Elo Ratings in Football — Creating Club Transfer Profiles
Hungarian-American professor Harpad Elo developed the “Elo Rating” as a means to rank chess players more fairly and accurately than previous models. His system was implemented by the U.S. Chess federation in 1960 followed by the World Chess Federation shortly after. While the formulas behind the ranking system might seem irrelevant to the average football fan, the concept can be applied to football (soccer) just as it is in chess. More specifically, I propose that Elo Ratings and Football (Soccer) Power Rankings can be used in a different way than ranking clubs. Instead, Elo Ratings provide data conducive to developing “transfer profiles” of professional football clubs. These transfer profiles are reverse-engineered reports based on a club’s incoming and outgoing players in a particular transfer window that put into numbers what takes a lot of words to explain. These transfer profiles are more accurate when applied over 3–5 transfer windows, for the duration of a manager’s tenure, a director of football’s career, etc.
What types of clubs do “Club X” sign players from? What clubs does “Club Y” send players to when they are no longer needed? This article aims to explain how the average reader can create club football transfer profiles and through these profiles, answer the above questions in regards to their favorite club(s) by using a metric that I have developed called “Net Elo.” While we know that top clubs scout hundreds of players a year, knowing where they ultimately sign from/transfer players to might lead to patterns over a few seasons.
What is Elo?
The general objective of the Elo rating system was to create a more accurate and fair system to evaluate the skill of chess players in relation to other chess players. The more wins one player has against quality opponents, the higher the score and vice versa. Therefore, the more matches any player plays, their Elo score should constantly self-correct to give them a more accurate score based on their skill compared to other players and past opponents.
There is no official rating system for Elo scores in football. Regardless, there are various organizations and websites that keep track of Elo and similar ratings based on their own particular formulas. Each formula gives more/less weight to different variables like home-field advantage, goal differential, etc. I try to refer to a few different sources to get a better idea of how clubs are performing over time. The primary source I use is Clubelo.com due to the fact I can see a club’s rating on any date where the website has data. More on this later. The only downside is that it is limited to European clubs although it includes clubs from smaller countries and lower divisions.
One of the others I refer to often is the FiveThirtyEight Global Club Soccer Rankings shown above. This source includes non-European clubs but not all lower division European clubs or clubs from smaller countries in Europe. Also, the FiveThirtyEight SPI rankings are not adjustable/searchable by date which makes it difficult to build transfer profiles unless you can record ratings when a transfer is made official. (This can be worked around sometimes by using the Wayback Machine: web.archive.org. The archive stores data from websites on any particular day and it is easier to find data for more popular sites. It is sporadically available for FiveThirtyEight but helpful sometimes.)
Building Club Transfer Profiles
The club transfer profiles are better built during the off-season (roughly May-August for European clubs) when clubs are not playing matches and do not move up and down. The profiling works by starting with one club and looking at all of the players they signed during a transfer window. The Elo rating of a selected club, “Club X” is used and the difference between “Club X” and the Elo rating of club(s) where new players are signed from, “Club A”, “Club B”, and “Club C”, is derived to what I have determined to be the “Net Elo”. The Net Elo is valuable as it gives the fan, player, and football professional an idea of where a club signs players from and sends players to. This information can lead to reverse-engineering and inferring what types of clubs are scouted more/less often, if a club has positive relationships with another club, or if a club has a formal/undisclosed developmental relationship with another club.
On the surface, a fan might think that FC Barcelona only signs players from lower level clubs because there are hundreds of clubs with lower Elo ratings than FCB but only a few with higher ratings at the moment. The fan would be correct. Similarly, Birmingham FC, currently playing in a lower division in England, might be expected to sign more players from better clubs where they might need minutes to prove themselves, recover from injuries, or if they are older, on their way to retirement and hope to play a few more years professionally. This is more difficult to predict because still there are hundreds of teams below Birmingham.
Where it gets interesting is looking at top tier, maybe not Champions League winning clubs, but still strong clubs, that are known to have amazing scouting networks. Among these, Borussia Dortmund, Sevilla, and the Red Bull clubs (Leipzig and Salzburg) come to mind. Where do these clubs find the players that turn into success stories that start at the club for a decade or are sold for €10–20+ million profit? I have broken down a few examples below that can be replicated and applied to multiple transfer windows to get a better idea of club profiles over time. This gives a clearer picture of long-term scouting and recruiting practices.
Atalanta Incoming Player Transfer Profile 2018/19
So, we have looked at the players that Atalanta brought in during the 18/19 season and a few things jump out at us. Prior to even considering the numbers in the chart we can see that a handful of players were brought in from “big” clubs like Inter, Benfica, AS Roma, and Chelsea. Even Zenit St. Petersburg could be considered a “big” club considering their recent domestic success. This might tip us off to think that Atalanta made some moves for players from high profile clubs. Also consider that players were signed from a lower-level Polish club, Serie B, the English Championship, and some mid-lower level Serie A clubs. Also, only 1 fee scraped past €10 million.
