Methodology

Three things are kept strictly separate: what the original site says, what our practice does, and what we estimate.

1. Facts about the original Krillion

• Sources: the official site’s own public pages (krillion.io homepage and FAQ), read on 2026-10-03 via an automated page reader.
• What we record: source URL, read date, and the specific claim it supports. Claims the source doesn’t make are not written.
• Limits: reading a page is not the same as completing a game on it. We have not played the official game as part of writing this guide, so we don’t describe undocumented mechanics, and free/paid boundaries are labeled as changeable.

2. Our question bank

• Every question states its constraints — the explicit boundary of what counts (e.g. “UN member states”, “through the 2025 ceremony”).
• Every accepted answer carries an acceptance rationale and at least one cited source with the date it was read. An answer’s sources prove it belongs to the set; they say nothing about how “rare” it is.
• Review status vocabulary: draft → source_checked → playtested → published. All launch questions are source_checked: compiled by the editor, then cross-checked answer-by-answer against the cited pages on 2026-10-03. No human playtesting has happened yet — we say so rather than inventing test logs.
• Known gaps: a few edge cases are documented in the open (e.g. Kosovo and Vatican City are included in one geography question on geographic grounds despite not being UN members — the question states this). Unresolved categories stay as drafts and are not served.
• The bank ships to your browser (the practice runs client-side), so it is readable by a determined player. That’s an accepted trade-off for a practice tool with no accounts and no leaderboard — it is not presented as cheat-proof.

3. Practice points (editorial, not “rarity”)

• Accepted answers score 1 point by default.
• Answers judged harder to recall can score 2 or 3 points — but only with a written per-answer rationale, and only as an editorial estimate. No source can prove real-world rarity, so we don’t pretend one does.
• Skips and timeouts score 0. Speed scores nothing. Each question settles once, capped at 3.
• Currently 93 of 335 answers carry tiered points with written rationale. If a future bank version has none, the site must be labeled BASIC_VALIDATION_ONLY instead of advertising a rarity experience.
• These points measure nothing official. They are not Krillion.io scores, not live player-percentage statistics, and cannot predict what the original would award you.

4. Player feedback we want (and can’t fake)

• Do the 2–3 point answers actually feel harder? That is exactly what code review cannot tell us — it needs real players. The per-answer rationale list doubles as the checklist for that future test.
• Missing answers you believe qualify: send the answer plus a source via the contact page. Accepted additions land in the next bank version (with the round-retirement notice that implies).

5. Analytics stance

Production builds run two analytics services — Plausible (self-hosted Community Edition, analytics.nitortian.com; cookieless) and Google Analytics 4 (property G-ND2ZZ4DCDC). Both receive the same six minimal events: practice starts, question resolutions (question ID + status + points), completions, replays, outbound-original clicks, feedback copies. Raw answers and personal data are excluded by construction; local and preview builds run with analytics fully disabled.