A study on predicting the results of Wordle guessing game based on time series and linear regression

Abstract

Wordle is a globally popular guessing game in which players guess a five-letter word in six or fewer attempts. To explain the changes in the number of players on a daily basis and to make predictions about the number of users playing on a future date, this paper uses the contagion model as the basis for an analogy to explain the reasons for the changes in the data. Since the contagion model is not very accurate in predicting the future, considering that the data itself is a set of time series data, this paper uses a time series prediction model to predict the future data and gets a good fit. In order to predict the distribution column of the number of guesses for a word, this paper performs regression fitting of word attributes and data distribution based on different eigenvalues of different words to get the change curve of eigenvalues, and considers the uncertainty of the model based on the obtained functional relationship, and then the model obtained by replacing the words to be predicted results is tested and error analyzed.