We propose the design of an observatory dedicated to the tactical analysis of handball, involving the use of dynamic graph to represent and study patterns of play. This innovative methodology was used to capture the complex interactions between players, their movements on the court and passing sequences. By using techniques for extracting frequent subsequences from a dynamic graph constructed from a gnomonic transformation (azimuthal map projection that transforms circles into straight lines.) of the court, it was possible to identify the most recurrent and therefore significant patterns of play. The extracted sequences were valued by various measures of interest: emergence rate and expected goal. These measures of interest introduced a hierarchy between the sequences, enabling them to be navigated. Within the observatory, the tactics discovered were illustrated by an intuitive graphical representation, giving coaches and analysts a better understanding of the trends, strengths and weaknesses of the tactics employed.