A Revised Entropy Based Algorithm to Solve Wordle
Yuqi Xia, Lehao Pan, Xinyan Lin · 2023
The previous entropy based algorithm for Wordle has a weak performance when encountering words with a common suffix. In order to solve this issue, we establish a revised entropy based system, adding a revised term to the entropy formula to amend the influence of suffix on the expected pattern distribution. We also consider the influence of KL divergence on the probability that a common suffix may result in a poor performance. The results show a 1.5% overall improvement with average attempts decreasing from 3.87 to 3.81. We also test our system's sensitivity on starting words and "salet" turns out to be the best starting word that suits our revised system.