Wordle clones and variations


What better way to start this blog than by chronicling my experience with the first real trend of 2022?


Play Wordle (powerlanguage)

Play Wordle (The New York Times)

I first heard of Wordle in early 2022. Despite having seen the infamous coloured squares on Twitter and even having been sent the link to the game, I found Wordle and starting playing on the night of the 15th of January; I told a friend: “you’ve gotta play this”. I was impressed by having guessed that day’s puzzle in only 5 guesses.

For the next few days and weeks after that, I started looking forward to solving each day’s Wordle just as it became available at midnight. During the first few days of puzzles I crafted my guesses carefully, getting up to a streak of about 11. Over time, I got better and better at the game and could often solve the puzzle in just a few minutes. At that point, playing the original Wordle was less interesting because I had a pair of good first guesses lined up – “early” and “opium” – which considerably narrowed down the range of possibilities, though I now know that these initial guesses aren’t exactly optimal as they completely exhaust the vowels. If I was still pursuing the goal of solving Wordle fast, I would think of a set of three guesses which together exhaust all the vowels and all the common consonants, but repeat no letters.

Wordle Unlimited

Play Wordle Unlimited

Wordle Unlimited is my own version of Wordle, reverse engineered to allow me to play as many times as I like. Rapid-fire, continuous Wordle puzzles remain fun, especially with others. The challenge for me is still to solve the puzzles as quickly as possible, rather than using as few guesses as possible.


Play Semantle

I started playing Semantle in late February – it proved to be a challenge. To date, I have solved about 5 puzzles in total, maybe 2 of those having been solved without the help of others.

Initially, my logic was to get a reasomably good guess, then to see what similar words moved me “in the correct direction”, that is to say moved me closer to the target word. I assumed that I would be able to “extrapolate” and continue moving “in the correct direction” to arrive at the target word. For example, if my initial good guess was “person” and a better guess turned out to be “engineer”, I would hope to be able to find a word that is to “engineer” as “engineer” is to “person”.

In reality, this logic was flawed. It ended up being difficult to get closer to the target word in any meeaningful way while still maintaining a connection with any of my previous guesses.

Here is an example of this strategy failing. “War” is an early good guess, but the target word is so dissimilar to it, that no amount of guessing conflict-related terms gets me any closer to the target.

# Guess Similarity
12 war 31.24
60 strategic 30.06
43 conflict 29.10
64 warlike 27.28
100 racist 24.01
106 united 23.72
52 army 23.58
18 genocide 22.72
57 strategy 21.75
101 segregation 21.73
49 naval 21.45
63 nuclear 20.48
17 criminal 20.43
21 aggression 19.80
58 financial 19.03
83 mathematics 17.75

A better strategy turned out to be to guess completely random words at first, ones which covered as many different ideas as possible, while using all the different parts of speech (this can sometimes make a big difference) until you have stumbled by chance into a word in the top 100 closest words to your target. You can make slightly better guesses if you know a few words in the top 1000 closest words, but you shouldn’t rely on this completely to inform your guesses to avoid repeating the flawed strategy outlined above.

However, once you have a guess in the top 100 or 200 closest words, you can confidently say that you are in the same domain as the target. The closer you are to the target, the closer your subsequent guesses should be to the current closest guess. It is often the case that there are many words here that share the same root, so you must be creative and think of the words derived or formed from your current closest guess. It is helpful to think “I see this word appear in a news article. What other words do I see?” – prior knowledge of the dataset helps and I don’t consider it to be cheating.

Here is an example of this in action (from the same game as above, game 28). “Conservative” is a guess in the top 100 closest words, and any similar word is bound to be close as well.

# Guess Similarity Getting close?
359 ideological 100.0 FOUND!
356 ideology 74.70 999
355 ideologies 67.39 998
301 liberal 59.31 993
292 conservatism 57.01 991
288 conservatives 51.76 969
316 libertarian 51.17 960
315 progressivism 50.19 941
309 socialist 49.45 931
287 conservative 48.94 923
310 socialism 46.35 845
317 libertarianism 46.34 843
307 nationalist 45.99 832

I may have made mistakes in typing out these tables. You can view a full list of the top 1000 closest words to “ideological”. Notice that the further away a given word is from the target, the further away it is to its neighbours in the list. I like to imagine that this is because there are a great deal more words that are far away from the target than words which are very close to the target; the far away words are allowed to cover a broad domain of language.


Play Worldle

I am rubbish at geography, so this wasn’t much fun, although I can guess familiar countries and try to get ever closer to the target by imagining where they all lie on a world map.

Quordle and Octordle

Play Quordle

Play Octordle

Quordle is arguably the most similar variation to the original Wordle game. Here, there is an even greater emphasis on not wasting the first few guesses and obtaining information about the most frequently-used letters of the alphabet. It did not pose a challenge for me, but the game takes markedly longer than the original Wordle and so provides a more rewarding experience.

Similary to the original Wordle, Quordle and Octordle give the player 5 “initial” guesses, after which they must be successful in their guesses, if they wish to complete all the simultaneous games.