The Most Common Letters in English (And How Every Word Game Exploits Them)

ETAOIN SHRDLU: The Accidental Phrase That Reveals Everything

If you've ever looked at old newspaper archives, you might have spotted a strange phrase buried in some articles: "etaoin shrdlu." It's not a typo. It's not nonsense. It's actually one of the most important concepts in understanding how the English language works.

In the era of hot-metal typesetting (roughly 1886 to the 1970s), printers used linotype machines to cast entire lines of type in molten lead. The keys on these machines weren't arranged like a modern QWERTY keyboard. They were sorted by frequency. The most common letters ran down the first two key columns from top to bottom: E, T, A, O, I, N on the left, and S, H, R, D, L, U on the right.

When a typesetter made an error, they'd quickly run their fingers down those two columns to waste the metal slug, producing the string "etaoin shrdlu" in the process. The whole line would then get pulled before printing. Usually. Sometimes it slipped through.

What that phrase represents is the 12 most common letters in English, in order of frequency, as measured by actual newspaper text in the late 1800s. More than a century later, modern computational analysis has confirmed the ranking is almost exactly right.

How We First Measured Letter Frequency

Samuel Morse figured this out before linotype machines existed. When he designed Morse code in the 1830s, he wanted common letters to have short codes and rare letters to have long ones. He walked into a print shop in Morristown, New Jersey and counted how many of each letter were in the type cases.

E had the most (hence a single dot). T had the second most (a single dash). Q and Z had almost none (hence the complex sequences we still use today). Morse's system worked because he was essentially doing the same thing modern linguists do: measuring how often letters appear in real written text.

Fast forward to 2012, and Peter Norvig at Google published a detailed letter frequency analysis drawn from a trillion words of English text. His findings matched Morse's 19th-century print shop count almost exactly. The language doesn't lie.

The Full Letter Frequency Table

The table below shows modern frequency data from large English corpora. The percentages represent how often each letter appears across general written English text (novels, newspapers, websites, everything combined). Fair warning: the exact figures shift slightly depending on the corpus and era measured — these are the best current consensus numbers, not a universal constant.

Look at that last column alongside the frequencies. There's no coincidence here. When Alfred Mosher Butts designed Scrabble in the 1930s, he followed Morse's lead: he went to his local library and counted letter frequencies in newspaper articles. The result was the tile distribution you see today. High-frequency letters got many tiles and low point values. Rare letters got one tile and high point values. The game is a direct physical model of English letter frequency.

The Scrabble Tile Distribution Is a Letter Frequency Map

This is one of those things that seems obvious in hindsight but genuinely surprised me the first time I thought about it consciously. I was explaining Scrabble to my nephew one evening in Pune, and he asked why there were so many E tiles. I started to say "because it's a common letter" and then stopped. I realized I'd played hundreds of games without ever thinking about why the bag is weighted the way it is. Scrabble's 100-tile bag isn't arbitrary. It's a frequency chart you can hold in your hands.

E has 12 tiles because it's the most common letter. Not because Butts liked the number 12. Q and Z each have just 1 tile because they barely appear in English text. H appears twice as often as Q in the language, and the Scrabble bag reflects that (H has 2 tiles, Q has 1). The whole tile distribution is essentially a compressed model of English letter usage.

This matters practically. When you draw tiles in Scrabble, you're drawing from a probability-weighted sample of the English language. A full S tile (only 4 exist despite S being the 7th most common letter) is so rare and valuable specifically because it's in the game less than its frequency would suggest. That's a deliberate design choice: S is underrepresented in the bag to prevent everyone from pluraling everything.

The Scrabble Word Finder and the Word Unscrambler both search through word lists that are themselves a product of this same letter frequency. Common letters combine in common patterns, which is why CRANE and SLATE are such strong Wordle starting words. They load up on high-frequency letters.

How Wordle Exploits Letter Frequency

The designers of Wordle (and the NYT team who now curate it) use letter frequency in two opposite ways. They exploit it when choosing accessible answers. Most Wordle words heavily feature E, A, T, O, R, S. And they use less common letter patterns to create the puzzles that spike the Twitter complaint feed.

Think about it: CRANE tests C, R, A, N, E. Together those five letters cover ranks 3, 9, 12, 6, and 1 in frequency. Five letters from the top 12. That's why it's such a strong opener.

But when Wordle uses an answer like NYMPH, TRYST, or CRYPT, it's doing the opposite. These words have unusual letter patterns, low-frequency letters, or even no conventional vowels. They're hard precisely because your starting word's high-frequency letters won't show up in them. Your typical "ETAOIN letters" become useless.

