Rarity means “how few people would say it”
In this game an answer’s value has nothing to do with how rare the thing itself is. A rare answer is one few players would think of when they read the prompt. A common plant can be a rare answer; a famous endangered animal can be a common one. Everything about scoring follows from that idea, and from the way each answer list is built. The format has a long history: the TV quiz Pointless made rewarding obscure correct answers popular long before browser games did.
How the lists are built
Every hand-written prompt has a list of accepted answers, written from scratch and ranked by hand from most familiar to least familiar. The question behind each placement is: how likely is an average English-speaking player to think of this? The top of a list is what most people say within a second or two. The bottom is what a specialist, an enthusiast or someone from a particular place might know.
Lists also include common alternate names for the same answer, so the regional or older name for something sits at the same position as its main name. Duplicates are merged, so each answer has exactly one position.
Generated letter puzzles (“starts with”, “ends in”, “contains”, “exactly N letters”) don’t have their own rankings. They are built from master lists, such as every country or every element, and keep the master list’s order after filtering to the answers that fit the letter rule.
What “percentile” means
When your answer is accepted, the game works out how far down the list it sits, as a fraction from 0 (the first, most familiar answer) to 1 (the last, least familiar one). That fraction is the percentile. It is about position, not popularity counts: the lists are rankings, not survey results, so a percentile says “this answer is ranked about two-thirds of the way down”, not “this many people gave it”.
Using position rather than raw rank means short and long lists are graded on the same scale. Being tenth on a list of twenty and being fiftieth on a list of a hundred are equivalent.
The grades
Every accepted answer gets one of three grades, and each grade has a fixed distance your cat’s UFO flies:
| Grade | Position in the list | Distance |
|---|---|---|
| S (Ultra rare) | The rarest 25% (75–100% of the way down) | 100 km |
| A (Rare) | The next 35% (40–75% of the way down) | 60 km |
| B (Common) | The most common 40% (0–40% of the way down) | 30 km |
| No answer | Skipped, or time ran out | 0 km |
Distances add up as you play. Every 700 km, the distance of seven S grades, the UFO does a victory lap and starts again. Because the grades are fixed, there is no bonus for being right at the very bottom of a list: anything in the rarest quarter scores the same 100 km. The one extra is the hot streak bonus: consecutive A or S answers add +10, +20, +30 and then +40 km each, until a B grade or a miss ends the streak. It rewards staying bold, not any single answer.
The gem answer
Each prompt has one hand-picked gem answer. It is chosen to be obscure but gettable: something a few players will genuinely think of, rather than the most extreme entry on the list. Gems usually sit around four-fifths of the way down, and a gem always scores an S. Finding one is marked with a star on the reveal and in your emoji grid. Nobody is told in advance which answer it is; that’s the fun of stumbling on it.
How your answer is matched
Before your answer is looked up, it is normalised so that trivial differences don’t cost you:
- Capitalisation and accents are ignored.
- A leading “the” (or “a”/“an”) is dropped.
- Plurals are matched to singulars and the other way round.
- Alternate names on the list are accepted, such as Holland for the Netherlands or Burma for Myanmar. In the Science pack, element symbols count, so Fe is iron.
- Small typos are forgiven on longer words: one wrong letter on most words of four letters or more, and up to two on long ones. Very short answers must be exact.
Because typo matching looks for the nearest listed answer, a misspelling can occasionally be read as a different answer. The reveal always says what your answer was read as, so you’ll know.
Complete prompts and open prompts
Some prompts are complete: every valid answer is already on the list. Every country, every chemical element and every Shakespeare play are examples. On these, an answer that doesn’t match is simply wrong, and it isn’t sent anywhere.
Most prompts are open. Nobody could list every cheese, cocktail, job or round object, so the list is a well-researched starting point rather than the last word. When you give an answer to an open prompt that doesn’t match the list, an AI checker decides whether it genuinely fits:
- Your browser sends only the prompt’s id and the answer you typed. Nothing else about you or your game goes to the model.
- The checker accepts real things that fit the prompt, allowing small spelling mistakes and common alternate names. Jokes, inventions, vague words and answers that only partly fit are rejected.
- Verdicts are cached. The first time an answer is checked, the result is saved; anyone who gives the same answer later gets the verdict instantly, without another AI call.
- A correct new answer scores an automatic S, because no one had it on the list before you, and it is saved to that prompt from then on.
- A rejected answer lets you try again while you still have time.
Letter puzzles built from open lists (cities, animals, fruit and vegetables) use the checker too, but the letter rule itself is checked exactly in code first. Only answers that already satisfy the rule are judged by the AI.
Why rankings are subjective
No ranking of familiarity can be perfectly objective. What’s obvious depends on where you live, how old you are and what you do: a cheese that’s on every supermarket shelf in one country is unknown in another. The lists aim at an average English-speaking player, which inevitably means some answers will feel too high or too low to you. That is part of the genre. If an answer you consider everyday scores well, enjoy the luck; if your specialist favourite lands in the middle, it’s because more people know it than you might expect.
Fairness in the Daily
The Daily gives every player in the world the same seven prompts, one from each pack, resetting at 00:00 UTC. Everyone is graded against the same lists with the same grade boundaries, and each player gets one attempt, so results can be compared fairly in the shareable emoji grid. Everyone gets the same 30 seconds per prompt, and the timer doesn’t change how answers are graded. When a player adds a new answer to an open prompt, it is checked once and the saved verdict is applied to everyone who gives that answer afterwards, so the same answer is judged the same way all day.
Where to go next
For the full rules and controls, read How to play. To put the scoring to work, start an Unlimited run on the home page, or choose a topic from the themed packs.