Assess whether stronger performance guarantees are achievable under a more realistic adversary

Ascertain whether price-update algorithms for multidimensional blockchain fee setting can achieve stronger performance guarantees—such as improved average-regret rates or tighter bounds—under a more realistic adversary model than an all-powerful adversary.

Background

Using an all-powerful adversary, the paper proves O(1/sqrt(T)) average regret and provides a matching lower bound, showing essential optimality in that model. The authors then pose whether a more realistic adversary could allow improved results.

This motivates investigating refined adversary models and corresponding analyses to determine if better rates or bounds than those established under the extreme adversarial model are attainable.