An Exploration of Wordle Game Data and Player Behavioural Habits Based on Time Series and GBDT Models

Abstract

The focus of the whole problem is to explore interesting characteristics of Wordle’ players’ behavior data over the year. In order to explain the daily variation in reported results and to explore whether lexical attributes affect the percentage of scores in the difficult mode, this paper introduced a time series model for specific predictions, finally coming to the conclusion that there is a significant correlation between tries in hard mode and commonness, while there is no significant correlation directly with the number of repeated letters. In addition, based on previous research, a model was developed to predict the distribution of reported results for games at future dates. This paper introduced the prediction model based on Gradient Boosted Tree (GBDT) and carried out specific training, where the training accuracy reached more than 95%. The models are conducive to understanding the development trend of wordle games and players’ game habits, and to providing better game suggestions for game developers.