Wordle-Solver – solves 99.7% of common five-letter English words | Hacker News Reader

Wordle-Solver – solves 99.7% of common five-letter English words

github.com

39 pointsjasonchao26 comments

Active related discussion: https://news.ycombinator.com/item?id=30050231

I've settled on the following word list to determine if 20 of 26 characters are present within 4 guesses:

  thank
  fuels
  crowd
  gimpy

If there are fewer than 5 letters present, "bevvy" will eliminate another two ('b' and 'v'), leaving j, q, x, and z as the remaining possible characters. These occur rarely (though "pizza", "bijou", and "vixen" can be challenging targets. More usually, it's words with either doubled characters, such as "onion" or "lalai", or those which have viable anagrams ("spams" and "spasm", "donor" and "rondo") that will throw me.

Another set I use is "slant", "prick", and "budge", with coices from "comfy", "whiff", and "vixen" depending on what the initial results were. I'm finding this less successful than my first set.

And for those who enjoy the hunt, yes, this does make the challenge quite mechanical.

This is (in almost all cases) ignoring the clues, I.e. non-hard-mode. I find the hard mode where every clue must be used to be far more interesting for exactly this reason. It makes it impossible to use canned guesses from the second word onward. You need to dig deep in your dictionary.

Even better to try to also use all negative clues and not re-use any gray letter either. This can’t be enforced with a setting however.

It's less ignoring clues than maximising the delivery of clues.

Hard mode is hard because repeating a known value reduces the values which can be tested.

If that's how you choose to play, then yes, go ahead.

But after applying the wordlist(s) I use above, what you have are clues to the hidden word, and that must still be guessed. It simply turns out that the letter space itself can be extensively tested.

Hard mode is interesting, but I think it's not better just because you can't use canned guesses. Canned guesses is a valid strategy (and I don't feel it's cheaty, as another commenter mentioned, you still need to guess the word even if you have 4 yellow letters).

In hard mode, you can be unlucky and had to guess based on luck. For example, if your first guess is POUND, and the hidden word is MOUND, you have no choice to simply guess one by one from the list BOUND, FOUND, HOUND, MOUND, ROUND, SOUND, WOUND. Which is not interesting (this is to showcase an extremely unlucky example in hard mode, not saying that hard mode is bad. Instead, this shows that you need to be more thoughtful of your guesses. I would say you would try to avoid the correct letters first, so you can eliminate as many letters as possible. Which, of course, makes it harder, and so it's called hard mode)

Also, from the original Mastermind game, guessing inconsistently with previous guesses is one of the best strategies too, for the same reason: allowing to remove more possibilities in one guess.

Which suggests that for Hard Mode, your best bet is to seek out less likely terms first.

The question is whether or not the goal is to achieve a very low tries score, or if you want to solve most or all challenges.

The canned-guess easy-mode strategy optimises for the latter.

Yep, that's what I mentioned too. Although too be strict, the word you have to guess to minimize the number of tries is the one that minimizes the worst case scenario. And, if a computer is doing this, it is achieved by considering all possible solutions, then try each reasonable guess for each possible solution to determine which guess would lead to the least number of try.

Now, a human cannot consider all possible solutions (especially in the first 2 guesses), so human strategy will be different from computer strategy. As we agreed, it is to guess the least likely word. And this canned guesses turns out to do just that as well. Here I assume some flexibility when using the canned list, that is, (1) the order is flexible, and (2) if you've found some confirmed letters already, you decide which of the canned word list would be most useful next.

For example, for me I use my canned guesses for the first two guesses, and decide the third based on the first two. If I still need to get more letters, I go through my list. If I think I got enough to start going towards the solution, then I do that. It is still an interesting game since there are still decisions to be made.

(Also, I think hard mode is not even solved yet by a computer, it might be possible that there is no strategy where all possible solutions can be guessed in 6 tries. This increased risk might be something that people find interesting too, so that's one of the appeal of hard mode).

(Also, as I said, this is not a critic about hard mode not being interesting. It is interesting. My point is to simply defend easy mode from the claims that it is not interesting [from what "hard mode is far more interesting" seems to suggest], and nothing personal. I appreciate hard mode too for its interestingness, and the people who like it better over easy mode. I'm not against people claiming that hard mode is more interesting than easy mode)

(Oh, I just realized you are the one posting on the canned guess list. I'm also using canned guess list. My list is AISLE, PHONY, TRUCK, and the fourth guess onwards I usually try to guess based on the clues so far)

If your goal is actually solving the current popular Wordle implementation, its complete word list (allowed words and words that might actually be chosen) can be retrieved from the game source.

You can get them both here:

https://gist.github.com/cfreshman/a03ef2cba789d8cf00c08f767e... https://gist.github.com/cfreshman/cdcdf777450c5b5301e439061d...

And you can solve 100% of them within five guesses.

Well you can solve 100% of them in one guess if you solve it in another browser first.

No - without knowing the answer in advance, there is no answer which requires more than five guesses.

I have a web UI that does the exact same thing. https://veerasundar.com/wordle - you input Wordle results and it suggest words. It sourced the list of words from the Wordle itself.

I made something very similar. https://cjboco.github.io/wordfinder/ - except mine you use normal keyboard to create your word and the on screen keyboard to exclude the characters. Think I prefer your way.

Do we have to take out the magic in everything?

that would be `JSON.parse(localStorage.getItem("gameState")).solution` ;)

I still play chess, daily.

I look forward to people playing with proof assistants as just a hard video game.

I suspect, for many of us here, writing solvers for games is more magical than playing them.

And trainers / save file hacking / locating ram addresses / editing config files / modding.

There is no magic in small puzzles. There is only truth, begging to be brought to light.

Fight your wordle bots at botfights ... current leader averages 4.021 guesses per word.

https://botfights.io/game/wordle

I wrote a similar thing, expect it's web-based: http://wordle.imedadel.com/

The list isn't limited to that of Wordle, so I guess I might update that.

I find these tools quite interesting and helpful especially since English is neither my first or second language.

I wrote a tool that can predict today's and tomorrow's answer without you having to input anything :p https://wordle-answer.vercel.app

What mechanic allows you to guess tomorrows answer?

Inspection of the game source code

I guess I was kind of asking for this, but I just spoiled the next two days of Wordle for myself clicking on your link. Oh, the woes of being a trigger happy link follower

Reply on news.ycombinator.com