In-Depth Analysis and Forecasting of Wordle Result Based on Prophet Time Sequence and CART

Wordle, a word guessing game created by Josh Wardle in 2021, challenges players to guess a five-letter word within six attempts, with color cues indicating the correctness of letters. This paper investigates the relationship between reported results and word attributes using a predictive model. A Prophet model is applied to account for non-linear trends in result data, outperforming traditional models like ARIMA in long-term forecasting. By using confidence intervals, forecasting accuracy improves. Correlation analysis identifies key word attributes affecting user guesses. A higher-order linear regression model is used to predict the guess percentage based on previous attempts, while a Random Forest classification model enhances first-guess prediction accuracy. Additionally, a word difficulty assessment model is developed, clustering user data into categories for more accurate evaluations based on word attributes. By analysing Wordle user data and constructing models, we found that Wordle's popularity is gradually decreasing. Meanwhile, players' guessing behaviours are significantly affected by word attributes. Prophet Model performs well in long time prediction, but the prediction accuracy for some specific conditions needs to be further verified. The multivariate linear regression model with polynomial extension performs well with a small amount of data, but more practical factors still need to be considered in practical applications. Overall, our study provides new perspectives and methods for understanding player behaviour and word difficulty in Wordle games.