How many Wordle words will Wordle guessers guess if Wordle's Wednesday Wordle word is "Eerie"?

Abstract

With the rising popularity of Wordle, people have eagerly taken to Twitter to report their results daily by the tens of thousands. In this paper, we develop a comprehensive model which uses this data to predict Wordle player performance and reporting given any word and date. To do so, we first decompose words into quantifiable traits associated with relevant difficulty characteristics. Most notably, we formulate a novel Wordle-specific entropy measure which quantifies the average amount of information revealed by typical players after initial guesses. We also develop a method to represent the distribution of player attempts, and hence the observed difficulty of a word, using just two values corresponding to the cumulative mass function of the Beta distribution. Consequently, we are able to use a preliminary Lasso regression to isolate the most relevant predictors of word difficulty, which we then use in a Bayesian model. For a given word and date, our Bayesian model predicts the distribution of the number of guesses by players (i.e. the reported player performance), the number of player reports, and the number of players reporting playing in hard-mode. To accomplish these three tasks, it is made up of three submodels which are conditionally independent given the data, making it efficient to sample from its posterior using Markov Chain Monte Carlo (MCMC). Our model is able to predict outcomes for new data and retrodict for old data, and we empirically demonstrate the success of our model through retrodictions on unseen data. Most notably, our model does not just provide such simple point estimates and prediction intervals, but full posterior distributions.