Open your favorite Wordle-style game, type your trusty opener, and hit enter. If that word is ADIEU, this simulation has bad news: in the game I tested, ADIEU isn't even a legal guess — and neither is CRANE, one of the most-recommended starters on the internet.

I know because I didn't guess. I ran the numbers. I took the complete answer bank and valid-guess list from a real daily word game — Practical Web Tools' Daily Word, a free browser puzzle with one five-letter word per day — and simulated every possible opening guess against every possible answer, all 14.4 million grading operations, using the game's actual duplicate-letter grading rules. Then I played out full games with the top contenders to see how much an opener really matters.

Try Daily Word Puzzle — free, no sign-up, and it runs entirely in your browser.

The results overturned three pieces of advice I'd internalized. Here's what the data actually says.

What I actually tested (and why the dictionary matters)

Most "best Wordle starter" articles rank words against one specific dictionary — the original game's 2,309 answers. But every Wordle-style game uses its own word bank. Daily Word, the game I tested, uses 503 common five-letter answers and accepts 1,235 valid guesses. That difference changes everything.

ADIEU is rejected. So are CRANE, SALET, TEARS, TONER, and PIUS. Type them and nothing happens.
AUDIO is legal but mediocre — it ranks #688 out of 1,235 valid guesses in this game.
And the word that beats everything isn't a vowel-maximizer at all.

The lesson before we even get to rankings: strategy advice doesn't transfer between games with different dictionaries. The only rankings that matter are the ones computed against the bank you're actually playing.

How you score an opening word (the honest version)

An opener's job is to shrink the possibilities as fast as possible. The metric I used is expected remaining answers — how many candidate words, on average, are still alive after you enter the guess and see the colored tiles.

Here's the intuition. Say your first guess leaves 20 possibilities on average. Each follow-up guess splits that pool further. A guess that leaves 40 possibilities on average is measurably worse — you've wasted a turn learning less. Lower is better, and small differences compound over six guesses.

I also tracked each opener's worst case (the biggest single bucket of answers it can fail to separate) and its green rate — how often it lands at least one letter in the correct slot on the opening guess. Those three numbers together paint a fuller picture than any single score.

The winner isn't ADIEU — it's IRATE

Ranking all 1,235 legal guesses by expected remaining answers, the top of the table looks like this:

IRATE — 10.3 expected remaining answers. Best of all 1,235 guesses, with the smallest worst-case bucket (24) in the entire game.
ARISE — 11.4. Nearly as good, same tiny worst case (28).
ALERT — 12.4. The best opener built from a word starting with A-L.
TRACE — 12.5, with a 50% green rate — the highest among the elite tier.
STARE — 12.7, TRADE — 12.7, RAISE — 12.8, SLATE — 12.9.

IRATE wins on both expected value AND worst case — a rare double crown that makes it the safest all-around opener in this game.

For contrast, AUDIO — probably the most popular vowel-first opener after ADIEU — leaves 41.3 candidates on average, four times IRATE's number, and its worst-case bucket is 96 words. In a game with only six guesses, that's the difference between a calm solve and a coin flip. Want a deeper look at the vowel-first theory and when it actually helps? Our Wordle trap strategy guide covers the two-green-letters scenario in detail.

The vowel myth, measured

The classic advice says: open with as many vowels as possible, because vowels appear in nearly every English word. The simulation confirms the premise but destroys the conclusion.

In this game's answer bank, 72.4% of answers contain A or E — the premise holds. But look at the letter statistics across all 503 answers:

E fills 11.7% of all letter slots — the most common letter by far.
A (8.9%), R (8.2%), T (8.0%), O (6.4%), I (6.0%), L (6.0%), S (5.8%) round out the top.
The best openers load up on E, A, R, T, I — which is exactly why IRATE, ARISE, ALERT, TRACE and SLATE dominate. They're not vowel-maximizers; they're frequency-maximizers.

