The problem of selecting m best players out of n candidates, through pairwise comparisons, is considered. Deviating from the standard models, it is assumed in this article that the outcome of a pairwise comparison (e.g., a match between two candidates) may be manipulated by collaborating participants: The stronger party may intentionally lose to the weaker party in order to gain group benefit. We discuss protocol design issues for such scenarios, and develop both possibility and impossibility results.
