An unfamiliar message invites a guess, but a useful first step is to record what the message actually contains. The analyzer above measures ASCII letters without choosing a cipher for you. Use it for classroom puzzles, cryptograms and your own practice messages. These classical methods are educational and do not protect private information.
A repeatable first pass
- Keep an untouched copy of the puzzle and any clue about its alphabet, key or layout. Paste the message above and press Analyze text.
- Check the number of counted letters and excluded text units. If the puzzle uses numbers, dots or a special alphabet, an A–Z analysis may omit the evidence you need.
- Read the letter counts, overlapping pairs and triples, and repeated-letter patterns. Decide which observations are relevant to your hypothesis.
- If a Caesar shift is plausible, inspect all 26 rows. Select a candidate to read the whole message, then verify it with the Caesar tool.
- If a repeating Vigenère key is plausible, compare the period table with the key-length guide. Treat peaks as starting points for further tests.
Exactly what the analyzer counts
| View | Convention | What to check |
|---|---|---|
| Letters | A–Z, case ignored; no accent folding | CAFÉ contributes C, A and F; the accented letter is excluded. |
| Pairs and triples | Overlapping fragments after joining all counted letters | AB AB becomes ABAB: AB occurs twice and BA once. |
| Word patterns | Separate runs matching A–Z; first 40 distinct runs | MEET and NOON have different patterns: 1.2.2.3 and 1.2.2.1. |
| Caesar rows | Subtract encryption shifts 0–25; preserve other characters | Shift 0 is included as the unchanged-letter baseline. |
| Periods | 1–16, only while every column has at least two letters | The table uses an arithmetic mean of the individual raw ICs. |
Input is limited to 20,000 UTF-16 text units, the browser string-length convention. An emoji commonly uses two units. The limit applies before filtering, so a large block of digits cannot bypass it. A rejected input leaves no old results visible; shorten it and submit again. Edits hide old results until you analyze the new text.
Read the raw index of coincidence
For N letters with counts n, raw IC is Σ n(n − 1) ÷ N(N − 1). It measures the chance of matching letters when drawing two positions without replacement. With fewer than two letters it is undefined. Some references multiply this value by 26; the workbench deliberately reports the unscaled value and labels it raw IC.
A count you can check by hand
BANANA!Caesar shift 3; preserve punctuationEDQDQD!BANANA has A = 3, N = 2 and B = 1. The matching-pair numerator is 3 × 2 + 2 × 1 = 8; the denominator is 6 × 5 = 30. Raw IC is 0.26667. EDQDQD has exactly the same count sizes and IC, even though the letter labels changed. This deliberately short example is not representative English prose.
What the scores cannot establish
A high IC does not distinguish a simple substitution from a transposition: both preserve the count sizes. Short text, repeated slogans, names and an unusual vocabulary can all distort comparisons. A flat distribution does not prove modern encryption or a particular classical cipher.
Caesar candidates are ordered by chi-square against a rounded English frequency reference. The score is Σ (observed − expected)² ÷ expected across all 26 letters. A smaller value is a closer fit to that reference, not a confidence percentage. Read every plausible row; verify words, clues and the complete re-encoded message. The period table similarly offers no automatic key recovery.