Information Theory-based Wordle Game Word Difficulty Classification and Dynamic Planning Optimization Research

Abstract

Wordle has gained popularity as a word-guessing game. This study aims to describe the difficulty attributes of target words in Wordle and categorize them based on their difficulty. After two rounds of word guessing, this paper calculated the average mutual information for the third-round correct answers. This data was utilized to build a K-means classification model, categorizing words into three distinct groups: easy, normal, and hard. For instance, the word "EERIE" has an average mutual information of 8.997 bits, categorizing it as a word of 'hard' difficulty. Given the vast number of words to process, the computation time was extensive. To address this, this paper employed dynamic programming, leading to a significant reduction in operation time. Additionally, a Monte Carlo simulation model was established to simulate potential player guessing patterns, validating the classification model's robustness. The model developed in this research offers fresh perspectives on strategy selection in Wordle games and the difficulty assessment of target words. It serves as a dependable guide for game developers when classifying word difficulty.