To examine test-retest reliability of two-dimensional measured frontal and sagittal plane kinematics during running, and to determine how many steps to include to reach and maintain a stable mean.Reliability study.Research laboratory.Twenty-one recreational runners.Lateral trunk position, contralateral pelvic drop, femoral adduction, hip adduction, knee flexion and ankle dorsiflexion during midstance, and foot and tibia inclination at initial contact were measured with two-dimensional video analysis during running for 10 consecutive steps for both legs. All participants were tested twice one week apart. A sequential estimation method was used to determine the number of steps needed to reach a stable mean. Intraclass correlation coefficients (ICC) and smallest detectable differences (SDD) were calculated.The minimal number of steps was 6.3 ± 0.3. Lateral trunk position, femoral adduction and foot inclination showed excellent reliability (ICC 0.90-0.99; SDD 1.3°-2.3°). Tibia inclination and ankle dorsiflexion showed good to excellent reliability (ICC 0.73-0.92; SDD 2.2°-4.8°). Hip adduction and knee flexion showed good reliability (ICC 0.82-0.89; SDD 2.3°-3.8°). Contralateral pelvic drop showed moderate to good reliability (ICC 0.59-0.77; SDD 2.7°-2.8°).Two-dimensional video analysis is reliable to assess running kinematics on different days. The mean of at least 7 steps should be included.