The aim of the study was to quantify the repeatability of the displacement of selected body landmarks while performing two different basic karate attacks. Seven karateka (three men, four women) with different levels of training and knowledge of karate were filmed with an optoelectronic computerized instrument (sampling rate 100 Hz) that allows the three-dimensional reconstruction of the movements of selected body landmarks. Thirteen landmarks (head, hips, upper and lower limbs) were analyzed while performing 10 repetitions each of choku-tsuki (straight punch) and oi-tsuki (lunge punch). For each karateka and punch, the average time of execution was calculated, and the standard deviations of each of the three spatial coordinates x, y, z were computed for each landmark. A total standard deviation of the single karateka and punch was also calculated. For all karateka, the execution of oi-tsuki took longer than the execution of choku-tsuki. For both punches and almost all landmarks, the largest repeatability (smallest standard deviation) was found in the vertical direction, while the smallest was found in the anteroposterior direction (direction of movement). In all karateka, oi-tsuki had a total standard deviation about 3 to 6 times larger than that measured during the performance of choku-tsuki. On average, women had a larger repeatability than men. The method employed in the current study allowed the quantitative analysis of the repeatability of two basic attacks in Shotokan karate by using a landmark based approach. The method could offer valuable help to karateka during training indicating which parts of the body do not repeat a selected movement with sufficient accuracy, thus assisting in the achievement of the most correct body form.