I Analyzed 2,315 Wordle Answers to Find the Best Starting Words

michaelbrooks20760-ux — Fri, 02 Oct 2026 01:21:12 +0000

Every Wordle thread eventually turns into the same argument: is it ADIEU? AUDIO? STARE? CRANE? Everyone has a favourite starter, and everyone swears theirs is obviously the best.

I got tired of guessing, so I pulled the letter frequencies across all 2,315 past Wordle answers. That's 2,315 words x 5 letters = 11,575 letter slots. If a letter shows up a lot in actual answers, a good opener should probably contain it — and if a letter almost never shows up, leading with it is just burning a guess.

Here's what the data actually says.

The letter frequency table

I counted every letter in every answer, regardless of position. Simple tally, no weighting, no clever modelling to start with:

Letter	Count	Letter	Count
E	1233	C	477
A	979	U	467
R	899	Y	425
O	754	D	393
T	729	H	389
L	719	P	367
I	671	M	316
S	669	B	281
N	575	G	262
		F	230
		K	210
		W	195
		V	153
		X	37
		Z	35
		Q	29
		J	27

A few things jumped out at me:

E is not close. With 1,233 appearances, E shows up in more than 10% of all letter slots. Any opener without an E is fighting the data from move one.
The top is vowel-heavy, but not only vowels. E, A, O, and I are all in the top seven, but R, T, L, and S are right there with them. A starter like AUDIO spends three slots on U, D, and I — letters ranked 11th, 14th, and 7th.
The bottom falls off a cliff. J (27), Q (29), Z (35), and X (37) combined appear fewer times than V alone, and way fewer times than a single mid-tier letter like C. Fun letters, terrible openers.

How I scored starting words

The scoring method is deliberately boring, because boring is easy to check:

Take a candidate 5-letter word with no repeated letters.
Add up the frequency counts for each of its five letters.
Rank every candidate by that total, highest first.

In code, it's basically:

score = sum(letter_counts[letter] for letter in word)

No position weighting, no information theory — we'll get to the limitations in a second. Here are the top scores it produces:

Starting word	Score	Why it scores well
ROATE / ORATE	4594	R + O + A + T + E — five of the top six letters
AROSE / SOARE	4534	Swaps T for S, still covers A, R, O, and E
IRATE	4511	The only top entry that gets I in there too
STARE	4509	S, T, A, R, E — the classic, and the data backs it up

Yes, long-time Wordle nerds will recognise ROATE and SOARE — they've been community favourites for a while, and it's nice to see a dumb frequency count land in roughly the same place as much fancier solvers.

STARE fans, you can feel vindicated too: 4,509 is only 85 points behind the top score, which over 11,575 slots is basically a rounding difference.

The takeaway: think in two guesses, not one

Here's where I think people misuse tables like this. The highest-scoring single word isn't the whole strategy, because your second guess should cover completely different ground.

The pattern that falls out of the counts is:

Guess 1 — vowel-heavy opener: Play something like ROATE, ORATE, AROSE, or STARE. You test A, E, and O (plus R and T or S) in one shot.
Guess 2 — consonant-heavy follow-up: If the opener comes back mostly grey, don't waste guess two on more of the same. Pivot to high-frequency consonants you haven't tried yet — L, N, S, C, H, D, P — with I or U if you still need a vowel. A word built from that second tier covers most of what's left.

Between those two guesses you can test 10 distinct letters, and if you pick them by frequency, they're the 10 letters most likely to actually be in the answer. That, more than any single magic starter word, is what consistently shrinks the candidate list.

Caveats, because it's data

Frequency is not the same as information. This count ignores letter position (E at the end is much more common than E at the start, for example), and it ignores how a word splits the remaining answer pool — which is what proper entropy-based solvers optimise for. A repeated-letter word can also score well on raw frequency while testing fewer distinct letters, which is why I excluded repeats from the ranking.

But as a sanity check on folk wisdom? It works. The starters people love mostly survive contact with the numbers, and the starters built on rare letters really don't.

So my new routine is boring, data-approved, and working pretty well: STARE or ROATE first, a consonant-heavy second word if it whiffs, and no more opening with words containing J, Q, X, or Z just for the vibes.

What's your starter — and does it survive the table above?

I keep the full stats and daily hints updated at WordleHintsToday if you want to dig deeper.

DEV Community: michaelbrooks20760-ux

I Analyzed 2,315 Wordle Answers to Find the Best Starting Words

The letter frequency table

How I scored starting words

The takeaway: think in two guesses, not one

Caveats, because it's data