Why Some Wordle Openers Are So Much Better Than Others

Short answer: letter economy. A strong Wordle opener asks five useful questions at once. It tests five distinct letters that appear in a large share of answers, and it places them where those letters usually sit. A weak opener spends the same five tiles on repeated or rare letters, so whichever colours come back, very little gets ruled out. Entropy is the single number that summarises the difference, and when you score every valid guess the same way, the gap from the top of the table to the bottom is enormous.

Five tiles, five questions

Every guess splits the remaining candidates into groups, one for each possible pattern of grey, yellow and green tiles. A guess is strong when it splits the field into many small groups, because then no single response can leave you with much left to sift. Three properties drive the split. Distinct letters make each tile probe something new. Common letters make the feedback vary from answer to answer, which is exactly what makes it informative. And positional fit puts letters where answers usually hold them, converting would-be yellows into greens and sharpening the pattern further.

Entropy rolls those three properties into one measurement: the average information a guess extracts, in bits, where each extra bit roughly halves the candidates you should expect to face. We scored all 15,783 valid guesses against the 1,352 answers in the curated answer list. The words at the top of that table share the same anatomy, and the words at the bottom fail in the same two ways.

A strong opener against a weak one

TARES tops the entropy ranking across every valid guess. MAMMA sits near the very bottom of the same table, at position 15,731 of 15,783. Both are legal openers, and both spend two tiles on vowels, close to the 1.78 vowels an average answer carries. What separates them is what the other tiles are doing.

The gap is not subtle. After TARES, about 28 candidate answers survive on average. After MAMMA, about 435 do, roughly 16 times as many. One word turns guess two into a precision strike. The other leaves guess two doing the job of a first guess.

Where MAMMA wastes its tiles

MAMMA spends five tiles on two distinct letters, three on M and two on A. From the second M onwards, every repeat can only refine a question the guess has already asked. Feedback on a duplicated letter mostly tells you how many copies the answer contains, which is a useful question late in a game and a wasteful one first. It is also a bet that rarely pays: only 31.7% of answers in the curated list repeat a letter at all, a pattern we unpack in our duplicate-letter research.

Rare letters are the other classic waste, and they can cost even more than repeats. JAZZY tests 4 distinct letters, twice as many as MAMMA, yet it extracts just 2.22 bits, lower still. J and Z appear in so few answers that the overwhelmingly likely outcome is a row of greys that eliminates almost nothing. A letter is only informative when a meaningful share of answers actually contain it.

Placement does the quietest work of the three. The most common final letter across the curated answer list is S, and TARES ends with an S while keeping its vowels in the interior slots English favours. Well-placed letters come back green rather than yellow, and a green pins both the letter and its position, which splits the surviving candidates far harder.

Entropy is the summary, not the secret

None of this requires memorising tables. Entropy condenses distinct letters, letter frequency and positional fit into a single score, which is why our ranked list of the best starting words sorts by it, and why the same few words keep surfacing at the top however you slice the data. If you want the measurement itself unpacked, our entropy research walks through exactly how the scoring works. The practical rule fits in one sentence: open with five distinct, common letters in natural positions, and save repeats and rarities for later guesses, when the surviving candidates specifically call for them.