Folks, I think this gradle algorithm is bad.
Without foreknowledge, how can an initial guess change in grade from one day to the next? Either it’s a good starting word, or it isn’t.
Yeah, Gradle’s grades are bogus. Gradle only compares how effective your guess was against the current puzzle solution, and that highly depends on luck. A starting word only scores an A+ if it narrows the today’s answer to 5 possibilities or less.
It should be grading you based on that word’s average performance against every puzzle solution you haven’t eliminated yet. The optimal play should always award an A+, and everything else should scale compared to how the optimal play did. Someday I’ll have to make a Gradle clone that does this.
If you pay a garbage starting word like MAMMA and it happens to win the puzzle in 1/6, it should still give you an F because it was a terrible idea.
A good grade is obtained by reducing the list of possible words as much as possible with each guess.
The best grades are going to depend on the answer that day. Your favorite starting word or words may not always be the best guess if they don’t reduce the possible list of words very much.
In my opinion, the first guess word should have a fixed grade against all of the words possible.
Then subsequent guesses would evaluate how much closer you got, ideally following hard mode rules
Folks, I think this gradle algorithm is bad. Without foreknowledge, how can an initial guess change in grade from one day to the next? Either it’s a good starting word, or it isn’t.
#Wordle1355 5/6* Grade: B+
🟩⬜⬜⬜⬜ D
🟩⬜🟨⬜⬜ D+
🟩⬜🟩🟩⬜ A
🟩🟩🟩🟩⬜ A+
🟩🟩🟩🟩🟩 A+
https://gradle.app/#4yCdUKmDWqE4QL6NvfjDY6fXm
Yeah, Gradle’s grades are bogus. Gradle only compares how effective your guess was against the current puzzle solution, and that highly depends on luck. A starting word only scores an A+ if it narrows the today’s answer to 5 possibilities or less.
It should be grading you based on that word’s average performance against every puzzle solution you haven’t eliminated yet. The optimal play should always award an A+, and everything else should scale compared to how the optimal play did. Someday I’ll have to make a Gradle clone that does this.
If you pay a garbage starting word like MAMMA and it happens to win the puzzle in 1/6, it should still give you an F because it was a terrible idea.
I think your second para nails the use case.
Of the first guesses how did I do? Of the possible second guesses based on hard mode rules and the feedback how did I do?
I could guess it in 2, but get a poor grade because it thinks my first guess sucks.
that’s super interesting so I did some research.
from the Gradle help:
In my opinion, the first guess word should have a fixed grade against all of the words possible. Then subsequent guesses would evaluate how much closer you got, ideally following hard mode rules