Hangman Lab
How We Pick Our Words
A word list is only useful if you can trust it: real words, spelled correctly, suitable for a classroom, and actually fitting the pattern or theme the page promises. Here is exactly how ours are made.
Difficulty scores come from a solver
Our solver plays hangman against every word. At each turn it looks at every dictionary word that still fits the pattern and guesses the letter that appears in the most of them, the way a careful player would. The number of wrong guesses it makes is the word's score: 0 to 1 is Easy, 2 to 4 Medium, 5 or more Hard. The same solver powers the hangman solver tool.
The scores surprised us. Short words are the hardest: across our dictionary the solver averages 7.51 wrong guesses on three-letter words but 1.15 on eight-letter words, because a short pattern fits so many words. That finding shapes our hardest words and strategy guide.
Sources
- wordfreq (Robyn Speer and contributors; data licensed CC BY-SA 4.0): how common each word is, measured across subtitles, books, Wikipedia, news and web text. It sets the grade bands for spelling lists and ranks our high-frequency word lists.
- The CMU Pronouncing Dictionary (Carnegie Mellon University, BSD-style licence): how each word is pronounced. Every phonics list is checked by sound, not just spelling, so said is not on our long-a list and bread is not on our long-e list.
- SCOWL (Kevin Atkinson, spell-checker word lists, permissive licence): confirms each word is a common English word and keeps proper names out.
- The Dolch word lists (Edward W. Dolch, published 1936 and 1948, public domain): our Dolch sight word pages.
We don't use Fry's Instant Words, which are a published product. Our high-frequency lists are our own ranking from the wordfreq data.
Spelling lists
Each spelling list is generated from a rule (for example: one syllable, the letters ai, and the long a sound), filtered to words common enough for that grade, and then checked by a person. Words that only match a pattern by accident are removed: pretend is not pre plus tend, and married is a y-to-i word rather than an ie word. No two spelling pages share the same words, except where a sound or rule naturally overlaps.
Themed lists
Category lists such as Christmas, dinosaurs or US states are written by hand, then every word is checked against the dictionaries above for spelling. Factual lists (states, countries, planets, dinosaur names) are checked against standard references.
What we leave out
Every list goes through a blocklist that removes profanity, slurs, adult and drug terms, and anything violent or grim, which matters more than usual for a game called hangman. We also leave out brand names and trademarked characters. If you find a word that slipped through, tell us and it comes off the list.