Minimax Hypothesis Testing for the Bradley-Terry-Luce Model

The Bradley-Terry-Luce (BTL) model is one of the most widely used models for ranking a collection of items or agents based on pairwise comparisons among them. Specifically, given n agents, the BTL model endows each agent i with a latent skill score α_i > 0 and posits that the probability that agent i is preferred over agent j in a comparison is α_i/(α_i + α_j). In this work, our objective is to formulate a hypothesis test that determines whether a given pairwise comparison dataset, with k comparisons per pair of agents, originates from an underlying BTL model. We formalize this testing problem in the minimax sense and define the critical threshold of the problem. We then establish upper bounds on the critical threshold for general observation graphs and highlight their dependence on two fundamental structural properties: principal ratio and edge expansion. Additionally, we derive lower bounds on the critical threshold for the special case of complete induced graphs, thereby demonstrating that the critical threshold scales as Θ((nk)^−1/2) in a minimax sense in this case. In particular, our test statistic for the upper bounds is based on a new approximation we derive for the separation distance between general pairwise comparison models and the class of BTL models. To further assess the performance of our statistical test, we prove upper bounds on conditional probabilities of error. Additionally, we derive several other auxiliary results over the course of our analysis, such as bounds on principal ratios of graphs, ℓ²-bounds on BTL parameter estimation under model mismatch, stability of rankings under the BTL model with small model mismatch, etc. Finally, we conduct several experiments on synthetic and real-world datasets to validate some of our theoretical results. Moreover, we also propose an approach based on permutation testing to determine the threshold of our test in a data-driven manner in these experiments.