Scoring
Tiers, points, and how rarity is worked out
Each prompt pays up to 100 points, so 7 prompts can pay 700 points. At 10 metres per point that is 7,000 m.
The six tiers
| Tier | Points | Rarity floor | Meaning |
|---|---|---|---|
| Fathom | 100 | 0.95 | The one in a million. Full depth for a single prompt. |
| Abyss | 64 | 0.86 | Almost nobody names this. Deep water. |
| Trench | 32 | 0.72 | Genuinely uncommon. You went looking. |
| Reef | 16 | 0.55 | A solid pull — most players never get here. |
| Shoal | 8 | 0.35 | Common, but not the first name on everyone's lips. |
| Minnow | 4 | 0.00 | The answer the whole school gave. |
How the rarity weight is derived
No tier is assigned by hand. Each answer carries a rarity weight between 0 and 1, computed from two measurements and then re-based against the other answers on the same prompt:
- How often the thing itself is encountered. An answer is resolved to its encyclopedic entity and scored by how much traffic that entity receives. This is the primary signal, because it measures familiarity with the thing rather than with the letters.
- How often the word appears in the language. Used as a cross-check and as the fallback when an answer has no entity to resolve to.
- Re-basing within the prompt. The two measurements above are absolute; what decides your tier is how rare your answer is relative to the other answers on that prompt. That is why the obvious answer for a prompt always lands shallow no matter how famous the prompt's subject is.
Why two signals and not one
Word frequency alone gets this badly wrong, and the failure is easy to see. As a phrase, “black rock” is far more common in written English than “sahara”, purely because black and rock are common words — so a frequency-only score would pay more for Black Rock than for Sahara when the prompt asks for a desert. The same failure hits homonyms: don, date and amazon are all frequent words and all obscure answers.
Measuring the entity instead fixes both. An obscure reference usually resolves to a disambiguation page, or to an article with little traffic, and both read correctly as unfamiliar.
What happens to an answer that is missing
Answer sets are the genuinely hard part of this format and they are never complete. Every answer typed — accepted or refused — is recorded, so an answer that many players give but the set does not contain can be found and added. If you have been refused for something you are confident about, see the contact page.
Full ladder
Minnow 4 · Shoal 8 · Reef 16 · Trench 32 · Abyss 64 · Fathom 100 — and a run is 7 prompts.