This paper presents an algorithm that was developed to identify statistical outliers from the scores of grading jury members in a large project-based first year design course. The background and requirements for the outlier detection system are presented. The outlier detection algorithm and the follow-up procedures for score validation and appeals are described in detail. Finally, the impact of various elements of the outlier detection algorithm, their interactions, and the sensitivity of their numerical values are investigated. It is shown that the difference in the mean score produced by a grading jury before and after a suspected outlier is removed from the mean is the single most effective criterion for identifying potential outliers but that all of the criteria included in the algorithm have an effect on the outlier detection process.