Worst-case versus average-case design for estimation from partial pairwise comparisons (Q2196209)

The authors study the problem of estimating the comparison probabilities from noisy pairwise comparisons under average-case and worst-case design assumptions. They also exhibite a dichotomy between both designs for permutation-based models, and they conjecture that a similar distinction may exist even for their parametric counterparts. Other relevant results about the problem of estimating the comparison probabilities from noisy pairwise comparisons are presented in the paper. In addition, the article ends with a discussion of open problems in this area of research.