# Commonness pools: provenance

One file, `common.json`, holding five pools keyed by length ("3" through
"7"), each a flat, alphabetized array of the ordinary English words of that
length. They exist to answer one narrow question: *of the words ENABLE1
says are valid at this length, which ones would most people recognize?*
They are not a second dictionary. A word missing from a pool is not marked
invalid anywhere on this site; it is only sorted after the words that are
in it, under a "valid but rare" fold.

They ship as one keyed file rather than five, under the same filename
wordgamelibrary's own pool uses (`common.json`). A dictionary-derived list
trivially contains ordinary words that happen to read like a claim (this
one holds "uptime" at length 6, for instance); that is a property of
English, not something this site asserts about any word in the pool.

## Source

Copied 2026-10-09 from `wordgamelibrary.com`'s own shared dictionaries at
commit `809a416` (`public/data/dict-N.json` for N = 3..7 in that site's
repository, the `w` key of each — the frequency-ranked flat list, sorted here
for this file). `809a416` is wordgamelibrary's main as of its own most
recent word-list cleanup; a pool copied from an earlier commit would bring
back words that cleanup removed. `tests/structure.js` pins this file's
SHA-256 to `4889fe47e210567b6e070f0d64b15d7190afd3a71ec0f2e9e18a5a003c3c786c`,
the hash of exactly this `809a416` copy, so a stale re-copy fails CI rather
than shipping quietly. wordgamelibrary built these pools for its own answer lists
by taking a general English word-frequency ranking, keeping only entries that
are both common and valid in ENABLE1, and stripping plurals and a short list
of blocked base forms (`BLOCK_BASES`) that would otherwise crowd out more
useful words. Word Sieve copies the result rather than re-deriving it, so the
two sites agree on what "common" means for a shared pool of words.

| Key | Length | Words |
|---|---|---|
| `"3"` | 3 | 600 |
| `"4"` | 4 | 1,600 |
| `"5"` | 5 | 2,400 |
| `"6"` | 6 | 2,400 |
| `"7"` | 7 | 1,900 |

Lengths 2 and 8–15 have no pool. `commonness.js` treats a word at an
unpooled length as neutral (neither common nor rare) rather than guessing,
and every page that sorts by commonness says so in its own method note.

## What this changes, and what it does not

On the rack trays (home, `/word-unscrambler/`, `/word-finder/`,
`/anagram-solver/`, `/word-maker/`) a plain plural or -s form of a pooled
word also counts as common: `retains` (from `retain`) and `inlets` (from
`inlet`). The pools leave most plurals out because a five-letter puzzle
answer is rarely a plural, but a rack's best answer often is one.
The rule is spelling only: -s on any pooled word (but not -ss, so `hiss` is
not read as `his` + s), -es only where English adds it, after s, x, z, ch,
sh or o (`dashes` from `dash`, `mangoes` from `mango`, but not `tares` from
`tar`), and -ies from -y. Because it reads spelling only, it cannot tell an
everyday form from a rare one: `easies` (from `easy`) opens just as `juries`
(from `jury`) does.
`/wordle-solver/` and `/5-letter-words/` keep the pools exactly as listed.

Every word in a pool is also checked against ENABLE1: the pool only ever
*reorders or tints* ENABLE1's own answer, it never adds a word ENABLE1 lacks
and never removes one ENABLE1 has. The "Likely answers" treatment on
`/wordle-solver/` is the one exception that needs saying out loud: it is a
*ranking aid*, not a claim that a puzzle's real answer is in the pool. The
pool itself contains ordinary words that are common in general English but
rare as word-game answers (`chang`, `ariel` and `arles`, for example, all
appear in the 5-letter pool) — common-first is a sort, not an answer key,
and no page on this site says otherwise.

## License

The pool is a derived, filtered selection of ordinary English words. The
ranking it is filtered from comes from wordgamelibrary's `tools/adapter.py`,
built on Norvig's `count_1w` word-frequency table (itself drawn from the
Google Web 1T corpus); that source's own licensing terms have not been
confirmed. This file ships so a visitor can read the exact words a sort is
built from.