AUDIO spends four of its five letters on A, U, D, I, O — and U is the waste. U appears in just a sliver of answers, and D is mid-tier at best. You're burning two slots to maybe-learn about rare letters while R and T — top-five letters both — go untested.

Opening with three vowels feels safe. The data says it's the slowest way to learn anything.

Why the dictionary keeps punishing famous openers

The deeper story is that answer banks differ in structure, not just membership. Three structural facts about this game's 503 answers shape the whole ranking:

S is the most common first letter (70 answers start with S, vs 44 for C and T).
E is the most common last letter by a landslide — 111 answers end in E, nearly triple the next finisher.
28% of answers contain a repeated letter (LL, SS, OO and friends), slightly higher than you'd guess.

That E-ending dominance is why IRATE and ARISE — both ending in E — carve the pool so finely, and why TRACE/SLATE (also E-final) post the highest green rates at 50%. If your opener doesn't test E's position, you're ignoring the single most informative slot in the game.

The repeated-letter rate has strategic value too: when you're down to two candidates and one has a double letter, remember the game grades duplicates properly — greens are assigned first, then yellows, left to right. A word like SLEEP against an answer with one E still shows one green and one yellow, not two greens.

What happens on turn two (this surprised me)

The most common IRATE outcome isn't all-gray — it's ⬛⬛⬛⬛🟨 (gray-gray-gray-gray-yellow), leaving 24 candidates. The best second guess there isn't a common word at all: it's KNEEL, cutting the pool to an expected 1.3 answers (best answer-pool pick: NEEDS, at 1.5).

After a full-gray IRATE (20 candidates remain), the optimal second is SCOLD — expected 1.8 remaining — though if you'd rather guess something that could win immediately, BOUND is the best in-pool choice at 2.0.

The single biggest skill gap between average and strong players isn't the opener — it's knowing when to sacrifice a turn on a "burner" word that can't win but splits the pool perfectly.

Does the opener really matter? I played 503 full games to check

Ranking openers by information is one thing. To see the endgame effect, I had a deterministic solver play out all 503 answers for each top opener — always guessing to minimize expected remaining answers, winning only by typing the answer itself.

TRACE and SLATE: 2.918 average guesses to win, zero losses.
ALERT: 2.946, IRATE: 2.948, LEAST: 2.960 — statistical ties.
ARISE: 3.006, RAISE: 3.024 — slightly behind but still excellent.
AUDIO: 3.189 with one six-guess scrape — the cost of ignoring R and T.
RATIO: 3.083 — solid, not elite.

A 0.27-guess edge (TRACE vs AUDIO) sounds small. But distribution matters more than the average: TRACE solved in 2 or 3 guesses 88% of the time; AUDIO needed 4+ guesses on 27% of answers. Over a week of daily play, the elite openers buy you one extra "solved in three" day and meaningfully shrink your loss risk on hard answers.

Honest caveat: among the top eight openers, the differences are within noise for casual play. The real cliff is between the top ~30 words and everything else — that's where average guesses jump.

Your practical pre-game checklist

You don't need to run simulations to use the findings. Before your next game:

Open with IRATE, ARISE, ALERT, TRACE, or SLATE. All five are elite in this game; pick the one you find easiest to build follow-ups around.
If you get mostly grays, don't panic-guess. After gray IRATE, SCOLD or BOUND are the calculated second moves; the all-gray pool still hides BLOOD, BLOCK, and BLOOM-type words.
Watch slot 5. E ends 22% of all answers here — a yellow E in another slot is a massive clue about the ending.
Expect double letters 28% of the time. If four of five letters are placed and nothing fits, the answer probably repeats one.
Don't import openers from other games. ADIEU, CRANE, and SALET die at the entry box here. Test your favorite in the game first.

And if you want a zero-stakes place to drill the system, play Daily Word free in your browser — no download, no account, and the daily answer changes at midnight UTC. When you're warmed up, the site also has word search puzzles and hangman if five-letter logic isn't your mood today.

