Wordle Game Result Prediction Based on XGBoost Algorithm

Wordle is a popular puzzle currently offered daily in the New York Times. The purpose of this paper is to develop a predictive model that predicts the percentage of different attempts made to solve the game. First, we calculated the five Spearman correlation coefficients representing the word attribute and the percentage of hard mode separately and found no significant relationship between the word attribute and the percentage of hard mode scores. To predict the percentage of different attempts in the results, we built a multimodal model. The Informer algorithm was used to analyze the time series, and the XGBoost algorithm was used to analyze the word attribute data, and then the model was improved using the attention mechanism and the sliding window mechanism. The final prediction of the percentage of attempts was made for the word “EERIE” on March 1, 2023.