The automatic observation of team-sport games such as soccer is a valuable application domain for computer vision research. A practically applicable system in this domain has to reliably solve problems such as initialization and tracking of camera models, real-time object recognition and multiple object tracking through occlusions. In the project, we developed the software system ASpoGAMo, a vision based system for extracting motion trajectories of all visible players from standard broadcast recordings of soccer games. It is also capable of infering the position of the ball. ASp0GAM0 solves the problem by splitting it into subproblems that interact in subtle ways: at first, the continuous estimation of camera direction and zoom factor is solved despite fast camera movements and almost homogeneously textured field background using robust optimization techniques. Then, visible players on the field plane are detected using model-based image cues such as appearance and expected size given the estimated extrinsic camera parameters. In the final step, estimated player positions are tracked and smoothed using a Rao-Blackwellized Resampling particle filter, which is capable of disambiguating tracked players after occlusions as well as generating predictions for future player recognition steps. Knowledge of player positions is also used as a key cue for minimizing false detections when tracking the ball during game-play. In contrast to the traditional method of using human scouts, the method is more objective and can take into account huge amount of data which cannot be handled by humans. Sports game analysis models as provided by ASpoGAMo open new and unique paths in both the sports science and its practical use. In this project we validated and analyzed the ASPOGAMO model together with the ASPOGAMO observation and interpretation system in the context of the German women football league champions FFC Turbine Potsdam, the professional team of FC Bayern Munich and FC Augsburg (2nd Bundesliga) as well as the recent World Cup games. The analysis consists of motion profiles, stress times, action profiles, tactic analysis and the evaluation of strengths and weaknesses of own and opponent team and in particular analysis of incorrect fitting.