Prediction of Scrabble Game Results Based on Gradient Boosting Decision Tree and time Series Model

An exciting puzzle game, wordle, is offered in the New York Times, in which players need to guess a five-letter word in no more than. To explain the daily number of reported result, this article established ARIMA time series model and obtained the smoothness test, model determination and fixed-order correlation data. After that, this article predicted the number of reported results on March 1, 2023, with the prediction interval [10268.457, 10644.897]. Meanwhile, this article selected four-word attributes and judged their relationship and the difficulty mode score by the Pearson correlation. Finally, this article built a GBDT model based on four-word attributes to predict the distribution of the reported results.