How to solve Wordle

A practical strategy guide built on the actual Wordle solution pool: the 2,309 words the New York Times draws from. Letter frequencies, opening theory, an elimination playbook for the middle turns, and the honest tradeoffs of hard mode.

How the game works (in case it doesn’t)

You have six guesses to find a five-letter word. After each guess, every tile turns one of three colors. Green means the letter is in the answer and in the right slot. Yellow means the letter is in the answer but in a different slot. Gray means the letter isn’t in the answer at all — with one wrinkle, which is that if you guess a word with two of the same letter and the answer only has one, you’ll see one colored tile (green or yellow) and one gray tile for the same letter.

The answer changes once a day at midnight local New York time. Everyone in the world plays the same puzzle, which is the social mechanic that made the game blow up in early 2022. If you want zero-knowledge coverage, our rules explainer walks through every tile, keyboard color, and the share grid.

Letter frequency: the data

Solving Wordle efficiently is mostly a letter-frequency game. Some letters appear far more often than others in the solution pool, so guessing them early eliminates more candidates faster. Here’s the actual frequency distribution across the 2,309 words NYT uses as solutions:

Letter frequency in the Wordle solution poolBar chart of how often each letter appears across the 2,309 possible Wordle answers. E is most common at 10.65 percent; J, Q, X, Z are the four rarest.E10.7A8.4R7.8O6.5T6.3L6.2I5.8S5.8N5.0C4.1U4.0Y3.7D3.4H3.4P3.2M2.7G2.7B2.4F2.0K1.8W1.7V1.3Z0.3X0.3Q0.3J0.2
Percent of all letter slots across the 2,309-word Wordle solution pool. Top 5 (E, A, R, O, T) highlighted.

E is the runaway winner at 10.65 percent of all letter slots. A, R, O, and T round out the top five, each appearing in roughly 6 to 9 percent of solution letters. The next tier — L, I, S, N, C — covers another 5 to 6 percent each. Together, those ten letters account for more than two-thirds of every letter in every solution.

The takeaway: any opener that covers four or five of the top-10 letters does most of the work of narrowing the answer space. That’s why words like STARE, AROSE, IRATE, ARISE, and TARES dominate “best opener” lists. We rank the top-20 of them, with the full methodology, on our starting-words page.

Vowels first, or consonants first?

A perennial debate. The vowel-first camp opens with something like ADIEU or AUDIO to cover four vowels in one guess; the consonant-first camp opens with something like STARE or SLATE, on the theory that consonants discriminate solutions more sharply than vowels.

The data favors the consonant-first camp, gently. There are 5 vowels and 21 consonants in English; vowels are over-represented per letter (because every word needs at least one) but under-represented per word (because most words only need one or two). An opener that spends three of its five letters on vowels spends those slots on a small alphabet. STARE spends one slot on a vowel and four on the highest-frequency consonants — it learns more, on average, per guess.

The exception is your second guess. If your first guess returned zero vowels, plowing into a vowel-heavy second guess like OUIJA or ADIEU is correct — you’ve confirmed the answer has uncommon vowels (probably U, I, or two of the same), and you need to find them quickly.

The information-theory shortcut

There’s a well-known math-channel video that frames Wordle as an information-theory problem and computes the theoretically-optimal opening word using expected information gain. The answer it lands on is SALET — not STARE, not CRANE — because SALET splits the remaining solution space more evenly across the 243 possible tile-color outcomes than any other opener does.

That’s the theoretical answer, and it’s correct under its assumptions. In practice, the difference between SALET and any of the top-20 letter-coverage openers is on the order of 0.05 of an expected guess — about one extra guess every twenty puzzles. For a human player who is going to spend most of their solving time on the middle three turns anyway, that’s noise. Use the opener you remember and like; the information-theoretic advantage is real but small.

The data viz that follows is more useful: it shows, for each of the top-10 coverage openers, how many greens you can expect on turn one and which positions tend to light up.

