SoccerAtlas: a web framework for visual soccer analytics

Kenny J.

data science  data visualization  sports analytics  soccer analytics  spatio-temporal data 

This thesis discusses the design and development of SoccerAtlas, a web framework for visual soccer analytics. During the past decade data science have entered the world of sports and large amounts of hi-fidelity soccer data are now readily available, collected by several companies through advanced semi-automatic sensing technologies. However soccer analytics is a young field of research, and the wealth of the collected data is not yet actually supported by adequate analysis tools. This is particularly true for soccer, which among all the team sports is the most difficult to quantitatively analyse due to the complexity of the play and to the low number of scores which determine the result of the game. SoccerAtlas is a web application that allows the soccer analyst to visually explore soccer data, interact with it, and gain new insights. It exploits several analytical functions to extract useful information from data, then visualising it in an appropriate way. This thesis discusses all the steps that lead to the final web framework, taking into account both technological aspects and visual communication aspects. Particularly, these steps include the definition of a theoretical framework which illustrates the abstract model of the data, the formalization of the analytical functions based on this model, the design and the implementation of the framework. Some examples of analysis performed on a real soccer game data are also provided in order to show the potential of the analytical tool. The resulting web framework is built using several technologies, such as Python, HTML, CSS and d3.js, the JavaScript library which is nowadays the standard for web-based data visualization projects.