I wrote a script to calculate the expected information gain for any possible guess. It has several caveats - it probably doesn't share exactly the same word li

I wrote a script to calculate the expected information gain for any possible guess. It has several caveats - it probably doesn't share exactly the same word list as the game, it assumes that all valid words are equally likely, etc. But apparently TARES is the most informative first guess in expectation. RATES is also very close, and feels a bit more plausible as possible answer. XYLYL is the worst.