Top-10 coverage openers. Each cell is the probability the listed letter lands as a green at that position; the right column sums to the expected total greens from that opener.
Word12345Expected greens
ORATE1.8%11.6%13.3%6.0%18.3%0.51
OATER1.8%13.2%4.8%13.8%9.2%0.43
ROATE4.5%12.1%13.3%6.0%18.3%0.54
REALO4.5%10.4%13.3%7.0%2.5%0.38
LATER3.8%13.2%4.8%13.8%9.2%0.45
ALERT6.1%8.7%7.7%6.5%11.0%0.40
ALTER6.1%8.7%4.8%13.8%9.2%0.42
TALER6.5%13.2%4.9%13.8%9.2%0.47
ARTEL6.1%11.6%4.8%13.8%6.7%0.43
RATEL4.5%13.2%4.8%13.8%6.7%0.43

None of these openers averages even a full green per game. That sounds discouraging until you realize the real value is in the yellows: a good opener that returns one green and two yellows has eliminated roughly 90 percent of the solution space in a single guess.

The elimination playbook

Once your opener returns its colors, the rest of the puzzle is an elimination exercise. The rules of thumb that work:

  • Keep greens in place. Always. If E is green at position 4, every subsequent guess has E at position 4. This isn’t just a hard-mode rule; it’s how you avoid wasting guesses re-discovering things you already know.
  • Move yellows to a new slot. A yellow A means “A is in the answer, but not here.” Putting it back in the same slot relearns nothing. Put it somewhere else.
  • Don’t reuse grays. Every gray letter is one fewer candidate. Using a gray letter again wastes a slot that could have tested a new letter. The exception is double-letter cases (one tile gray, one tile colored) — in those, the gray means “not another one,” not “not in the word.”
  • Test the highest-frequency unknown letters next. After your opener and second guess, the letters you haven’t tested are usually some mix of L, I, N, C, U, H, D, P, M. Build a guess that uses three or four of those.
  • When stuck, burn a turn for pure information. On turn four or five with three candidates left, guessing a word that tests the distinguishing letters — even if that word can’t itself be the answer — is usually the fastest path to a green-row finish. The exception is hard mode, where this strategy is illegal; see below.

Common mistakes

The mistakes that cost most players the most turns:

  • Repeating a gray letter. Common enough that it has a meme. Always cross-check your candidate word against the keyboard before submitting.
  • Ignoring position info on yellows. A yellow E at position 1 doesn’t just mean “E is in the answer.” It means “E is in the answer and isn’t at position 1.” That second clause matters when the answer has only one E.
  • Locking onto the first plausible word that fits. Confirmation bias is a real Wordle killer. When you see a word that matches your constraints, check the other candidates that also match before guessing. The puzzle rewards patience, especially on turn five and six.
  • Forgetting about double letters. Roughly 6 percent of solutions contain a repeated letter (FLUFF, PARER, MAMMA). When you’ve eliminated most of the alphabet and the remaining candidates won’t fit, ask yourself whether the answer might have a doubled letter you haven’t accounted for. Some of the historically hardest puzzles fall into this trap; see our list of the hardest Wordle answers for examples.

Is hard mode worth it?

Hard mode enforces two rules that good players already follow: confirmed greens must stay in place, and revealed yellows must appear somewhere in subsequent guesses. What it forbids is the “burn a turn for information” play described above.

If you average four guesses per puzzle in normal mode, hard mode probably pushes you to 4.3 or 4.4 — not because the extra rules are harder, but because the information-burning move is sometimes the right one. If you’re solving for streak length more than puzzle elegance, stay in normal mode. If you want a sharper challenge, flip it on and live with the occasional unwinnable position. Our hard-mode guide covers the full ruleset and the strategy adjustments that matter.

Want to skip ahead today?

The strategy is what the algorithm already does — pick a starting word with maximum information gain, calibrated to how many guesses you have left.

Use the hint tool →

Looking for the answer to a past puzzle? The Wordle archive has every NYT puzzle going back to early 2022, each with the answer, definition, and the hints our algorithm would have offered that day.