If you are joining this series for the first time, be sure to check out Part 1 and Part 2 in this series on machine learning in Major League Soccer. To recap, using a k-means clustering technique (a machine learning technique to look for similarities within data and group things based on closeness), we grouped MLS players into 27 unique player-roles.
A Brief Intro...or Recap?
You might be asking yourself the question of “why are you doing this?” First and foremost, I believed that the methods used currently to group players together were too broad and not refined enough to describe and distinguish between players like Zlatan Ibrahimovic and Adi. Both players are big, strong target players who hold the ball up and get into the box to score goals. The problem here is that while we can group based on quality of play, but that is too simplistic and only reflects results rather than smaller details like direction of passes, type of passes played, type of shots, and so on and so forth. Zlatan belongs to a role that I have classified as “Goal-Scoring Machines” (GSM), while Adi sits within the “Physical Attacking Specimen” role (PAS). In the PAS role we see players like Gyasi Zardes, Adi, Jozy Altidore, Jordan Morris, and Nemanja Nikolic; and to contrast that the GSM role contains players like Wayne Rooney, Josef Martinez, Robbie Keane, and Sebastian Giovinco. When watching a soccer game we can separate these players-roles with little difficulty but the majority of these players are typically called “strikers” or “forwards”, which leaves much to be desired. Goal scoring plays a large part in this role’s classification, as does shot efficiency, but statistics like fouls committed, tackles attempted, and length of key passes. Some of those are rather intuitive while others may not be exactly what we think of when considering a top goal-scorer.
All of that is a long-winded way of saying that i believe we can create a more comprehensive model to classify players. The model used here contains only 5 seasons of MLS player data ranging from 2014 to 2018. A more holistic approach would account for more statistics as well as bringing in data from other leagues to compare and contrast. The difficulty in doing such a thing would be somewhat high as that type of analysis would necessitate a weighting scale for leagues worldwide, but that is a task for another time. My task was clear when I started to look at player positions, I would attempt to create a better method, and here we are a couple of weeks later!
Part 3 - Revisiting Team Change & Growth
A person who I greatly respect suggested that perhaps MLS Cup winning teams were not the only comparison that would be valuable to readers, and I completely agree. In this section I will briefly analyze two teams that made astounding jumps in points earned from one season to the next - Montreal Impact in 2015 and Chicago Fire in 2017.
Montreal 2014 to 2015
In July of 2014 Montreal acquired a player named Ignacio Piatti who would arrive in Canada later that year. In his first full year of play for the Impact, Nacho Piatti was a revelation and firmly placed himself into a role in which the Impact had no one on the roster. Other players who fit into this role are Diego Rossi (LAFC), Aleksandar Katai (CHI), and Luciano Acosta (DCU). Additionally, the team picked up Marco Donadel who occupied a deep-lying role in the midfield - another role that the club did not have in 2014. Lastly, the club added another attacking player - Dominic Oduro. This doubled the players they had in this role, adding to Jack McInerney and providing a burner that could run past, through, and around defenses. These three changes helped contribute to the enormous turnaround that the club had, gaining 23 more points in 2015 than they earned in 2014.
Chicago Fire 2016 to 2017
The Chicago Fire of 2017 is another example where we can look at our player-role clusters and observe changes made by the team from one season to another that lead to a dramatic turnaround. First, we should note two major acquisitions that the Fire made between 2016 and 2017: Nemanja Nikolic and Bastian Schweinsteiger. Those two players came into the team and made an immediate impact on the team as a whole, as well as moving players form one position to another within the team dynamic.
Nikolic came to the fire and immediately took on a role that the team did not have previously. He fit into the category with Fanendo Adi and Jozy Altidore, as a physical specimen sitting highest up the field. This helped the Fire have an outlet man and a player who would put the ball in the net pretty regularly. Schweinsteiger came in and forced Matt Polster to move out of the holding midfield role, where he had won Rookie of the Year mind you, and play a different role for the Fire. The upgrade in quality and experience is apparent in this example, and we should note that the role group that Polster occupied was not filled but rather replaced with a deep-lying playmaker role (the same role that Michael Bradley played for Toronto FC’s championship winning team).
The two new players mentioned above influenced the Fire in a predominantly positive way, but there were a few other players who made changes and contributed to the team’s turnaround. Dax McCarty came in before the 2017 season and slid in next to Bastian Schweinsteiger. Those two players essentially replaced Razvan Cosis and Matt Polster, providing an non-trivial upgrade in the midfield while not changing the team layout. Dax brought energy, hard work, and veteran savvy, finding himself clustered into the same player-role shared by former FC Dallas’ homegrown Victor Ulloa.
