Efficient Computation of Rankings from Pairwise Comparisons
M. E. J. Newman; 24(238):1−25, 2023.
Abstract
This study focuses on the ranking of individuals, teams, or objects, based on pairwise comparisons between them, using the Bradley-Terry model. The estimation of rankings within this model typically involves an iterative algorithm first introduced by Zermelo almost a century ago. However, we present an alternative and similarly simple iteration that yields identical results but is significantly faster—over a hundred times faster in some cases. We demonstrate the effectiveness of this algorithm through various example data sets and provide insights into its convergence.
[abs]