So, there's another evil Wordle thread on Hacker News today (https://news.ycombinator.com/item?id=29862597) and over there, I've been refining my algorith for optimal play. Here's the current approach:
R A I S E (0 green, 0 yellow, 168 words remain)
B L U D Y (1 green, 0 yellow, 13 words remain)
C O U N T (2 green, 1 yellow, 2 words remain)
V O U C H (4 green, 0 yellow, 1 word remains)
P O U C H (5 green, 0 yellow, 0 words remain)
But what was interesting was at one point of development, it did this:
L A R E S (0 green, 0 yellow, 576 words remain)
T O N I C (1 green, 0 yellow, 50 words remain)
B O O D Y (4 green, 0 yellow, 5 words remain)
D E G U M (0 green, 2 yellow, 1 word remains)
G O O D Y (5 green, 0 yellow, 0 words remain)
See, the "BOODY" has 4 greens, but then it went and guess "DEGUM" to eliminate a lot of possibilities!
One additional feature I was looking at for a personal implementation was picking the target word- how is it being chosen from the remaining letters after the player's first guess? Are you using a word that purposely uses fewer common letters from the set of remaining words?
No, I'm looping through all possible guess and targets and grouping possible targets by matching scoring. Then I select a guess on which one has the smallest set of possible words afterwards. For example, "SERAI" seems to be a good first word. The reason why is that even if the adversary selects the worst-case scoring of (in this case, it would be no green, no yellow), there's only 168 remaining words it could be.
