ETAOIN SHRDLU: How Letter Frequency Wins Word Games

For most of a century, the fastest way to see English letter frequency was to look at a Linotype machine. Its keyboard arranged keys by workload, and the first two columns read E-T-A-O-I-N and S-H-R-D-L-U. When an operator fumbled a line, the habit was to run a finger down those columns to fill it out and flag it for discard — and "etaoin shrdlu" slipped into actual print often enough to become typesetters' folklore. That nonsense string is still, near enough, the frequency table of English. Everyone who's good at word games carries some version of it in their head, whether they know it or not.

The table

Count letters across any large pile of English text and the same order settles out:

Those top six letters account for roughly half of all English text, which is the single most useful fact in this article. Probe E, T, A, O, I and N and you've probably touched most of any hidden word.

Position matters as much as presence

The table hides a second table. S sits around seventh overall, yet it starts more English words than any other letter — the S section is the fattest in every dictionary you'll ever open. E runs the opposite way: common everywhere, but it especially loves the final slot, where silent E and endings like -SE, -TE and -LE pile up. Y is a nobody in running text and a star at the end of five-letter words. Among pairs, TH leads the language, with HE, IN and ER close behind.

Position is why a word like STARE is better than it looks. It isn't just five common letters; it's common letters standing where they like to stand — S up front, E at the back, R rolling around the middle. Strong players think in positions, not just letters.

An opener is a question, not a bet

In Word Master you get six guesses at a five-letter word, and the tiles answer back: white for the right letter in the right place, gray for a letter that's in the word but elsewhere, dark for absent. The beginner instinct is to hunt vowels — open with AUDIO and hope something lights up. That instinct is wrong in an instructive way. A first guess isn't an attempt to be right. It's a question, and good questions are ones whose answers you can't predict. AUDIO mostly asks which one vowel the word contains, so it sorts the dictionary into a few lumpy piles. An opener built from E-T-A-O-I-N-S-R material splits it into many small piles instead, and whatever the tiles say, you've eliminated most of the language in one move. Three dark tiles that each kill a fat slice of the candidate list beat two white ones on a word you still can't name.

Push this far enough and it stops being philosophy and becomes arithmetic. In 2022 two MIT researchers ran the original Wordle's entire game tree, and the mathematically best opener came out as SALET — a medieval helmet — averaging about 3.42 guesses, with familiar favourites like CRANE and SLATE a whisker behind. Look at what the top openers share: high-frequency consonants, one or two vowels, no repeated letters, everything parked in its favourite position. Word Master runs on its own word list, so there's no single sacred opener here, but the recipe travels unchanged.

Keep asking questions on guess two. If SLATE comes back all dark, the strong follow-up is five fresh letters — CORNY, chasing the next tier of the table — not a rearrangement of the corpse of guess one.

Anagram games flip the table

In Word Wheel and Word Connect nothing is hidden. The letters sit in front of you, the job is to see words in them, and frequency knowledge changes costume: what matters now is how English assembles words from parts. Suffixes first — -ER, -ING, -ED and -S do enormous work — then prefixes like RE-, UN- and OUT-. Hunt families, not words: find one -ER word and its -ED and -S cousins are usually hiding in the same letters. In Word Wheel, where every word must use the centre letter, read that letter's habits before anything else. A centre R wants -ER endings. A centre G means go straight for -ING.

Codeword is the hybrid where raw frequency comes back swinging. Every letter has been swapped for a number, and the opening technique is a census: count how often each number appears, because the most common one is nearly always E, with T and A behind it. Doubled numbers narrow things fast — LL, SS, EE and OO are the usual suspects — and a three-cell word ending in your suspected E is begging to be THE. This is frequency analysis, the same technique that broke ciphers for a thousand years before it started winning puzzles, and it works for the same reason: English can't help being itself.

Two families of game, one table. In guessing games the table writes your questions; in anagram games it predicts the answers. Even Word Search bows to it from the other side — the fastest solvers scan the grid for a target word's rarest letter, because a Q or a Z stands out of the noise the way an E never will. The table is small. Learn the first six letters and where they like to stand, and every word game on this site gets a little easier at once.

More from the blog

Or skip the reading and just play something — everything here is free, in your browser, no download.