The application of low-cost L1 GPS receivers integrated with micro-electro-mechanical system (MEMS) inertial measurement units (IMU) allows the continuous observation of position, velocity and orientation which opens new possibilities for comparison orf athletes' performance throughout a racecourse. In this paper, we compare loosely and closely coupled integration strategies under realistic racing scenarios when GPS is partially or completely masked. The study reveals that both integration approaches have a similar performance when the satellite constellation is completed or the outages are short. However, for less than four satellites, the closely coupled strategy clearly outperforms the loosely coupled approach. The second part of the paper is devoted to the important problem of system initialization, because the conventional GPS/IMU alignment methods are no longer applicable when using MEMS-IMU. We introduce a modified coarse alignment method and a quaternion estimation method for the computation of the initial orientation. Simulations and practical experiments reveal that both methods are numerically stable for any initial orientation of the sensors with the error characteristics of MEMS-IMU's. Throughout the paper, our findings are supported by racing experiments with references provided in both, the measurement and the navigation domain.