Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: A Wordle for the command line (Spanish and English)

github.com

Post reply on HN
Print view View original

11–18 of 18 posts

Re: Show HN: A Wordle for the command line (Spanish and English)

#11

Post by gpvos »

bryceneal wrote: post #2

Neat! One thing that has interested me about this game is whether or not there is an optimal first word. Is there is a word that contains the 5 letters which most commonly appear in the language? Preferably where more of the letters appear at their most common index? If so it seems one should use this word as the first guess each time given we have no other information other than the length of the word.

I usually keep ETAOIN SHRDLU in mind. ATONE would be a good one. It shouldn't really be too hard to exhaustively calculate an optimal strategy for the complete list of five-letter words; my /usr/share/dict/words has only 6806 of them.

Edit: for all the five-letter words in the above list, the order of frequency is ESAROILTNDUCYMPHBGKFWVZJXQ. Wordle has a different list.

Re: Show HN: A Wordle for the command line (Spanish and English)

#12

Post by gullywhumper »

japhib wrote: post #5

Cool! I’d recommend tweaking the dictionary a bit though … here are the first 5 words in the English one: aahed aalii aargh abaca abaci I don’t think I would ever think of putting in any of those words, so if that was the solution, I’d just be stuck. (Are they actually real words?)

The list of acceptable wordle words is available. I'm sure someone else has made the effort to make it a little more accessible.

view-source:https://www.powerlanguage.co.uk/wordle/main.e65ce0a5.js

Re: Show HN: A Wordle for the command line (Spanish and English)

#13

Post by banana_giraffe »

bryceneal wrote: post #2

Neat! One thing that has interested me about this game is whether or not there is an optimal first word. Is there is a word that contains the 5 letters which most commonly appear in the language? Preferably where more of the letters appear at their most common index? If so it seems one should use this word as the first guess each time given we have no other information other than the length of the word.

oater, orate, or roate are probably the optimal first selections. Given Wordle's word list, they let you make the largest prunes the list of possible answers.

Re: Show HN: A Wordle for the command line (Spanish and English)

#14

Post by powersnail »

japhib wrote: post #5

Cool! I’d recommend tweaking the dictionary a bit though … here are the first 5 words in the English one: aahed aalii aargh abaca abaci I don’t think I would ever think of putting in any of those words, so if that was the solution, I’d just be stuck. (Are they actually real words?)

Apparently, aahed is the past tense of aah. [webster](https://www.merriam-webster.com/thesaurus/aahed)

Edit: typo

Re: Show HN: A Wordle for the command line (Spanish and English)

#15

Post by gpvos »

bryceneal wrote: post #2

Neat! One thing that has interested me about this game is whether or not there is an optimal first word. Is there is a word that contains the 5 letters which most commonly appear in the language? Preferably where more of the letters appear at their most common index? If so it seems one should use this word as the first guess each time given we have no other information other than the length of the word.

The order of letters in the guess word may also be relevant, to give you as much information as possible (green is better than yellow; there may be branches in the search tree where twice the same letter in a word is advantageous).

Re: Show HN: A Wordle for the command line (Spanish and English)

#16

Post by rgovostes »

bryceneal wrote: post #2

Neat! One thing that has interested me about this game is whether or not there is an optimal first word. Is there is a word that contains the 5 letters which most commonly appear in the language? Preferably where more of the letters appear at their most common index? If so it seems one should use this word as the first guess each time given we have no other information other than the length of the word.

There have been a few attempts to answer this.

One approach is to look at which guess, averaging across all answer words, gets the most hits. You'd expect a word that uses several of the most common letters ETAOIN etc. This helps you quickly discover letters that appear in the answer. Using this strategy, a common first guess is SOARE, explained here:

https://bert.org/2021/11/24/the-best-starting-word-in-wordle...

Missed letters provide information as well, though. Instead of optimizing for the most green/yellow hits, you could divide the entire solution space by the 3^5 possible green/yellow/gray patterns generated by the initial guess. Which initial guess divides most evenly, so you can reliably eliminate 99% of the possible answers? That would be ROATE, explained here:

https://medium.com/@tglaiel/the-mathematically-optimal-first...

Note that ROATE and SOARE are very similar to one another. There are 729 T's in the Wordle wordlist and 669 S's, so ROATE seems like the slightly better answer.

(If going by the first strategy, why SOARE and not ROATE if there are more T's? Probably because there are 3x as many words that start with S than R in the list, so it optimizes towards locking in the first letter.)

Post reply on HN