Rovers estimate their position (“positioning”) and map their surroundings, but these tasks are complicated by measurement errors, which accumulate. This effect is severe in the vertical channel, as drift in the rover's elevation estimate is particularly difficult to mitigate. One of the main contributors to this drift is pitch misalignment between the rover's velocity vector and its Inertial Measurement Unit (IMU). This motivates our contribution: a method for calibrating an IMU's pitch misalignment. Our method does not require an initial ground calibration, external infrastructure, a source of ground truth pose, accounting for wheel slip, or sensors beyond an IMU and wheel encoders. We simply require the rover to drive in loops on approximately planar (but not necessarily level) terrain. This can be done autonomously in a planetary environment, requiring little oversight from mission operations. Our theoretical analysis, simulation results, and physical tests prove our method's effectiveness. Additionally, we analyze the effect of terrain, IMU quality, and calibration path on the calibration. Though our idea to drive in loops is simple, our main contributions are that we prove why our method works, analyze it extensively in theory, simulation, and practice, determine what parameters do and do not matter, and provide a comprehensive argument for applying our method to any wheeled rover.