Five answer tiers shown in Daily Dive #104
After completing Dive #104, we opened the first prompt’s answer review and checked each of its five tier filters. The following representative entries were displayed for its water-bird category. These are observed review results, not five answers we personally submitted.
| Answer tier | Representative answer | Displayed points | Depth at 10 m/point |
|---|---|---|---|
| Krillion | Sungrebe | 100 | 1,000 m |
| Deep Cut | Abbott’s Booby | 85 | 850 m |
| Rare | Adélie penguin | 60 | 600 m |
| Schooler | Albatross | 30 | 300 m |
| Plankton | Duck | 10 | 100 m |
The depth column applies the official 10-metres-per-point rule. The review confirms these entries and values in this dive; it does not disclose the full scoring algorithm or guarantee that future categories and revisions will remain identical. See our Dive #104 record for the responses we actually entered and the distinction between submissions and reveals.
The final dive rating uses your total score
The same results screen included a separate guide called “the bearing.” These ranges describe the whole seven-prompt dive. They are not the points awarded for one answer.
| Total points | Dive rating |
|---|---|
| 0–150 | Plankton |
| 151–250 | Schooler |
| 251–350 | Rare |
| 351–449 | Deep cut |
| 450+ | One in a krillion |
For example, our 200-point total falls within the displayed Schooler range, although individual answers in that run returned Rare, Schooler or Plankton. Keep the single-answer label, total-score rating and depth separate when comparing results.
How Krillion determines answer rarity
The official FAQ describes a process that depends on the question. For some prompts, a familiarity database measures how often a word or name appears in reading and writing, allowing mechanical scoring. Other prompts receive a combination of LLM assistance and a manual review.
The method can therefore vary by category. The FAQ explains the general approach, rather than publishing every answer’s underlying familiarity value or a formula for reproducing the full system at home.
The answer set is a separate part of that process. The creator gathers relevant datasets, such as dictionaries, encyclopedias and wiki pages, then merges and cleans them. Whether an entry is included and how an included entry is scored are different questions.
A score is not a live player-percentage report
Do not read a tier or point value as “this percentage of today’s players submitted my answer.” The published explanation describes familiarity data and curation; it does not provide a conversion from points to current submission frequency.
The game does display player-comparison information: our sampled results showed how many other players chose particular answers and a separate percentage comparison for the final score. Read those fields directly. Their presence does not establish that they are the formula behind an answer’s rarity tier.
Your own familiarity is different again. An astronomy reader may recall a particular moon immediately while another player has never heard of it. Neither reaction establishes the value Krillion will return.
Why the Krillion gem is not simply the most obscure answer
The creator explicitly says the gem is selected for more than rarity. At the obscure end of an answer set, distinctions become difficult; the selection also considers whether an answer is fun, satisfying and realistically discoverable by at least some players.
Consequently, finding a deeper cut does not establish that the game chose its gem incorrectly. Nor does memorizing an extremely obscure name guarantee the gem. The player-facing goal is still to find an appropriate, less familiar answer, but the special selection includes editorial judgment.
The FAQ also explains “too clever” as a supposedly obscure choice that is actually a familiar go-to. Treat that as a reason to examine your first unusual association, not as a reliable way to calculate a penalty or fixed point value.
Missing answers and aliases can change the interpretation
A plausible answer that is absent from the set has not demonstrated a low rarity tier. The creator acknowledges that fair answers can be missed when assembling the sources. First check the prompt’s restrictions, then distinguish “not recognized” from an accepted answer with a low score.
Autocorrect also checks aliases, so the name returned may differ from what you intended. Compare the exact prompt, recognized answer and awarded points together. If you report an omission, include a source establishing that the item belongs to the category. Our answer-matching guide explains the creator’s example.
Our practice scores measure a different skill
The Category Recall Lab awards 10 points for the first distinct match, 15 for the second and 20 for the third. Seven prompts allow up to 315 practice points. These numbers reward finding more different answers, not choosing a rarer one.
The lab uses independently authored questions and sourced reference lists. It normalizes inputs and checks listed aliases, but does not run Krillion’s autocorrect, familiarity database or gem selection. A lab match does not guarantee official acceptance, and its points cannot be converted into an official tier.
Use the exercise to widen what you can recall, then apply the boundary–candidate–subcategory routine to an actual prompt. Neither practice nor that routine promises a particular official result.