Ranking and selection for pairwise comparison
针对只能通过成对比较来评估设计方案的问题，提出了一种基于最优计算预算分配的统计排序与选择方法，并通过数值实验验证了其能有效减少仿真次数。
Abstract In many real‐world applications, designs can only be evaluated pairwise, relative to each other. Nevertheless, in the simulation literature, almost all the ranking and selection procedures are developed based on the individual performances of each design. This research considers the statistical ranking and selection problem when the design performance can only be simulated pairwise. We formulate this new problem using the optimal computing budget allocation approach and derive the asymptotic optimality condition based on some approximations. The numerical study indicates that our approach can reduce the number of simulations required to confidently identify the best design.