Where do we fit a football player in order to maximize their success? : A tool to help football scouts analyze players

Sammanfattning

In the world of football, AI and more exactly machine learning is used ever more frequently,where machine learning has not only become a huge part in team analytics but also in scouting.In today’s football, where teams are playing a lot more games and where the transfer windowhas become a tool in helping teams find players to minimize load and manage fatigue on theirown squad, it has at the same time become more common to find new players underachieve intheir new team. Therefore this report investigates an architecture helping the world of football toevaluate the most important variables for player- and team-success for a newly transferred player,analyzing where the player might experience success on the field. The aim of the project will beto identify what the key to a successful transfer is. This was achieved by identifying commonpositional transitions, thereafter creating machine learning models for each transition and usingplayer qualities as target variables to get the effect of a skill-set for the player in the new team. Thearchitecture then predicted qualities for three different transitions for a player, giving an exhaustiverundown on the position where the player experiences the most success. A team model wascreated, to gather the effect a player has on the playing style in order to gain a measure of teamsuccess. Although the results of the study proved that predicting qualities of a player makinga transfer to a new team works, with both models increased the prediction compared to theirbaseline models, providing an increased amount of information to use when making a scoutingdecision. Some positions were harder to differentiate, such as central defenders and full backand strikers and wingers, meaning that the positions rely on the same qualities available creatinga more difficult prediction.