Follow

What's the best first guess in wordle? Did some quick analysis on the hardcoded lists
- Word with the highest likelihood of each independent letter being correct: "slate"
- Word that rules out the most incorrect words: "canoe"
jon-e.net/blog/2022/01/10/word

"slate" is good because it has a bunch of common letters in the places they're most frequently found, (left image is counts of letters in each position). Almost all of the best words by this measure start with "s" and end with "e" ("y" is also overwhelmingly common at the end)

"canoe" is good because it has three vowels, with the relatively uncommon "c" in its most common position, and "e" in its most common position. This gives you a lot of information about what vowels are in (or not in) a word. It has a median reduction of ~90% of incorrect words.

oh ya also ty to @qntm@twitter.com for spotting that the lists were hardcoded in the source, they've made a v twisty adversarial version i can't quite wrap my head around the strategy for twitter.com/qntm/status/147998

the game works sort of like how the entropy reduction calculation here works, selects a random correct word after each guess from the list of possible remaining words, given the preceding guesses and correct words. Using the maximally entropy-reducing words works, but then...?

in another life I would be a morphologist. wondering how those patterns of letter frequency depend on word length. how much more entropy 6 letters has (eg. do constraints on morphemes relax and open more of the relative combinatoric space of letter combinations, or less?)

also the words in the game from what I understand are a p arbitrary subset of words based on how recognizable the devs thought they were, wonder how much that influences the information available in a given guess

@brennen @jonny the notion of the best first guess in wordle depends on playstyle; a penpangrammatic approach to constructing five-guess sets renders the choice of first one moot while also constraining the use of individual words with a bunch of common letters, as you will need some of them elsewhere. But I seem to be mostly alone in my desires not to go guess golfing .

@brennen @jonny also entertaining that one of the words of my unbeaten set is in fact "nymph", which it describes as one of the worst words, but it is a delightfully-useful word in using up almost a fifth of the alphabet as compactly as possible in a way that leaves something for the other four words

@brennen @jonny although I am interested in the rest of that "worst words" set that starts NYMPH INBOX ETHOS so as to be able to play "hard mode" without the Puritan punishment of being forced to bear all your correct positional guesses along with you

@earthtopus
@brennen
lost me a bit, but of course everyone has their own playstyle! tried to get that across by saying "depends on how you measure it" but was unclear. this was mostly me just puzzling on the entropy space and constraints placed by English morphology. I said it elsewhere but not optimizing against the win condition and enacting your own structure on the game is its own real joy ❤️

@earthtopus @brennen
specific to nymph, that's just with the independent probability of letters in positions, which I was mostly doing as a warmup and sort of red-herring bc of course letter probability isn't independent -- they're in words!

@jonny @brennen

yeah, sorry, I was a little manic, but basically: I use the maxmum of five wrong guesses every single time designed to eliminate possibilities, and then solve the puzzle with the information provided.

If ABCDE FGHIJ KLMNO PQRST and UVWXY were words, imagine me using them. If there weren't four green or yellow boxes after this, the remaining letters have to be repeated letters or Z.

@jonny @brennen

Given the curation of the potential set of prompts, I spend very little time actually playing the game and more time thinking about weird words and refining my guess sets towards the dream of 25 distinct characters (I'm up to 24!) With my playstyle ruling out most of the alphabet anyway words like NYMPH become incredibly useful because of/in spite of the low probability of containing letters at more common positions

@davidoclubb
:) am not a phonotactician but do research in phonetic perception and have been a computational morphology hobbyist for awhile. the phonology of the language creating a constraint space for its letter sequences is exactly why I think this is fun to think about :)

@jonny oooh blimey you are way more into the science of this than me! I pick a random start word each day

Sign in to participate in the conversation
social.coop

The social network of the future: No ads, no corporate surveillance, ethical design, and decentralization! Own your data with Mastodon!