A Wordle bot that plays using principles inspired by Approximate Bayesian Computation.
Essentially the bot chooses a word which is expected to maximally 'narrow down' a given prior/posterior density of words with each guess until a point where it is left with a single word of suitably high posterior density.
To use, call python wordle_abc.py. You'll also need to include a dictionary of new-line delimited
valid words in a file called vocab.txt, and a CSV of word frequencies (from which prior
distributions of words are derived) in word_frequencies.csv (first column word, second column
frequency, with a header included).
Copies of vocab.txt or word_frequencies.txt are not provided out of respect for underlying
licenses of sources from which I derived them for my personal use.
A decent dictionary is pretty important. Using /usr/share/dict/words didn't cut the mustard in my
testing.