The 26 letters by frequency
Peter Norvig counted the letters of the Google Books corpus: 743,842,922,321 word mentions, 97,565 distinct words and 3,563,505,777,820 letters. Because the count is over running text rather than a dictionary, a letter in a common word like "the" or "and" is counted every time the word appears, which is what makes T and H so strong. The table gives the share of all letters for each of the 26.
| Rank | Letter | Share | Rank | Letter | Share |
|---|---|---|---|---|---|
| 1 | E | 12.49% | 14 | M | 2.51% |
| 2 | T | 9.28% | 15 | F | 2.40% |
| 3 | A | 8.04% | 16 | P | 2.14% |
| 4 | O | 7.64% | 17 | G | 1.87% |
| 5 | I | 7.57% | 18 | W | 1.68% |
| 6 | N | 7.23% | 19 | Y | 1.66% |
| 7 | S | 6.51% | 20 | B | 1.48% |
| 8 | R | 6.28% | 21 | V | 1.05% |
| 9 | H | 5.05% | 22 | K | 0.54% |
| 10 | L | 4.07% | 23 | X | 0.23% |
| 11 | D | 3.82% | 24 | J | 0.16% |
| 12 | C | 3.34% | 25 | Q | 0.12% |
| 13 | U | 2.73% | 26 | Z | 0.09% |
The top three letters cover 29.8 percent of text, the top six 52.3 percent and the top ten 74.2 percent, so the remaining sixteen letters share the last 25.8 percent. The order of the first thirteen spells ETAOINSRHLDCU, which, as Norvig notes, slightly violates the typesetters' standard order ETAOIN SHRDLU. The five rarest, K, X, J, Q, Z, add up to 1.14 percent, and E on its own is about 139 times as common as Z. Vowels account for 38.5 percent of letters, which is why a five-letter guess needs two or three of them to say anything about most words.
Position changes the order
Frequency over all positions is not the same as frequency at the start or the end of a word, and Norvig breaks the count down by position as well. The page gives the position rankings as graphics rather than numbers, so only the facts it states in words are repeated here. The most common first letter is T, not E: E is about four times less frequent as a first letter than it is elsewhere, and N about three times less common at the start than overall. O is the most common second letter. E comes back as the most common last letter and is also very common in the third and fifth positions. For a five-letter game that is the key point: a letter's overall share says how often it appears somewhere in a word, but whether to put it first or last is a separate question, and the five-letter word finder counts both.
The most common letter pairs
A bigram is two letters in a row inside a word. The top 20 from the same count, as a share of all letter pairs:
| Rank | Pair | Share | Rank | Pair | Share |
|---|---|---|---|---|---|
| 1 | TH | 3.56% | 11 | TI | 1.34% |
| 2 | HE | 3.07% | 12 | ES | 1.34% |
| 3 | IN | 2.43% | 13 | OR | 1.28% |
| 4 | ER | 2.05% | 14 | TE | 1.20% |
| 5 | AN | 1.99% | 15 | OF | 1.17% |
| 6 | RE | 1.85% | 16 | ED | 1.17% |
| 7 | ON | 1.76% | 17 | IS | 1.13% |
| 8 | AT | 1.49% | 18 | IT | 1.12% |
| 9 | EN | 1.45% | 19 | AL | 1.09% |
| 10 | ND | 1.35% | 20 | AR | 1.07% |
TH and HE lead, and "the", the most common word in the corpus at 7.14 percent of all word mentions, contains both; IN, ER, AN, RE and ON are the pieces of common endings and prefixes. Pairs matter for a different reason than single letters: once a yellow or green H shows up, TH, HE, CH and SH are the places it is likely to live, and a guess that puts H after T or before E tests the pair rather than the letter alone. The crossword solver works from exactly these partial patterns, listing every word that fits a shape like ?h??e.
