changing sport where substitutions can change games and formation changes can be made on the fly. A coach might instruct their team during half-time to completely change their style of play simply because they’re losing, while another coach might maintain the same formation but instruct certain players to switch positions to better suit the game. Capturing these complexities and in-game switches is difficult to do in a visualization, especially when that visualization is based on data that might not have that information readily available. This is why most passing networks only go until that team’s first substitution: if they kept going after the sub, there would be 12 nodes on the field!
Unfortunately, tactical changes are less easy to document than subs: as stated earlier, a LW and RW might switch sides at half-time, and because the average locations are exactly that — averages — their dots might appear in the middle of the chart, despite the fact that neither player played there throughout the match. A striker might occupy a position just on top of the opponents 18-yard-box while her team has possession, but because she rarely plays passes from that position, her dot might be nowhere near there. In essence, this core issue can be summed up by a few points:
Passing networks are based only on on-ball actions. They do not track player movement or action that happens off the ball, which limits the scope of what we can measure.
Passing networks can be impacted by many things: red cards, substitutions, injuries, etc. and so can be limited to only small portions of the match because a team had to make an early sub.
Tactical changes and formation changes are difficult to show in a single passing network, and they can skew averages such that players appear to have played in positions they didn’t actually play in.
The issues above exist for most passing networks, and are very important to keep in mind while viewing them, but they do not mean that passing networks are entirely useless. Passing networks are certainly better than nothing, and are often incredibly useful in showing general trends not just across a game but also across a season. Much like analytics as a whole, instead of thinking of them as the end-all-be-all, or a perfect description of a match, think of them rather as a tool in a toolbelt, one data point amongst many that can help us understand a team to a greater extent. With all of this in mind, it’s time to dig into what these graphics can tell us about how the MLS season has played out so far.
MLS Passing Networks
Now let’s look at passing networks for each team in MLS since the beginning of the 2024 season. Instead of looking at players’ average positions, I’ll be going by position, using that team’s most played formation and only considering passes played while that team was playing in that formation. For example, if a team played 20% of their minutes in a 4-4-2 formation and 80% of their minutes in a 4-3-3, I’ll only be looking at passes played while that team was in a 4-3-3, regardless of what player was playing what position. I’ll be looking to see what trends emerge, maybe explain why certain teams struggle, and try to break down various tactics based solely off of passing networks.
Here’s our first set of four passing networks (I’m ordering these roughly by position in the supporter’s shield rankings, as of writing), and I’m going to start with Inter Miami. Miami is by far one of the most attacking teams in MLS. Their game plan goes as follows: concede within 15 minutes, score five goals in a row. Their passing network on the other hand shows a pretty balanced approach: their most common formation is a 4-3-3 with Busquets at the heart of the buildup, the LW remaining high, and Messi at RW dropping in to collect passes and start the attack. There’s a slight lean towards playing more balls out to the LB instead of the RB, this makes sense, I’d much rather have Jordi Alba on the ball than Marcelo Weigandt (no shade to Marcelo). I think given some of the chaos that has surrounded Miami this season, it’s safe to say that their approach is pretty consistent albeit simple. They undeniably have the player quality, so there’s no real need to get fancy tactically.
In contrast, FC Cincinnati and Pat Noonan play their classic 3-4-1-2 which has resulted in so much success over the last few seasons. With a rock solid back line and attacking wing backs (notably Luca Orellano on the left), FCC rely less on a central distributor like Busquets and look more to play it into Lucho Acosta, who has become their most advanced player on the pitch despite being listed as a CAM. According to the passing network, their forwards can participate in holdup play, but are not as integral to ball progression and so are involved less in the graphic. It’s clear who their focal point of attack is, and for good reason, Acosta has 21 G+A from 16.3 xG+xA in 1736 minutes.
