It is challenging to design general robot soccer coordination behaviors that address individual states. We have successfully followed a case-based approach to define behaviors for a single soccer robot. In our multi-robot system we now distinguish retriever robots that access the case library, reason about the situation, and select the most appropriate cases. They communicate with the other robots and they all execute the retrieved case in a coordinated way. We evaluate our approach with two robots demonstrating that the robots successfully coordinate and the number of passes during a game highly increases compared to an approach with an implicit coordination mechanism.