Wordle Solver
A simple solver for Wordle puzzles that uses letter- and word-frequencies to narrow down possible guesses.
Approach
Wordle maintains two word lists: a list of 2,315 solution words and 10,657 possible words. The union of these two lists forms the set of all valid guesses; other 5-letter words will be rejected from the game. This combined set forms the initial universe of possible words. The solver narrows this set until a solution is found.
For each guess, the solver first calculates the letter frequencies of the possible words. The candidates for the next guess begin as the set of words that contain the most frequent letter. From there, this set is narrowed by iteratively performing an intersection of the set with the next-most-frequent letter. The intersection operations continue until either the set has fewer than 10 guesses or the list of letters is exhausted.
When comparing the solution list to the possible word list, it appears that the words in the solution list are more common than the possible list, with the possible list containing some proper nouns and archaic words. To account for this, the solver selects a guess from the candidate list by finding the word that appears most frequently in the American National Corpus (ONC).
If the guess is incorrect, the solver will ask for the green and yellow letters. The solver will use this information to filter the universe of possible words. The words must contain any yellow letters, not contain yellow letters in positions that have already been considered, contain green letters in the correct spots, and not include any previous incorrect letters.
Once the universe is narrowed, letter frequencies are re-calculated and guessing repeats.
Running the solver
The solver requires a recent Go install (>1.11).
In the root directly, simply run go run cmd/main.go.
The solver assumes two files exist in the current directory: 1) words.txt, which contain the possible words to guess and 2) counts.txt which contains space-separated word frequencies.
