• September 21st, 2016

Is it possible to predict the transfer value of a football player? During my studies of Business Mathematics and Informatics, I gave it a shot.

In anticipation of our master’s thesis, we were asked to conduct a small study on a topic of our choice. Being a football fan, I decided to investigate the possibilities of predicting the transfer values of Dutch football players. To do this, I used a multiple linear regression model. Regression is a statistical technique for analyzing data when there is a possible correlation between the data. Multiple regression means that there is one dependent variable (in this case the transfer value) and multiple independent variables (in this case characteristics of the player, of the old and new club, and the period in which the transfer takes place). I composed the dataset myself, using the internet and football yearbooks that I found in my father’s closet. I used the data of 5 successive years of Dutch transfers. To come up with possible variables, I researched multiple English studies to make a model as complete as possible. The model is referred to here:

“Thursday 26 March
What determines the transfer fee of a football player?
For years, scientists have been searching for the ideal formula to calculate the transfer fee of a football player. Why was Christiano Ronaldo worth 94 million euros when he transferred from Manchester United to Real Madrid in 2009? In the Netherlands, Jarno Suijkerbuijk attempted to build a calculation model during his studies of Business Mathematics and Informatics at the Vrije Universiteit in Amsterdam. In 2010, he researched over five seasons of the Dutch football league. After a lot of calculating, he came up with a formula in which you have to insert 32 characteristics of a player and his football club. If a player scores a lot of goals or often gives an assist, he is obviously more valuable. Fewer red cards indicate a higher value, just like white skin (!). The value of a player also increases if interest has been shown by a rich club. A convenient formula then? Quite a bit. But in only 70 percent of the cases did the model more or less predict the right transfer fee. So actually football players are like works of art: they are worth whatever you can get for them.”

The above quote is taken from the Quest calendar (Quest is a popular Dutch “brains and entertainment” magazine). Of course, I regard it as quite an honour that Quest decided to place my research in their calendar. However, this particular editor had scanned the results too fast. Part of my conclusion  specifically stated that racism factors like skin colour and nationality are not significant when determining the transfer fees of football players in the Netherlands. (The same applies to the number of red cards.) A painful mistake. And given the exclamation mark that the editor placed, this is precisely the information that the editor wanted to score with.

The existence of this calendar was actually not known to me nor was I asked to contribute to it. My research was 4 years old by the time the calendar was published! However, the above article does show that mathematics and personal passions can be nicely combined. As a matter of fact, I still use the kind of model (linear regression model) from the football research in my current job. Although I now use it to perform data-analysis on pension entitlements and insurance premiums.

Jarno Suijkerbuijk is a mathematician and works as a Business Analyst at RiskCo.