Authors: Jaelyn Eisenhart, Adi Rosenstock, Joe Tomaskovic

jaelyneisenhart2023@u.northwestern.edu, adirosenstock2026@u.northwestern.edu, josephtomaskovic2027@u.northwestern.edu

Known for its extreme competition and global popularity, the English Premier League has star players from around the world who are measured by their in-game performance. Clubs are constantly evaluating their players to properly manage their organization competitively, fiscally, and strategically. Because of the EPL’s relegation rule, in which the clubs that finish toward the bottom of the table are at risk of being sent down a division, teams prioritize any form of competitive advantage, whether that is with players on the field or in the transfer market. Valuation metrics and analyzing data are key ways for teams to achieve an edge in such a dominant league.

We want to uncover which players in the English Premier League were undervalued during the 2022-23 season by looking at their offensive value and annual salaries. We started by gathering our data from fb.ref. To determine offensive value, we used fb.ref’s non-penalty expected goals plus assisted goals per 90 minutes (npxG+xAG/90). This metric includes expected goals, taking various factors into account like shot location, defender positions, angle, or type of play. It quantifies the likelihood of a goal-scoring chance without counting penalty kicks which occur from a stationary ball at the 12-yard line. While quality PK takers are very important for a team to have, we were hoping to look at players who could generate more creative chances during the run of play. The metric also includes expected goals assisted which measures how likely an assist will lead to a goal-scoring shot. Scoring is important, but passing the ball to the person who scores is arguably just as important, and we wanted a measurement that took quality assists into consideration. It is important to note that the metric we base our analysis on was per 90 minutes which helps standardize the numbers regardless of the amount of minutes a player has on the field, and makes it easier to compare players’ offensive contributions to scoring goals.

Our data collection consisted of the top 5 players per EPL team with the highest npxG+xAG/90 during the 2022-23 season and their respective annual salary for the given year. As the league has 20 teams, there were 100 players analyzed. While this number seems small, it gives us a snapshot of the given season and relevant players in the top league. In a future study, we could expand this pool to other leagues and across multiple seasons for a broader picture. Looking at Manchester City, the players with the top npxG+xAG/90 were Erling Haaland, Kevin De Bruyne, Riyad Mahrez, Phil Foden, and Jack Grealish. Their non-penalty expected goals plus assisted goals per 90 minutes were .92, .72, .55, .52, and .51, respectively. The relevant columns we needed to conduct our analysis were team, season, player, npxG+xAG/90, and salary.

Initially, we created a graph to determine if there was any correlation between the players’ salaries and npxG+xAG/90. NpxG+xAG per 90 minutes is seen on the x-axis and salary is listed on the y-axis seen in Figure 1. In a perfect world, a player with a higher npxG+xAG/90 would have a higher salary. That would mean they were getting compensated for creating more success offensively. However, soccer has many contributing variables that affect a player’s offense production per game and season, so it isn’t that simple. The red trendline shows a slight positive correlation and even though it is increasing, it is very weak with clear contradicting data points.

After determining that npxG+xAG/90 and salary did not have a strong linear relationship, we decided to conduct a further analysis directed toward specific players. We compiled all the players with a npxG+xAG greater than .7, which encompassed the top 10 out of 100 players. As we can see from the table, Reiss Nelson had the top metric with 1.04. He is an Arsenal academy player who played in 11 games during the 2022-23  Premier League season. From those games, he scored 3 times. He was an impact substitute who made the most of his opportunity as a 24-year-old. Nelson wasn’t a regular starter, but he had the highest npxG+xAG/90 because the measurement is based on per 90 minutes of play. By this metric, Arsenal should play him in Premier League action more often and see if he continues to bring a strong offensive presence. There are also big-time EPL stars in this list including Erling Haaland, Gabriel Jesus, Mohamed Salah, and more. The wide range of salaries highlights that certain players are undervalued based on their offensive value in the form of non-penalty expected goals plus assisted goals. For instance, Deniz Undav had the sixth highest npxG+xAG/90, yet the lowest salary by far.

We used the same data to create another visual of the npxG+xAG/90 separated by each team in the English Premier League. In Figure 3, the legend on the right has the team and their respective color. The names listed are the top 10 players from the graph above. This group of players stands out with a higher npxG+xAG/90 than the rest of the player pool. Some teams have deeper rosters with multiple players above a .7 npxG+xAG/90. Additionally, the graph can make a case for why players should be awarded less or more money for their performance. It would make sense if Reiss Nelson got an increase in salary and more opportunities in the EPL because of his contribution to the offense when he did go in. Coaches and general managers need to recognize that they can get players who create the most offensive value at a price that makes sense for their budget. They might need a center forward that contributes to the offense by a certain percentage but within a specific price range. They can use a model like this to examine center forwards in the transfer market and use npxG+xAG as one of the factors. 

Our analysis helped us determine the relationship between the offensive value a player contributes and how much they are worth for a given season. However, there are a few limitations and notes we wanted to mention that could improve future analysis. We used non-penalty expected goals plus assisted goals per 90 minutes for our metric to determine offensive value. This is not the only thing that determines if someone is a good player. For instance, even though Marcus Rashford isn’t one of the top 10 players in the league with the highest npxG+xAG/90, you still want the 2022-23 PFA Fans’ Player of the Year on your team. If we wanted to expand our analysis in the future and look at trends over time, we could compile data for more seasons from the English Premier League. In a future analysis, we could divide the players up by position to look at how position correlates with offensive value. Even though most of the players we looked at were forwards and attacking mids, a more defensive-minded position that gets forward a lot could create a larger npxG+xAG/90 than a player whose starting position is higher up on the field. If we were to repeat this study, we could include a threshold of a minimum of 25 games which would have excluded Nelson from the data pool as he only played 11 games. It is also important to note that the salaries were gathered from fb.ref and some were deemed as unverified estimations. We assumed this number was an accurate salary number, but we recognize the numbers could be slightly off which could affect our results.

 

Sources

Fbref. “Football Statistics and History.” FBref.com, fbref.com/en/.

Sam. “Investigating Team-Adjusted Statistics.” Medium, 13 Mar. 2023, gustasam5.medium.com/investigating-team-adjusted-statistics-ef2dd71a8d7a. Accessed 25 Feb. 2024.

Tweedale, Ali. “Expected Goals: Explained.” The Coaches’ Voice, www.coachesvoice.com/cv/expected-goals-xg-explained/.