The document summarizes a car racing competition using the open-source racing simulator TORCS. It describes the goals of learning or designing controllers to race as fast as possible alone or against other drivers. It provides details on the TORCS simulator, the competition API and framework, the sensors and effectors available to controllers, submissions received, and results of evaluating the submissions in single-controller and multi-controller races on different tracks. The best performing controllers were able to learn to outperform programmed controllers, though some overfitting and issues with avoiding opponents and recovery were observed.