Effective Wordle Heuristics --- Ronald I. Greenberg

This is a companion web site for the paper

Ronald I. Greenberg. Effective Wordle Heuristics.

Based on that work, a daily recommended Wordle strategy is provided. These daily recommended strategies take into account the past Wordle history and assume that solutions are not reused. These daily recommended strategies are provided as of the first public day of Wordle on June 19, 2021, but as if the current list of possible solutions was in effect from that time. This current solution list used by the New York Times official version of Wordle contains 3158 words. The strategies provided here make guesses guesses only from this list of 3158 words, but does not restrict to hard mode.

The strategies recommended here do not use information about the frequency of words in the New York Times or elsewhere, unlike the way that WordleBot works. (See Introducing Wordlebot and Introducing Wordlebot 2.0.) Nonetheless, it has been observed that the New York Times is rarely using words introduced when it expanded the list of possible solution words to 3158 words, so when all else is equal in heuristic evaluation, a guess is chosen from the 2302 older words neither added nor removed by the New York Times. (WordleBot does not use the history of past solutions, so users with a good memory or the tool here may be able to do better than WordleBot. This can be especially true if as you go through the recommended strategy you look at the end game situation and consider choosing a more common word from among those remaining as possible solutions.)

The C code used to create a new recommended Wordle strategy each day is also provided. (Additionally, shell scripts grab the new Wordle solution each day and invoke the C program to generate a new heuristic strategy.) Based on the New York Times release of each day's Wordle word early the prior morning (in US Central Daylight Savings Time), a recommended strategy is created within a couple hours of that time to guess the Wordle word two days hence.





Last updated on 26-Sep-24 at 11:24 CDT by Ronald I. Greenberg (rig at cs.luc.edu)