jonesnxt / kilordle

24 stars 20 forks source link

Reduce number of guesses based on "autocomplete" mechanic #13

Open sehrgut opened 2 years ago

sehrgut commented 2 years ago

Per #11 it appears the autocomplete mechanic is intentional to avoid having to skim columns of letters that are effectively solved when all positions have been guessed for a given word. However, this results in the game being winnable in significantly fewer (10-15%) of the 1005 available guesses.

I'm not sure exactly what the best way to determine this is, but either a fixed number of guesses based on statistically how many are needed for the average game, or even "5 guesses more than the minimum" (which appears what the 1005 was intended to be) for the SPECIFIC puzzle generated in a session, might be useful.

martinkozle commented 2 years ago

Or you could just leave it to be an unbounded number of guesses and have a leaderboard for the fewest guesses for example. It would be pretty annoying to be near the end and for the game to end because you failed to solve it in the minimum number of guesses. The game isn't really replayable enough for someone to retry it if they fail.

sehrgut commented 2 years ago

Oh, I like this idea! It also helps build towards an empirical difficulty target, instead of trying yet another model that's possibly flawed like the n+5 model is.

JAicewizard commented 2 years ago

You could also remove guesses based on auto-completed words, so if you complete 5 words 5 guesses are removed

JohnyDL commented 1 year ago

I can guarantee a full solution with 37 words (almost certainly only a handful of them are actual words in the grid), solving less than that right now seems unlikely but might be doable. A leaderboard would simply consist of the people who've done the math/written a program with the game in it's current format.

However I can see potential solutions to this if the autocomplete mechanism is limited, if the autocomplete solutions are counted to the solution it actually makes the game more challenging in line with the n+5 model after all the autocomplete is just following the instructions it's fun probably why it exists in the first place.

You could then have a couple of different game 'modes' or potential ways to score: Wordle Original is the lowest number of words (including auto-completes) should actually be possible to get 1000s on that score with getting lucky on the initial words Kilordle Classic is the lowest number of words input (as it exists right now with 37 being the 'optimal' but likely hitting 1030 on the wordle classic score mode) Dual Scoring Your total score is actually both the Kilordle Classic and Wordle Original Scores added together, the aim being you want to minimize the number of words you put in for a balance of both scores. You want to auto-score as many words as possible by adding the fewest words manually, and only adding words manually if they're correct answers.

With the green and yellows auto-completes: 4 greens + 1 yellow 3 greens + 2 yellows (one open location can't be one yellow) 2 greens + 3 yellows (one open location can't be two yellows another open location can't be one of the same yellows) 1 green + 4 yellows (ad nauseam) 5 yellows (ad nauseam)

There could in effect 7 play modes, no auto completes, green auto completes, and 5 levels of yellow autocompletes, (Doesn't solve the double letter answers by default but for everything else that's just about paying attention to the clues on screen it could) each with 3 different scores to master certainly makes the game replayable