He very specifically tests all his arrows for how much they flex, and if it's outside his tolerance, he doesn't use it (as explained in the video). In this way, he is able to normalize the deflection of the arrows he uses, and correctly predict the deflection offset.