Real Salt Lake is less structurally rigid than Miami and Cincinnati and play a 4-2-3-1 with an inverted LW, advanced RW, and more involved central midfielders. The attack is lopsided — they complete more passes on the right side than the left — but deadly: RSL has the second most goals in MLS behind Miami. With Ruiz out injured, their CAM is less of an integral part of play, and almost acts as a RAM, with Luna often drifting in from the left to play as a LAM. This allows Katranis to get further up the field from LB, and plays into Phillip Quinton’s defensive and passing strengths at RB. Ultimately, it seems to be working quite well for RSL and specifically Arango at CF, who has been putting up MVP-caliber numbers this season.
To round out the top four, let’s take a look at LAFC. Plenty has been said about their lack of possession compared to years past, the weird spot they’re in as they wait for Giroud to join in the summer, and what a terrible choice that was by Ilie (I’m referring to his dreadlocks), but for now I’m just going to focus on their formation structure. With Bouanga and Olivera on either wing, it makes sense for them to be the most advanced players on the field, while Bogusz has played a pretty standard false nine (it’ll be interesting to see how that position changes in the passing network when Giroud arrives). LAFC also have one of the strongest on-paper midfield cores in MLS, and their setup is pretty solid in the network. Actually, now that I think about it, I’m pretty sure this LAFC team is exactly what you see when you select “4-3-3 false 9” on FIFA.
LA Galaxy have been dangerous in attack and at the same time can’t help but be leaky in defense. Their right side is deadly — just look at that triangle between Yamane, Pec, and Delgado, I mean talk about overloads. Paintsil and Joveljic have been so threatening in attack. Riqui Puig has shown no signs of slowing down. The only knock you can give them is a tendency to concede more than they should and tie games they could be winning.
Despite failing to make the playoffs last year, NYCFC have been impressive this season, sitting near the top of a strong Eastern Conference. Hannes Wolf, Agustin Ojeda, and Santi Rodriguez have been pivotal in attack, forming almost a front 4 alongside Bakrar or Mijatovic. In possession, this creates something akin to a 4-2-4, with pinched-in wingers and space for their fullbacks to progress. Although they have a far less potent attack than RSL or Miami, NYCFC have conceded just 22 goals from 26.1 xGA. Their defensive strength can probably be attributed to a good rest defense, but likely moreso to an excellent Matt Freese, who has 8.0 (!!) goals added (g+) — good for first in MLS goalkeepers — on 72 saves (third in MLS).
For a team that has been embroiled in drama surrounding their DP star, has a first year head coach, and has their supposed top striker sitting on two goals through 15 games, Minnesota has not looked bad this season at all. Robin Lod has been truly incredible for this team, and you can see that in the passing network. Whether it’s lining up at RCM and advancing into a second striker role or playing as a wide and high RW, Lod has been all over the place and has five goals and seven assists to show for it. The Finnish Messi debate is over, sit down Robert Taylor.
The Red Bulls have had a good start to the season this year after barely scraping into the playoffs last September. They are third in the Eastern conference and sixth in the league according to ASA’s xPoints model. Emil Forsberg has proved to be an incredible offseason acquisition despite the hit-or-miss nature of signing aging stars. Structurally, they’ve played a straightforward 4-4-2 with Forsberg and Carmona pinching in to act as dual 10s — this allows Tolkin and Dylan Nealis to move forward, both of whom have shined this season. It’s fair to say this has been working for the Red Bulls, they have 11 goals and seven assists on 14.2 xG+xA from Morgan and Vanzeir, and Forsberg has fit in very nicely with six goals of his own.
Is the Columbus Crew the new 2022 Seattle Sounders (except without a CCL trophy)? Probably not; they run an almost perfectly executed 3-4-2-1, and when healthy, there’s nothing stopping this team from repeating.
Toronto, on the other hand, play the exact same formation in a completely different way. Instead of playing with their two AMs inside and behind the CF, they play them out next to their (awfully narrow) WBs, allowing for interplay in half-spaces instead of centrally or in wide areas.
As for Vancouver, I think the most normal thing about their passing network is their back line. I honestly don’t even know what to make of their network, maybe this falls into the aforementioned issues category, unless they're playing a new formation with three LWs.
