I made one, not the optimal strategy though, I simply minimized the expected number of possible solutions for the next step. I repurposed my Mastermind solver since it's basically the same game. Just like Mastermind, minimax is probably better than what I did, but it does solve every Wordle in at most 5 guesses, on average 3.4 guesses.
Surely it differs from Mastermind (which I might not be remembering well as I didn't like it) in that the answers must be words in the game's dictionary rather than arbitrary colour codes? That seems to change tactics considerably.
I agree it changes the experience of playing the game for a human quite a bit, and you'll use different tactics, but for the kind of solver I made it doesn't change much - I had to change very little code to make it work.
From the perspective of the solver it just means the starting set of potential solutions is every word in the dictionary, instead of every permutation, but it doesn't really make much of a structural difference. Either way a solver can just brute-force its way through every potential solution since there aren't that many.
The other difference is that Mastermind tells you only how many of the pegs are correct, but not which ones. Wordle tells you which ones. That is easier, but it's counterbalanced by Wordle having a larger solution space. There's only 1296 possible solutions for classic Mastermind (4 pegs, 6 colors).
I think using the expected logarithm of the number of solutions (entropy) makes more sense from an information theory perspective. I wonder if it works better in practice.
Surely it differs from Mastermind (which I might not be remembering well as I didn't like it) in that the answers must be words in the game's dictionary rather than arbitrary colour codes? That seems to change tactics considerably.