The takeaway (and a challenge)

The internet's favorite openers fail on arrival in a game with a different dictionary, the vowel-first religion measurably underperforms, and a word most players have never opened with — IRATE — wins on both expected and worst-case information. None of that was knowable from vibes; it took grading every guess against every answer under the game's real rules.

My challenge to you: take your current opener and play one week with IRATE or TRACE instead. Track your guess counts. If your average doesn't drop, your follow-up strategy is the bottleneck — and that's exactly the skill the section above on burner words trains.

The eight faces of IRATE (keep this cheat sheet)

Because IRATE's outcome distribution is lumpy, a handful of tile patterns cover most games. Here's what each common first row actually means, straight from the simulation — candidate counts and real example answers from the 503-word bank:

⬛⬛⬛⬛🟨 (E yellow, 24 candidates): the most common outcome. The answer has an E but not at the end: think BELOW, BENCH, BLEND, CHECK, DOZEN. Best second guess: KNEEL.
🟨⬛⬛🟨⬛ (I and T yellow, 22): I and T are in the word, displaced: BUILT, FIGHT, JOINT, LIGHT, LIMIT, MIGHT.
🟨⬛⬛⬛⬛ (I yellow only, 21): I floats somewhere else: BILLY, BLIND, BUILD, CHILD, CIVIL, CLICK.
⬛⬛⬛⬛⬛ (all gray, 20): no I, R, A, T, or E at all — the answer is consonant-heavy: BLOCK, BLOOD, BLOOM, BOUND, CLOCK, CLOUD. Best second: SCOLD; best winnable: BOUND.
⬛⬛⬛⬛🟩 (E green, 18): you've banked the ending: CHOSE, CLOSE, CYCLE, GLOBE, HENCE, HOUSE.
⬛⬛🟨⬛🟨 (A and E yellow, 14): A and E both present, both displaced: AHEAD, ANGEL, BASES, BEGAN, BLEAK, CHEAP.
⬛⬛⬛🟨🟨 (T and E yellow, 14): CHEST, DEBUT, EVENT, FLEET, GUEST, HOTEL.
⬛⬛🟨⬛⬛ (A yellow only, 13): ALBUM, ALLOW, ALONG, AMONG, APPLY, BADLY.

Notice what the all-gray row really says: not "bad luck," but "the answer avoids five of the most common letters in English" — which itself carves the pool down to 4% of the dictionary. Even your worst first row is a 20x reduction.

FAQ

What is the single best opening word? In this game, IRATE — it leaves 10.3 candidate answers on average (best of all 1,235 legal guesses) and has the smallest worst-case outcome (24 candidates). ARISE, ALERT, TRACE, and SLATE are statistically close behind.

Why is ADIEU a bad opening word here? Two reasons: it's not in this game's valid-guess list at all, and even in games where it's legal, spending four letters on vowels wastes slots — R and T are both top-five letters in the answer bank, and ADIEU tests neither.

Is AUDIO a good opening word? It's legal but ranks #688 of 1,235. It leaves 41.3 candidates on average — four times more than IRATE — because U is a rare letter in answers and the word ignores R and T entirely.

Do the same best openers work in every Wordle-style game? No. Rankings depend entirely on each game's answer dictionary. Openers like CRANE and SALET aren't even valid guesses in this game. The reliable transferable principle is to cover common letters E, A, R, T, and ideally end your opener with E.

How much does the opening word actually matter? Among the top openers, less than legend suggests — roughly 0.1 guesses between elite words. But between an elite opener and a poor one like AUDIO, the gap is about 0.27 average guesses and a large shift in how often you solve in 2-3 turns versus needing 4+.

Should I use a "burner" word that can't be the answer? Often, yes. After a gray-heavy first guess, the mathematically best second guess is sometimes an unlikely word (SCOLD, KNEEL) that splits the remaining candidates nearly perfectly. Sacrificing a turn to halve the pool twice is how strong players protect their streak on hard days.