The signings listed above have been segmented in the chart below with corresponding Elo ratings of the departing club and Atalanta on the date of the transfer as reported by Football-Italia.net. Data was not plugged in for transfers where the data involved a South American club or club from a lower division not ranked by Clubelo.com.
The numbers to pay special attention to are the Net Elo per transfer on the far right of the table, the Average Net Elo of all signings, and the Total Net Elo. The Net Elo is determined by subtracting the Elo Rating of the other club(s) from the base club (Atalanta) here. The closer to 0, the closer in Elo rating the clubs were at that time. As it approaches a very low negative number, the player was signed from a much better club. Conversely, as the number gets higher positively from 0, the player was signed from a much worse club according to Clubelo.com. A high, positive Avg. Net Elo means players were signed on average from much worse clubs.
Atalanta’s Elo rating fluctuated barely over the summer window due to some clubs having pre-season and early season league matches prior to the start of Atalanta’s season. Aside from the “eye test” of the transfers, noting that players were signed from a handful of “big” clubs, the numbers in these transfers do not seem too big of a deal. The loan of Pasalic from Chelsea and signing of Tumminello from AS Roma stand out as the highest profile incoming players based solely on their departing club. But these were just two on the list of 11 players considered in the Average Net Elo and Overall Net.
Another interesting statistic to note are the relatively small differences in rating between Inter, Zenit St. Petersburg, and Benfica compared to Atalanta. At this point in time, Atalanta’s performance was trending upwards compared to a stagnating Inter. This means 4 of the players came from comparably-ranked clubs at the moment of their signing; this is not even half of their signings.
Lastly, there are 5 players left. These 5 players dramatically brought the average Net Elo higher and the Overall Net much higher. While clubs like Sampdoria and Udinese are respectable top-flight clubs in Italy’s Serie A, the difference between them and Atalanta is greater than the difference between Atalanta and top clubs like Chelsea and AS Roma. Atalanta was ranked #29 overall during this timeframe, Chelsea at #13, and Roma at #12. In Elo Rating, Atalanta was an average of 107.5 points apart from these clubs. A fairly big difference between just 16 and 17 rankings.
With the average of Atalanta’s calculable signings from this window, +81.18, we can see the “average club” or type of club Atalanta seemed to sign from most, and likely scout most during the search process. At the time of the 18/19 window, Atalanta was roughly 75–85 points better than #66 Stuttgart and #69 Southampton and #70 BSC Young Boys. This results in a more drastic difference in ranking compared to the “big” signings Atalanta made in this window.
Again, this is just one window. However, Atalanta seemed to scout many lower-ranked teams in 2018/19. These are by no means bad teams, but perhaps they were clubs that could not hold on to more promising players or players that Atalanta determined to be necessary for their creative sporting project during that season.
RB Leipzig Outgoing Player Transfer Profile 2018/19
Here we see players transferred entirely to Europe (except Hapoel Tel Aviv, perhaps.) This is a big help in using Clubelo.com to develop a comprehensive profile for the outgoing transfers in this window. We can generally see that there are a lot of lower-tier clubs with really the only exceptions being Liverpool, RB Salzburg, and Brighton. We also have some second and regional division teams that might not be included. So, let’s look at the numbers!
Looking at this set, we have an even higher Average Net Elo and Total Net. This is also definitely even lower than it would be if the lower-division/regional league ratings were available and plugged in. We can see that really aside from the headline transfer of Naby Keita to Liverpool and the goalkeeper Köhn to RB Salzburg, every departure left to a worse club. RB Leipzig was ranked #51 of all teams on Clubelo.com. The lowest ranked club they transferred a player to was #432 Dunajska Streda. The average clubs RB Leipzig could be expected to send players to are Huddersfield, Genoa, and Guingamp, ranked between #130, #132 and #134 overall.
RB Leipzig sent only a handful of players to much better clubs in this window. Otherwise, players left to much worse clubs in 2018/19. This is interesting in comparison to Atalanta, who signed the majority of their players from much worse clubs except a couple of bigger names. These two clubs likely share similar transfer budgets, wage bills, and work hard to keep top performers to build long-term projects.
Conclusion
These profiles are difficult to measure in live time with limited access to transfer data/when a transfer is made official. Also, different sites use different metrics and there is no uniform official ranking to go off of — one might need to compare multiple source rankings and average the ∆ between multiple sources to have the most well-rounded number for a particular signing. Also, one window gives a very limited look into a club’s scouting and their signings. All sorts of factors go into signings but applying this process to one window gives an idea of what clubs/leagues are being scouted or where players are moving to. Applying this to 3–5 year periods and/or the career of brilliant sporting directors like Michael Zorc at Dortmund or Monchi at Sevilla are future topics I hope to write about eventually. This profiling concept is just one piece of the puzzle that could be used by fans and football professionals as they attempt to comprehend the evolving transfer market.
Thank you for reading. Please feel free to check out my analysis of various players like Nordi Mukiele and Eduardo Camavinga as well as my article series on modern football player transfers. I am also open to conversations of all sorts on twitter: @rschmidtke8.