The most common Wordle letter patterns analysis shows that E appears as the final letter in a disproportionate share of Wordle answers, while S appears far less often than its frequency would suggest (the NYT deliberately limits answers ending in S to prevent trivial plurals). Letter frequency in Wordle isn't perfectly mirrored from general English usage. It's curated.

What This Means for Your Starting Word Strategy

Every discussion about the best Wordle starting words ultimately comes back to letter frequency. SLATE, CRANE, SALET, RAISE: all of these are strong because they pack the highest-frequency letters into five tiles. They maximize your expected information gain on guess 1.

The ETAOIN ordering is a useful quick sanity check. A good Wordle starting word should contain as many of E, T, A, O, I, N as possible. CRANE hits C, R, A, N, E (3 of the top 6 frequent letters and 2 others that are still fairly common). SLATE hits S, L, A, T, E (A, T, E from the top 10 plus two solid letters). ADIEU hits A, I, E (the vowel-heavy approach, testing 3 top-5 letters but leaving many consonants untested).

The vowel-only approach (like ADIEU) is controversial because English words need consonants too, and the common consonants (T, N, S, R) are what really separate candidates. Going heavy on vowels in guess 1 loads up on information about a small part of the alphabet. Most serious Wordle players have moved away from ADIEU for exactly this reason.

My take: ADIEU is the word that sounds smart but plays dumb. The data is clear — you're trading a shot at T, N, and R (ranks 2, 6, and 9) for an extra vowel you probably didn't need. If you're still using ADIEU, try CRANE or SLATE for a week and see if your average improves.

Letter Frequency Varies by Context

The 12.7% figure for E is an average across all types of English writing. But frequency shifts depending on what you're reading. Academic papers use different patterns than casual text. Poetry and formal writing differ from news articles.

More interesting for word game players: frequency changes by word position. E is the most common letter overall, but it's especially dominant in the final position of English words. S is common overall, but it's particularly strong at the start of words. The Google Books Ngram Viewer lets you explore how even specific letter combinations have shifted in frequency over centuries of English writing.

The Corpus of Contemporary American English (COCA), a database of over 1 billion words from modern English text, is the most precise tool for this kind of analysis. Researchers who study word game strategy have used COCA data to confirm that the traditional ETAOIN ordering holds up against modern usage, though the exact percentages shift slightly depending on whether you're measuring spoken vs. written English.

English Isn't the Only Language With This Pattern

Every language has its own letter frequency hierarchy, and it's genuinely fascinating how they diverge. In Spanish, A overtakes E as the most common vowel. In German, E is still dominant but the frequency of N and R rises dramatically because of grammatical endings. In French, E is so dominant that entire avant-garde novels have been written without using it (the literary movement is called the lipogram).

This cross-language variation is why foreign-language versions of Scrabble have completely different tile distributions. French Scrabble has more E tiles and fewer tiles overall for letters like K, which barely appears in native French words. The game adapts to whatever the language's actual frequency map looks like.

It also explains why learning a foreign language tends to feel hardest in the first weeks, because you're not just learning new words but learning an entirely new frequency distribution. The patterns your brain built around ETAOIN don't transfer. For the history of how these patterns developed in English specifically, the history of anagrams traces how word play has always been tied to these frequency properties.

How Algorithms Use Letter Frequency

Word solvers don't just search through dictionaries randomly. The fastest solvers use frequency-weighted indexing, where high-frequency letter combinations get searched first. When you type letters into any words starting with search tool, the results you see first are typically the most common words by letter combination frequency, not alphabetical order.

When I was building AnagramSolver.io, I ran a frequency count on every letter appearing across the solver's word list. The results mapped almost exactly onto the ETAOIN table — not because I designed it that way, but because the word list is just English, and English follows these patterns reliably. The only outlier was S, which appeared less often than general text frequency would predict, because the list excludes most plurals. The core algorithm behind every word unscrambler (sorting letters into a canonical key and matching against a pre-indexed dictionary) is explained in detail in the how word solvers work post. The short version: the reason LISTEN and SILENT and ENLIST are all instant matches for the same search is because they share the sorted key "EILNST." That key is made up of letters that sit right in the middle of the frequency table. Not the rarest letters, not the most common single letters, but the consonant-vowel blend that makes English words work.

And the reason English has silent letters at all is partly a frequency problem. As pronunciation shifted over centuries while spelling stayed fixed, certain high-frequency letters like K and G ended up written but silent in combinations like KNIGHT and GNAW. The letters are there because they were once pronounced and reflected actual speech frequency. The language moved on without updating the spelling.