How scoring works
Every prompt has an answer set. We rank each accepted singer by how well known they are, then score your answer by where it lands in that set. Rare wins; obvious sinks.
The short version
- One prompt, one answer set. The set is a hand-checked list of real singers who fit every condition.
- Rank the set. Each singer gets a fame score; the set is sorted from least known to most known.
- Score the rank. Your answer's position becomes a tier and a point value. The same singer can rank differently under a different prompt.
The six tiers
| Tier | Points | Where your answer lands |
|---|---|---|
| Hadal | 100 | A curated gem — 1–2 per prompt, chosen for the story as much as the rarity. |
| Abyssal | 80 | The rarest ~10% of the set (excluding gems). |
| Midnight | 55 | Between the 10th and 50th percentile of rarity. |
| Twilight | 35 | Between the 50th and 80th percentile. |
| Sunlight | 20 | Between the 80th and 95th percentile. |
| Surface | 10 | The most obvious 5% of the set. |
Percentiles are computed inside the prompt's answer set, not across the whole game. A singer who is a deep cut in one prompt can be a surface answer in another. That is the point.
Gems are not always the rarest
The top tier is reserved for 1–2 curated gems per prompt. Past a certain depth, rarity becomes hard to distinguish — and a completely unknown name is not fun to type. The gem is the answer we would most want a player to find, and it always scores 100 even if a handful of answers in the set are technically rarer.
This is why a result card can say Rarity #3 of 35 · Curated gem · always 100. The rank is honest; the gem is the reward.
Where the ranking comes from
The long-term fame score uses open data: Wikidata sitelinks across language editions, Wikipedia pageviews, and manual overrides for the gaps. The current prototype uses hand-estimated scores while the data pipeline is being built.
- No Spotify or streaming scraping. No audio, no previews, no platform terms to fight.
- No player votes. The rank is fixed for the dive, so the first and last player see the same scoring.
- No photos, lyrics, or bios. Only names and factual attributes.
Why not live player percentages?
Live percentages are seductive and bad for a daily game: scores would move while you play, whoever goes first would define the scale, and on day one there is no crowd to measure. We use a fixed editorial rank for the dive. Once there is traffic, live crowd stats — “x% of players said this” — will be shown next to the score, not used to calculate it.
What doesn't count
- A band name when the prompt asks for a singer.
- A singer who misses one of the conditions (era, region, scene, award).
- A misspelling we cannot resolve to an accepted alias.
- A made-up name, or a real person who is not a singer.
Rejected answers cost no points. You can retry until the timer runs out.
Corrections
Answer sets are hand-checked but never perfect. If a fair singer is missing, or an answer is wrongly accepted, email hello@getdeepercut.com with the prompt and the name. Corrections are reviewed in batches.