Lastly, the Fire pushed Arturo Alvarez into a slightly different role in 2017. He took on fewer players per game but succeeded at dribbling opponents at a higher rate, his passing rate increased, and he turned the ball over about 65% LESS in 2017 than he did in 2016. Additionally he played a higher percentage of his passes short (70% in 2017 vs 64% in 2016). All of that said, he played about 800 fewer minutes in 2017, and likely benefitted from having two new, talented midfielders who have a good grasp of spacing and positioning working around him. His role became less important to the team overall, as he went from a playmaker role to one of a role-player: occupied by players who are on the ball less, make fewer passes, take fewer shots, and take fewer risks with the ball. Schweinsteiger and McCarty taking over the majority of the midfield work and Nikolic spearheading the attack all contributed to the dramatic turnaround that the club had from 2016 to 2017, gaining 24 more points from one year to the next. Looking at the player-roles we can see some striking changes in these two teams who did not win championships but, with a few key changes were able to turn the fortunes of their club around in just one offseason.
Young Player Growth & Projections
An application that I have not spoken much about is player development and how the player-role clusters can help us see players developing and taking on more important roles within their teams. Let’s take a look at a couple of examples of players growing and blossoming into key players for their teams in the past few seasons.
Player Growth #1 - Cristian Roldan - Seattle Sounders
In 2015 Cristian Roldan was the SuperDraft pick that the Sounders so desperately wanted, and they moved up to land him. During that first season he played a solid role for the team and helped the team before finally anchoring himself into a position next to Ozvaldo Alonso in 2016. His rookie year, Roldan occupied role number 14 in our clusters, which is a very generic role for midfield players who do very little that is exceptional. The passing percentage tends to be mediocre, key passes, interceptions, tackles are all average and the only thing that tends to be high with this group is the loss of possession category. As a rookie, this is understandable and Roldan learned much going into his second year.
By 2016 Cristian Roldan had shaped himself into a very solid contributor and has occupied the same player-role for the past three seasons, the “windshield wiper” role. Championship teams tend to have at least one if not two of these players on their roster and with Roldan , the Seattle defense was protected and the attacking players had constant service played to them. His development is what one would hope for when looking at young players typically associated with the “holding midfielder” role.
Player Growth #2- Justen Glad - Real Salt Lake
Glad is still quite young, at just 22 years of age, but has already grown into quite a competent defender in his few years with the club. Though he won defender of the year with RSL in 2016, his biggest transition came in that offseason when he transitioned into the player-role that is shared with players like Matt Hedges, Chad Marshall and Laurent Ciman. After such a short time playing, the transition into playing so similarly to several solid veteran defenders is no small task. Going from a player-role grouping that included Daniel Steres and Aro Tarek to one held by player who are often considered among the best in the league shows tremendous growth and understanding of the game.
These two players changed over the course of an offseason and provided a large boost to their teams with their higher-level of play. Let’s see if we can find any players that we can compare to these guys and project them forward to see who might break out this year! I will give you a few players, where they were in 2018 and what roles I think they will project into by the end of the 2019 campaign based on their player-roles.
Player Projection #1 - Alex Muyl - New York Red Bulls
After the each of the past three seasons, Muyl has found himself in a group of players that put up workman-like performances and average player rankings among the typical statistics. Already in his fourth year with the Red Bulls, he is still only 23 years old. This season he seems to have started off rather well with two goals in three appearances, is attempting more dribbles (and succeeding on more of them as well), and perhaps most telling is that he is being dispossessed at a relatively low rate - at half of his career average. If Muyl continues his play and this new run of form, his player-role will transition into the versatile attacking midfielder role. While not the flashiest nor the highest profile role, this one is common among successful teams and these players provide a variety of things to teams from goals to defending to keeping possession. This would be a big get for the Red Bulls if they can have another young player develop into a valued asset.
Player Projection #2 - Cristhian Paredes - Portland Timbers
At 20 years old, Paredes is poised to break out this season. The loanee from Club America has already matched his goal total for all of last season, but more importantly he is helping in some important ways for the Timbers. His poor touches are down to 1⁄3 of the number from 2018 last year, his times dispossessed is down, key passes is up, and perhaps most importantly he is being fouled at nearly four times the rate of last season. To be fouled, one typically needs to be near the ball, and that is a good thing for a midfielder. Last season Paredes was grouped with players like Tommy Smith and Tony Tchani but this year he is projecting to be in a more important role, joining the company of Tyler Adams and Anibal Godoy.
Player Projection #3 - Aaron Herrera - Real Salt Lake
And lastly, we have Aaron Herrera - a product of Real Salt Lake’s high quality youth system. Starting the 2018 season as the first choice left back for the club, Herrera had plenty of opportunity to develop and this season he looks to turn a corner and establish himself as one of the best left backs in the league. Last season he was a true left back grouped among players like Oniel Fisher and Jerome Thiesson but with his increased aggressiveness both offensively and defensively, Herrera is poised to take on the wingback role that is so crucial in the modern attacking game. Having already passed his 2018 season total in shots and assists, Herrera has planted himself into an important role for the club in the offensive end of the field while doing work on the defensive end as well. His passing percentage is down but that comes with taking more risks with the ball.
Time will tell with these players if they establish themselves into more elite roles or not, but there are three players that our clustering analysis has identified as likely jumpers from one role to a more exclusive role. Thanks for reading this analysis series, if you would like to see more pieces like this then please let me know. I have a couple lined up for the season but would like your feedback either way. Please be sure to contribute in the comments and let us know what your thoughts are about the advancements in soccer analytics and what applications you would like to see in the future.