The Lifecycle Change Model based on Wordle analysis and study

Abstract

Wordle is a popular game, and I did a detailed analysis of the Wordle dataset provided by the New York Times to better understand the game. First, we proposed a lifetime change model to describe the number of reported results over time, and through analysis we found that the property of "whether a word contains repeated letters" affects the fraction percentage of the difficult pattern. We then constructed a GloVe-LBP neural network to predict the reported percentage of a given solution and performed a detailed analysis of the model confidence, loss function, training adequacy, and generalization ability. Next, we constructed a K-Means and z-score classification model to classify the difficulty of words, through which the "EERIE" difficulty level was level 5. The relationship between the number of reported results and the number of difficult patterns in the data set was concluded and analyzed.