In material wear experiments, such as the Ball Mill Abrasion Test (BMAT), it is crucial to know the number of specimens to be used in each experiment to produce accurate and reliable results. In BMAT, however, a structured way of determining the necessary number of balls per type of material to be included in the experiments has not been considered to date. This article shows, using statistical tools and adhering to the standards that govern these experiments, that the optimal number of balls per type of material to be included in each experiment is 6. In addition, it is shown that if the initial differences between the balls are reduced, it is possible to obtain more accurate and reliable results.