Letter Frequency: The Count That Cracks Ciphers

Letters are not equally common. In a long stretch of English text, E appears roughly twice as often as most letters, Z almost never shows up, and that imbalance is not trivia — it is the fulcrum that lets you read a substitution cipher without ever knowing the key. Letter frequency analysis is the oldest trick in cryptanalysis, and it starts with a simple count.

What gets counted — and what does not

The Letter Frequency Counter tallies exactly the 26 letters A–Z. Digits, punctuation, and spaces are ignored, so the percentages describe the true letter distribution rather than being diluted by symbols. By default it folds case — A and a add to the same tally — because frequency analysis cares about letters, not capitalization; switch case sensitivity on when you specifically want to study capitalization patterns.

One honest limitation follows from that rule: accented letters are not counted. An é is not one of the 26 letters A–Z, so text heavy in accented characters will show a lower letter total than its length suggests. For English text this rarely matters; for other languages it changes the profile entirely.

A real count, and a warning about short texts

Counting the famous pangram "The quick brown fox jumps over the lazy dog" gives 35 letters covering all 26 of the alphabet. But look at the top of the ranking:

total: 35 letters, 26 unique

o: 4  (11.43%)     e: 3  (8.57%)
h: 2  (5.71%)      r: 2  (5.71%)
t: 2  (5.71%)      u: 2  (5.71%)

O beats E — the opposite of English's long-run profile. That is the lesson every beginner learns the hard way: short texts lie. Frequency analysis is a law of large numbers; a pangram is engineered for alphabet coverage, not natural balance, and even an ordinary sentence wanders far from the average. The profile below only emerges over paragraphs and pages of ordinary prose.

The English profile and how it cracks ciphers

Across ordinary English, the rankings are famously stable: E leads at roughly 12%, then T around 9%, A around 8%, with O, I, and N close behind — the sequence remembered as ETAOIN. In a monoalphabetic substitution cipher, where each letter is swapped for a fixed substitute, the ciphertext preserves those proportions perfectly: the most frequent ciphertext letter is the strongest candidate for E, the next for T, and so on. Combined with short-word patterns (single-letter words are almost always A or I), the count turns an opaque cipher into a solvable puzzle within a page of work. The full walkthrough is in Cracking Caesar, and the family of ciphers it defeats is surveyed in Classical Ciphers Explained.

Case sensitivity, top-N, and ties

With case sensitivity on, counts split into separate entries and sort by count descending, with alphabetical order breaking ties. Count Aa Bb A and you get:

A: 2    a: 1    b: 1    B: 1

The top-N option trims the table to the N most frequent letters — useful for eyeballing the head of the distribution without scrolling through 26 rows. The top five letters of the pangram above, for instance, come out O-E-H-R-T: a reminder that the head of a short text's distribution is a suggestion, not a verdict.

Letters versus words

A letter count answers "which characters dominate this text"; a word count answers "which vocabulary dominates it". They diverge usefully: word frequency reveals subject matter and style (see What Word Frequency Reveals), while letter frequency reveals the underlying statistical skeleton that ciphers cannot hide. If you enjoy counting things that should be random, the same statistics appear in a different costume at the byte level in Byte Entropy and Frequency.

Try it yourself

Paste a long paragraph of normal English into the Letter Frequency Counter and check the head of the table against ETAOIN; then paste a single sentence and watch the ranking scramble — that contrast is the whole lesson about sample size. Next, run a short piece of Caesar-shifted ciphertext and try to recover the plaintext using only the counts, following the method in Cracking Caesar.