Football scouting system

Abstract

The research carried out focused on the development of a weighted ranking mechanism for football players. Through the use of a public dataset, a tool was implemented to allow football scouts to easily analyse and discover players which may be of interest. The FIFA dataset which was rich in data, high in data volume and also recent was used throughout the research. This dataset was subsequently enriched by the inclusion of additional players attributes from an external source. All the players were categorised according to their playing position and then data analysis followed for each position. Attributes which were found to be relevant for each position were selected and taken into consideration for the ranking mechanism. The data analysis phase was initiated by performing a weighted average for all the attributes selected. Further attributes were selected through identifying positive associations with other relevant attributes. This was achieved through the use of the Pearson correlation coefficient by analysing the coefficient value for each pair of attributes. All the attributes were given a weighting, by calculating the geometric mean for each attribute and normalising the weights out of 100. An overall ranking for each player was generated by applying the attribute weights in the weighted average method. Data visualisations were adopted to visualise the player information in a variety of ways. Real-time data was also included through the integration of public data feeds into the system. The player rankings were evaluated for each position using Root-Mean Square Error with the FIFA and Football Manager datasets. This was applied by performing a comparison between the system's ratings and the ratings of FIFA and Football Manager. The results in comparison with the FIFA dataset were quite satisfactory for all positions, whilst the results for the Football Manager dataset varied accordingly for different positions.