[2207.00076] Efficient computation of rankings from pairwise comparisons

We study the ranking of individuals, teams, or objects, based on pairwise comparisons between them, using the Bradley-Terry model. Estimates of rankings within this model are commonly made using a simple iterative algo…

Efficient computation of rankings from pairwise comparisons \name M. E. J. Newman \addr Center for the Study of Complex Systems University of Michigan Ann Arbor, MI 48109, USA Abstract We study the ranking of individuals, teams, or objects, based on pairwise comparisons between them, using the Bradley-Terry model. Estimates of rankings within this model are commonly made using a simple iterative algorithm first introduced by Zermelo almost a century ago. Here we describe an alternative and similarly simple iteration that provably returns identical results but does so much faster—over a hundred

related reading

• Bradley–Terry model - Wikipediaen.wikipedia.org
• Why Momentum Really Worksdistill.pub
• Whole-History Rating: A Bayesian Rating System for Players of Time-Varying Strengthremi-coulom.fr
• Elo rating system - Wikipediaen.wikipedia.org
• Energy-Based Preference Model Offers Better Offline Alignment than the Bradley-Terry Preference Modelarxiv.org
• [1809.09087] Implicit Maximum Likelihood Estimationarxiv.org
• Prompt-to-Leaderboardarxiv.org
• [2503.10990] Statistical Impossibility and Possibility of Aligning LLMs with Human Preferences: From Condorcet Paradox to Nash Equilibriumarxiv.org
• Distortion of AI Alignment: Does Preference Optimization Optimize for Preferences?arxiv.org
• 2012.00174arxiv.org
• [arxiv] an algorithmic framework for fairness elicitationarxiv.org
• Consensus Multiplicative Weights Update: Learningto Learn using Projector-based Game Signaturesarxiv.org