How the words are chosen
Why your word wasn’t on the list, and why a rarer one was.
The complaint
Every word game gets the same two messages. That is a word, why didn’t it count? And, less often but more bitterly: how was anyone supposed to know that was a word?
Most games answer neither. There is a list, the list is not explained, and players are left to assume it was assembled by someone who does not play. This page is the explanation — the method, the judgement calls inside it, and the places where it is still wrong.
One thing first, because it defuses half the complaints: every real word in the grid counts. The list you have to clear is a subset. Anything else you find is an extra — it scores exactly the same, it is simply not demanded of you. Nothing real is ever rejected. The only question this page is about is which words the game is willing to require.
The obvious rule, and why it fails
The natural answer is frequency: common words are required, rare ones are extras. It is what everyone assumes is happening, and it is what we tried first.
Rather than trust it, we measured it — scoring every possible frequency cut-off against a large published set of required and bonus words from an established game in this genre. The best threshold available still gets about one word in five wrong.
Not wrong at the margins, either. The two groups do not merely overlap, they invert:
The scale is logarithmic, so those two numbers are further apart than they look: Marc is roughly 250 times more common in written English than abominably. The rarer word is the one players are required to find.
A rule that produces that result is not slightly miscalibrated. It is measuring the wrong thing.
Names are what break it
Word-frequency data counts every appearance of a string of letters, including the times it was somebody’s name. Marc, Bailey, Marcel, Wigan, Bing, Mina and Liang all look like common English words in the data. They are common people and common places.
Sort by frequency and every one of them is promoted to the required list — which is precisely the thing players complain about most in this genre. The obvious rule does not just fail; it fails by manufacturing the exact grievance it was meant to avoid.
So frequency was demoted. It still sets how hard a given day should be, which is a job it does well. It no longer decides what is fair.
What we ask instead
A different question, and a simpler one: does a dictionary list this as an ordinary lowercase word? Not is it common — is it a word at all, as opposed to a name.
That single change fixes the inversion. Abominably is in the dictionary as a word. Marc is not — it only ever appears capitalised, because it is a person. One is eligible to be required; the other can still be found, and still scores, but the game will never insist on it.
Plurals nearly broke it
Dictionaries list base forms. Oat is in there; oats is not. Team is in there; teams is not.
Applied naively, the new rule therefore rejects almost every plural in English as “not a word” — which would be the single worst failure this game could produce. So an inflected form inherits eligibility from its base: oat makes oats legitimate.
And that is a trap
Inheritance has to be narrow, and each of these was a real bug found while building it:
- Win must not make winy a word.
- Gail must not make gaily one either — that is a coincidence of spelling, not a relationship.
- Only nouns and verbs take an -s. Without that restriction, cute produced cutes and usual produced usuals, and both reached a required list.
- Only some stems take -es. Getting that wrong derived butes from but, and bates from bat.
Plurals and tenses inherit. Invented spin-offs do not. Genuinely derived words like abominably and chirpy are in the dictionary in their own right and need no help.
Why we don’t simply ban names
The obvious fix to the name problem is a blocklist. It is also unworkable, for a reason that is genuinely interesting: English surnames are English words. That is where they came from.
Baker, Miller, Mason and Walker are occupations. Hill, Wood, Stone and Dale are landscape. King, Ward, Bell, Cook, Price and Young are ordinary vocabulary. Filter surnames and you gut the language.
The dictionary test already draws the line correctly, because it asks whether the word carries a live meaning of its own. It keeps mark, grace, hope, rose, iris, jay and drake — all real words that happen also to be names. It never admits Thomas or David, which are only names.
A short hand-written list covers the handful where the dictionary is more generous than modern usage: brit is genuinely a young herring, and trey is genuinely a three-pipped card, but nobody meeting them in a grid reads them that way. Around a dozen words. They remain findable as extras.
Words we will never make you find
Slurs and profanity are never required. This is not squeamishness about the dictionary — they stay findable as extras, because refusing to recognise a real word is the failure this whole method exists to prevent. They are simply never demanded of you.
This rule was written for a reason. An early Saturday board required a word that is a slur in ordinary use, and a player could not have completed that day without finding it. A general-purpose profanity list had not caught it, because such lists are built to moderate abusive text and tend to miss clinical-looking words.
There is a wider category too. Racist, slave, corpse — unremarkable in a newspaper, a jarring thing to be required to find in a game people play over breakfast.
Where it is still wrong
A method described only by its successes is an advertisement. These are the known failures, at the time of writing:
- A handful of marginal plurals survive — anis, anas, infos — technically defensible, genuinely irritating to meet.
- Heros gets through as a nonstandard plural.
- Chai and chia are absent from the dictionary we use, so loanwords of that kind land in extras rather than the list. That is the safe direction to be wrong in, but it is still wrong.
Each of these is a judgement that could go the other way, and some of them probably will. If a word looks wrong to you in either direction, that is worth telling us — real examples are what moved every rule on this page.
The short version
Every real word scores. The required list asks one question — is this an ordinary word rather than a name — and then works quite hard to get plurals right without inventing words that do not exist. Frequency decides how hard a day is, not what is fair. Nobody is ever made to find a slur.