You would need a significant sample size to make any sort of conclusion from such a probabilistic process. Then there's the issue of how you would actually grade/compare.