Which letters eliminate the most words?
Original LimiLetter analysis of 9,586 distinct words shows which letters appear in—and can eliminate—the most candidate words.
What we measured
We combined the six active category files documented by the game engine—English, books, technology, fall, food, and nature—then removed duplicate words. For every letter, we counted how many distinct words contain it at least once. Repeated letters count once because one Limi eliminates a word whether the chosen letter appears once or several times.
The result covers 9,586 distinct four-to-nine-letter entries. Category files contain 7,000 english, 2,409 books, 598 technology, 117 fall, 148 food, 159 nature entries before cross-category deduplication.
| Rank | Letter | Words | Share |
|---|---|---|---|
| 1 | E | 5,457 | 56.9% |
| 2 | S | 4,298 | 44.8% |
| 3 | A | 4,204 | 43.9% |
| 4 | R | 3,786 | 39.5% |
| 5 | I | 3,619 | 37.8% |
| 6 | O | 3,024 | 31.5% |
| 7 | N | 3,008 | 31.4% |
| 8 | T | 2,916 | 30.4% |
| 9 | L | 2,797 | 29.2% |
| 10 | D | 2,125 | 22.2% |
| 11 | U | 1,994 | 20.8% |
| 12 | C | 1,844 | 19.2% |
| 13 | P | 1,594 | 16.6% |
| 14 | G | 1,568 | 16.4% |
| 15 | M | 1,516 | 15.8% |
| 16 | H | 1,394 | 14.5% |
| 17 | B | 1,173 | 12.2% |
| 18 | Y | 1,119 | 11.7% |
| 19 | K | 845 | 8.8% |
| 20 | F | 803 | 8.4% |
| 21 | W | 766 | 8.0% |
| 22 | V | 566 | 5.9% |
| 23 | Z | 214 | 2.2% |
| 24 | X | 196 | 2.0% |
| 25 | J | 160 | 1.7% |
| 26 | Q | 97 | 1.0% |
Why the most common letter is not automatically the best move
LimiLetter protects the target. Any letter inside the target is unavailable, even if that letter would eliminate thousands of other words. The live board is also ranked: eliminating a rare letter can be decisive when it removes the few candidates standing immediately above the target.
The Rust engine therefore solves a more specific problem than global frequency. It finds the smallest set of target-safe letters that hits every higher-ranked decoy. This report describes the landscape; an optimal route describes one puzzle.
Reproducible methodology
This page is generated from the version-controlled word files used by hangman-core. The build accepts lowercase ASCII entries from four to nine letters, deduplicates them across the six active sets, and counts presence—not occurrences—of each letter. Re-running npm run build:content regenerates the table from the current files.