How long English words are
Counted by mentions, the mean word length in the corpus is 4.79 letters, and three-letter words are the most common length. Five-letter words, the length of the daily word games, are 10.7 percent of running text, while words of four letters or fewer are 55.9 percent.
| Length | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Share | 3.0% | 17.7% | 20.5% | 14.8% | 10.7% | 8.4% | 7.9% | 5.9% | 4.4% | 3.1% |
Counted by distinct words instead of mentions, the picture flips: the mean length is 7.60 letters and seven-letter words are the most common length, because long words are many but each is used rarely. That is also why letter frequencies differ between running text and a word list. In a list of five-letter words each word counts once, so the letters of the few very common short words no longer dominate. A tool that scores guesses should count letters in the candidate words, not in prose.
What this means for a first guess
In a five-letter game the first guess has no information to work with, so its job is to produce information: as many letters as possible that are either confirmed or ruled out, in positions where the result tells you the most. Three principles follow from the tables, none of them tied to any particular game's answer list.
- Five different letters. A repeated letter tests one letter twice. "FUZZY" tests four letters, three of them rare.
- Common letters first. A guess built from E, T, A, O, I, N, S, R will hit most words somewhere. The score column below is simply the sum of the running-text shares of the distinct letters in each guess.
- Positions that match position frequency. T at the start, O in the second square, E at the end: each letter goes where it most often sits, so a green is more likely and a yellow carries information about where the letter is not.
| Guess | Distinct letters | Vowels | Sum of letter shares |
|---|---|---|---|
| STONE | 5 | 2 | 43.1% |
| NOTES | 5 | 2 | 43.1% |
| TEARS | 5 | 2 | 42.6% |
| RAISE | 5 | 3 | 40.9% |
| QUEEN | 4 | 3 | 22.6% |
| FUZZY | 4 | 1 | 6.9% |
The score is a rough guide, not a ranking of real openers: it uses prose frequencies, ignores position and ignores the game's list. It is enough to show the shape of a good guess (STONE scores 43.1, FUZZY 6.9) and to see why anagrams of the same letters, such as STONE and NOTES, score the same and differ only in where they place each letter.
What the five-letter word finder does
The five-letter word finder applies these ideas to the words that are still possible after each guess, which is where they are most useful. You enter the green, yellow and gray letters from your game, and it filters a list of 4,662 common five-letter words down to the ones that fit. For the suggestion it counts, among those remaining words, how many contain each letter and how many have it in each square, and scores every candidate by its different letters, so the suggested guesses are the likely words that also test the most frequent letters. The page explains this under How the next guess is chosen. When many words are left it also offers a word to narrow them down, picked from all five-letter words for letters that about half of the remaining words contain, so that whatever colors come back, the list shrinks a lot. It does not know any game's answer list or the day's answer; it works from a public domain word list.
For the same questions outside a daily game, the word unscrambler lists every word a set of letters can make, longest first, the anagram solver finds the words that use all of them, and the word frequency counter counts how often each word, rather than each letter, appears in a text you paste in.
Frequently asked questions
What is the most common letter in English?
E, at 12.49 percent of all letters in running text, followed by T at 9.28 percent and A at 8.04 percent. The figures come from Peter Norvig's count over the Google Books corpus, about 3.6 trillion letters. The five vowels together make up 38.5 percent of all letters.
What is the rarest letter?
Z, at 0.09 percent, then Q (0.12), J (0.16), X (0.23) and K (0.54). The five together account for 1.14 percent of letters, so E alone is about 11 times as common as all of them combined.
Which letter most often starts a word?
T, according to the same count, which is also the second most common letter overall. Position changes the picture: E is about four times less frequent as a first letter than it is elsewhere in a word, and N about three times less, while E is the most common last letter. O is the most common second letter.
What is the most common two-letter combination?
TH, at 3.56 percent of all letter pairs, then HE at 3.07 percent and IN at 2.43. The word "the", the most common word in the corpus at 7.14 percent of all word mentions, contains the first two.
What is the best first guess in a five-letter word game?
There is no single best word, and the answer depends on the game's word list, which this guide does not claim to know. The principles are clear, though: five different letters, each common in five-letter words and in the position where it sits, with two or three vowels. The five-letter word finder scores guesses this way from the words that are still possible after each round.