Current driver assistance systems or forward-looking safety systems mainly address traffic scenarios with cars travelling in the same direction or being stopped. These scenarios are - considered from a dynamical point of view - comparatively easy to handle due to the limitations of the relevant scenario parameters (relative velocity, possible accelerations, …). In the future it will be necessary to address oncoming traffic scenarios as well. These oncoming scenarios are responsible for a lot of critical accidents and the potential benefit is very high if one is able to reduce the crash severity in these scenarios. The problem remains that these scenarios are highly dynamical and therefore difficult to evaluate and handle.The following questions are of interest: How should a system be designed to be able to handle these situations?What are the critical scenarios which define the performance of the whole system?What are the limitations which cannot be overcome?We show a method based on simulation and data mining techniques which is able to address these questions. A stochastic approach based on different use cases is used to generate the relevant data. The method described is generic and can also be used to evaluate, develop or improve other safety or driver assistance systems.