Charlotte is playing a well-spaced 4-2-3-1 but could really benefit from a DP signing at CAM… or I guess 350k in GAM?
Again, it’s interesting to see the same formation executed wildly differently between teams. All four teams here are playing a 4-2-3-1, but none are alike. Austin FC has a very solid network with both wingers high up the field, but could probably benefit from a more involved nine.
Colorado has benefited greatly from Mihailovic’s arrival and Basset and Navarro have been on fire. I imagine they’d like a lot more out of their wingers, though.
Portland’s passing network is slightly less advanced up the field than others, with Evander clearly all-encompassing in attack (including the CF node) — though offense has not been the issue for them this season.
To round out the four, Houston is completely lopsided and playing with a very advanced Dorsey at RB, Bassi pushing into the half-space, and Aliyu keeping the width.
DC United’s offense
If there was ever a passing network that perfectly described Route One football, it would be D.C. United’s (granted, they’ve played only 500 minutes in the formation). I mean holy crap. I guess, if anything, it encapsulates their game plan this season: get the ball to Benteke.
Philadelphia is all over the place, but has clearly made it work — Gazdag already has over twice as many non-penalty goals as he had all of last season.
Nashville’s passing network does a pretty good job of summing up their season so far: could do with more attacking threat, not super exciting, although they haven’t played as many minutes in that 4-2-3-1.
The Sounders have a pretty well-balanced passing network but have suffered from a lack of any real production from the wings.
Orlando provides a lot of space wide for their fullbacks and plays with a front line of four, using the wingers as inverted attackers. Despite pretty high expectations coming into the season, this group hasn’t been able to fully click yet — just 19 goals scored this season.
CF Montreal is really trying to be Columbus right now, just not to huge effect. More than anything, their defense has let them down this season, so that will definitely need to be an area of improvement.
St. Louis is also playing a pretty standard 4-2-3-1, however, they tend to play with an inverted RW and high LW. Klauss played a big role last season at CF, but just hasn’t hit those numbers (yet) this season. You can tell he’s involved in play, though.
Atlanta United is another interesting case. On paper, they have a very good squad, though that might be changing with the departure of Giakoumakis and potentially Almada as well. They’re set up in a pretty standard 4-2-3-1 with slightly more narrow wingers, however, it will be interesting to see how or if that changes with Piñeda’s departure.
FC Dallas is also in the three at the back crew, just haven’t been executing it as well as they should be; the attack has been lackluster (setting aside their most recent games) and the defense has been poor. Musa also hasn’t been putting up the numbers they expect from him, though judging by his position in the passing network, that might be in part due to lack of service.
The Fire haven't quite been that bad in defense, but they haven’t been good either. A much more balanced network across the back line, but their verticality is low.
Sporting KC is about as well spread-out as you can get for a team near the bottom of the table, but you can tell they need more out of their attacking trio and back line.
The Earthquakes have made some good signings this season and have improved their attack, but that defense has been killer. They’re not maintaining enough possession (2nd to last in the league) but I will say they’ve been screwed on the xG: 30.5 conceded but 45 goals let in. That’s just unlucky.
And last, but not least, New England. No massive surprises here, Carles Gil is still the man for the Revolution, though they really need to be getting more from Vrioni. Gil is the only one on the team with more than two goals, and total the team only has 15. It’s just sad. Otherwise, pretty standard 4-2-3-1 with no major tactical wrinkles.
Conclusion
As much as I enjoy passing networks — and formations — there comes a point where they can take you no further. Although the past ~3400 words have been a fun exploration into what passing networks can tell us about the game of soccer and the teams that play it, that point has been reached. I still love the passing network; it’s useful at a glance, nice to look at, and good for going a little bit deeper than a team sheet. They’re not perfect, however, and there are certainly a lot of ways you could improve them, though that’s a piece for another day.
Hopefully this article was informative, whether you heard of passing networks yesterday or whether you’ve grown sick and tired of seeing them all over your feed. Thanks for reading!
For more content, follow me, @mlsstat, and @nwslstat on twitter.