Order Selection Prophet Inequality: From Threshold Optimization to Arrival Time Design

TL;DR

This paper improves the competitive ratio for the order selection prophet inequality from 0.669 to 0.725 by introducing a novel arrival time design framework, advancing the understanding of optimal order and arrival time strategies.

Contribution

It presents a new algorithmic framework translating order selection into arrival time design, achieving higher competitive ratios in prophet inequalities.

Findings

Achieved a 0.725 competitive ratio in the order selection prophet inequality.

Developed a continuous arrival time design approach simplifying the problem.

Obtained an optimal 0.745 competitive ratio in the i.i.d. model.

Abstract

In the classical prophet inequality, a gambler faces a sequence of items, whose values are drawn independently from known distributions. Upon the arrival of each item, its value is realized and the gambler either accepts it and the game ends, or irrevocably rejects it and continues to the next item. The goal is to maximize the value of the selected item and compete against the expected maximum value of all items. A tight competitive ratio of $\frac{1}{2}$ is established in the classical setting and various relaxations have been proposed to surpass the barrier, including the i.i.d. model, the order selection model, and the random order model. In this paper, we advance the study of the order selection prophet inequality, in which the gambler is given the extra power for selecting the arrival order of the items. Our main result is a $0.725$-competitive algorithm, that substantially improves the state-of-the-art $0.669$ ratio by Correa, Saona and Ziliotto~(Math. Program. 2021), achieved in the harder random order model. Recently, Agrawal, Sethuraman and Zhang~(EC 2021) proved that the task of selecting the optimal order is NP-hard. Despite this fact, we introduce a novel algorithm design framework that translates the discrete order selection problem into a continuous arrival time design problem. From this perspective, we can focus on the arrival time design without worrying about the threshold optimization afterwards. As a side result, we achieve the optimal $0.745$ competitive ratio by applying our algorithm to the i.i.d. model.