This document discusses answering why-not questions for top-k queries. It begins with background on why-not questions and the history of explaining missing answers to queries. It then introduces the problem of answering why-not questions for top-k queries by explaining how to refine a top-k query to include a missing tuple in the results. The authors propose a sampling-based approach to find an optimal refined query with the minimum penalty by sampling weightings and stopping progress operations early. They present experiments on NBA data and synthetic data to evaluate their approach and optimization techniques.