Optimal Sets of Questions for Twenty Questions

Abstract

SIAM Journal on Discrete Mathematics, Volume 38, Issue 1, Page 412-452, March 2024.
Abstract. In the distributional Twenty Questions game, Bob chooses a number [math] from 1 to [math] according to a distribution [math], and Alice (who knows [math]) attempts to identify [math] using yes/no questions, which Bob answers truthfully. Her goal is to minimize the expected number of questions. The optimal strategy for the Twenty Questions game corresponds to a Huffman code for [math], yet this strategy could potentially uses all [math] possible questions. Dagan et al. constructed a set of [math] questions which suffice to construct an optimal strategy for all [math], and showed that this number is optimal (up to subexponential factors) for infinitely many [math]. We determine the optimal size of such a set of questions for all [math] (up to subexponential factors), answering an open question of Dagan et al. In addition, we generalize the results of Dagan et al. to the [math]-ary setting, obtaining similar results with 1.25 replaced by [math].