Kinematic measurements made at the head centre of gravity (CG) are most often filtered using SAE J211 recommendations, but there is little guidance for measurements taken at locations other than the CG. This study introduces a new method for determining optimal kinematic filtering. We then use this method to establish optimal cutoff frequencies for lowpass filters applied to six-degree-of-freedom (6DOF) signals when measured external to the CG. We conducted laboratory head impact tests using a pendulum impactor striking a bare headform to emulate impact scenarios observed in on-field environments. Impact conditions included four impact energies, two impact durations, and four impact locations. We collected triaxial linear accelerations at the headform's CG and triaxial linear accelerations and angular velocities at the teeth. We compared measurements at the teeth to ground truth determined through rigid body dynamics equations of motion. Teeth measurements were filtered with over 100,000 combinations of linear acceleration and angular velocity cutoff frequencies. Mean bias and variance were both considered when quantifying error. A range of suitable filter combinations was identified to minimise bias and variance, with the recommended filter combination being 300 Hz for linear acceleration and 300 Hz for angular velocity (CFC 180).