Optimle is a Wordle variant I made in February 2022, back when everybody was making Wordle-inspired games. A lot of Wordlelikes change one of the same few parameters of the game concept, especially "what you're guessing", or sometimes "what feedback you're given", "how your guesses are restricted", and so on. I wanted to find a design parameter that I hadn't seen anyone else pull the lever on yet.
Rather than starting from an empty grid, each puzzle starts from a prepopulated table of guesses that serves to narrow down the set of possible words. You have two guesses remaining to "guarantee" a win. What exactly does "guarantee" mean? Put yourself in the shoes of the unluckiest Wordle player: if there's just one word they're not considering when picking their penultimate guess, that'll inevitably be the secret word.
More precisely, consider the asymmetric roles played by the guesser (you) and the evaluator (what gives you feedback on your guesses). In Wordle, the evaluator compares your guess to the day's secret word. In Optimle, the evaluator instead behaves adversarially within the bounds of the game's rules. It considers all the possible secret words that are valid given the responses to previous guesses, then responds with whatever feedback allows the most words to remain valid. (Absurdle explores this same idea, but in a more open-ended way.)
If you haven't tried it yet, I encourage you to give today's puzzle a shot before reading on! The rest of this post explores the strategy involved in solving these puzzles, which you may want to discover for yourself first.
Let's look at the example from the game's help screen:
Resist the temptation to immediately start entering guesses; there are always at least 3 valid secret words at the start.1 Do take some time to enumerate them in your head or notepad, though, and try to be exhaustive. This can be an irritating task for certain sets of green letters;2 as a commiseration, only common words are considered to be valid possible secret words, and typing an "uncommon" but real word will tell you, before pressing Enter, that the word you've typed is not in the secret word list. You can use this feature to separate uncommon from common words, but not non-words from words.
With only two guesses remaining, you need to carefully tune your first guess such that the adversarial evaluator can't give you an ambiguous response. Once you think you've got all the possible secret words, find a word such that every possible response from the adversarial evaluator fully disambiguates the secret word. For the NORTH / FORTH / WORTH example, you want a word that includes at least 2/3 of the letters N / F / W. For example, SWORN would work: either the W, the N, or neither will light up, mapping to each of the 3 possible words.
Once you've locked your first word in, you should have enough information to guess the secret word. If there's nothing else it could be, you win! Otherwise, the game will tell you what else the secret word could've been, and you can retry the puzzle with that information in mind. (Your "score" is how many attempts you took to find a solution - getting it on the first try is a nice badge of honor, but you have unlimited attempts, so don't be afraid of a miss.)
Knowing the variable letters (N / F / W in the example above) isn't always enough to find a winning first guess. Consider the following puzzle:
Spoilers for the first half of the strategy: the possible words here are CASTE, HASTE, TASTE, and WASTE, making the variable letters C, H, T, and W. So, let's find a word that has at least 3 of those letters and get guessing:
Aha, the T lit up! Let's lock it in...
...and now we've lost our first try. The T was more than just a variable letter - it was also one of the letters already revealed in the puzzle! Having a T light up yellow doesn't give us enough information to distinguish TASTE from WASTE in this case.
We have two ways to deal with this on the first guess:
You may notice the numbers 2 3 4 in the top-left corner of the page. Clicking each of these reveals a different puzzle where the prepopulated guesses leave you with that number of green letters. I feel that 3 yields the ideal game balance for an experienced player, while 4 serves as a gentler introduction for newcomers, and 2 as an extra challenge for masochists.
Puzzles with multiple unknown letters aren't just harder during the initial "find all the possible words" phase; they introduce new edge cases to the second phase, similar to the one described above. You may need to improvise a notation system using pen & paper for particularly thorny cases, where variable letters are shared across the unknown spaces, possibly further complicated by yellow letters in the prepopulated table. One player documented his thought process for an extra-hard 2 puzzle in this thread:
This is by far my favorite aspect of Optimle: even once you understand the general strategy, the constraints imposed on your first guess are often entangled, requiring a deep understanding of the original game's mechanics and attention to detail when mentally evaluating possible guesses.
I wrote all the logic for this game in Rust (my first Rust project!) because I knew I would want a highly performant engine for generating "interesting" puzzles by brute force. A portion of this Rust code is compiled to WASM for the browser, ensuring that the offline puzzle generation & online guess evaluation behave exactly the same. Nearly the entire game came together in a single weekend, although I added the "this word isn't in the secret list" hint and the 2 / 3 / 4 puzzle variants after the initial release.
1 Also, the puzzle generator is tuned to never allow a valid possible secret word to fully disambiguate the other 2+ valid secret words. This can happen with puzzles that give you 3 or fewer green letters, but they'll never show up as an Optimle puzzle.
2 Yesterday's 4 puzzle had "BILLY" as one of the possible secret words. Proper nouns that are not only real words, but "common" according to Wordle's dictionary, are the bane of my existence.