What 75 Saved Krillion Dives Actually Asked
This site’s first 75 saved Daily Dives contain 525 prompt instances, 524 normalized unique wordings and 129,424 recorded answer entries. Those three numbers answer different questions: how many slots were played, how many wordings repeated after punctuation normalization, and how many entries appeared across all saved rosters.
A breakdown you can reproduce
Instead of assigning overlapping subjects by judgment, group each prompt by the number of entries actually saved. A “prompt instance” is one slot on one date; the same wording on another date counts again.
| Entries per prompt | Prompt instances | Share of 525 |
|---|---|---|
| Under 30 | 57 | 10.9% |
| 30–99 | 252 | 48.0% |
| 100–499 | 149 | 28.4% |
| 500 or more | 67 | 12.8% |
Download the 525-row prompt dataset (CSV) to filter by date, wording, roster size, tier count or maximum score. Every row includes its exact archive link. “Country” occurs in 73 prompt texts, but that literal count is not an exhaustive geography classification.
Constraints change the task
Some prompts add a mechanical filter: starts with a letter, ends in a suffix, excludes a letter, requires a minimum size or uses an official-name rule. A letter filter can narrow a country list to nine entries yet still leave thousands of animal names. The category and the constraint must be read together.
The saved extremes show the spread:
- Smallest roster: 9 entries, tied three ways. Countries beginning with L on August 18, countries beginning with I on September 12, and countries with nuclear weapons on September 17.
- Largest roster: 9,373 entries. The September 24 animal-without-A prompt.
- Median roster: 79 entries. Half of the 525 prompt instances had 79 or fewer saved entries and half had 79 or more.
That range is why a raw answer count needs its prompt beside it. Nine entries can describe a closed country list; thousands can describe a broad lexical filter.
What the gem counts show
Fourteen prompt instances had no One in a Krillion entry in the saved roster. At the other end, 112 prompt instances had two. Across all 525 instances, the archive stored 623 entries in the top tier.
A missing gem caps the mathematical maximum for that prompt below 100. It does not say anything about how many players attempted the question or how difficult they found it. The perfect-700 analysis separates availability from player performance.
Method and limitations
The totals are computed from the dated JSON sheets stored by this site for July 16–September 28, 2026. Punctuation and case are removed only for the repeat check. Answer entries are counted as roster records; they are not unique people, submissions or searches. The CSV contains one row per dated prompt, including tier totals and its largest saved score. To reproduce the totals, count rows, sum the entries column, and normalize the prompt column by lowercasing and replacing punctuation with spaces. These are fixed findings through September 28; later sheets are not silently added to this study.
For a different view of the same data, see the roster-size analysis. To inspect any source prompt and its original tier groups, use the Archive.