A game about the world. Data you can understand.

Who runs Mapnesia

Mapnesia is an independent project created and maintained by Nicolas. It brings together geography quizzes, language recognition and anonymous game statistics to make exploring the world more engaging.

The reference catalogue is curated for the game and checked for consistency between questions, scoring and statistics. Source data can still contain errors, and language sentences need further native-speaker review. Mapnesia is not affiliated with the organizations supplying its reference data.

For questions, corrections or data enquiries, contact [email protected]. For a correction, please include the game mode, country or language, and a supporting source where possible.

Where the reference data come from

Country boundaries and shapes

Country boundaries and shapes come from the public-domain Natural Earth Vector 5.1.2 country dataset at 1:50 million scale.

Countries and capitals

The playable catalogue contains 193 UN member states. Capital names were derived from the “Capital city” field in the UN Statistics Division’s World Statistics Pocketbook 2025, Series V No. 49. Nauru has no capital question because this reference gives no legally designated capital. South Africa has three recorded capitals.

Capital coordinates and localized place names

Capital points use GeoNames WGS84 coordinates from the cities15000 extract downloaded on 28 September 2026. GeoNames IDs are retained to trace each match. GeoNames also supplies localized capital spellings where available. Its data are used under CC BY 4.0.

Flags

Flag SVGs are bundled from FlagCDN / Flagpedia and matched using ISO country codes. The upstream collection is based on Wikimedia Commons vector files. Mapnesia serves its saved assets locally; it does not fetch live flags during play.

Languages and example sentences

The catalogue contains 200 language entries and three example sentences per entry. It starts from a frozen, curated workbook extract; the original workbook is not distributed with the project. Examples combine original sentences written for that catalogue with sourced translations from the Universal Declaration of Human Rights, including the UDHR translation archive. Corrections and source references are retained. Low-confidence sentences are excluded from new sentence games.

Language–country associations

Mappings and primary-language preferences use a reviewed snapshot of Unicode CLDR 48 territory-language information plus sourced editorial overrides. Spoken-language coverage includes minority and second-language communities; it does not always mean official status. Varieties remain separate, and a parent language’s coverage is not automatically assigned to every dialect. Map → language accepts any of the saved primary languages offered for that country.

Geographic statistics

Continents, land areas, land borders and curated country centroids are stored as reference data. Miss-distance calculations use a visual centroid of the main landmass, such as metropolitan France or the contiguous United States. A centroid is a game reference point, not the nearest border or a capital.

Reference data are versioned with the application. They do not automatically follow upstream changes. Saved challenge questions and grading resources preserve the answers used when a challenge was created.

Disputed territories and country coverage

The map follows the geometry in the bundled Natural Earth dataset. Natural Earth’s default boundary policy generally reflects de facto control. Mapnesia uses the same map for every interface language; changing language does not change its boundary viewpoint.

Map coverage and game eligibility are separate. Only the 193 UN member states are game targets. Other mapped areas, including Palestine, Taiwan, Kosovo and Western Sahara, can appear on the map without being playable targets. Clicking a non-member area explains that it is outside the game; statistics maps show these areas without country statistics. The optional origin selector uses the broader map catalogue.

These choices define a consistent quiz catalogue. Displaying a boundary, name or flag does not express a position on sovereignty or recognition, and simplified map geometry may not reflect recent changes.

How statistics are calculated

Statistics count game attempts, not unique people. Eligible answers from unfinished games contribute; completed-game comparisons use only completed games.

• Accuracy. Clicks, choices and “I don’t know” are recorded separately; unanswered prompts are excluded. Including unknowns: correct ÷ (correct + wrong + unknown). Excluding unknowns: correct ÷ (correct + wrong). Game scores and geography-table accuracy include unknowns as incorrect; language tables and post-game summaries also show accuracy without them. Click accuracy uses map clicks only.
• Time and distance. Timing uses clicks and choices taking 0–60 seconds inclusive, excluding unknowns; longer answers still count in non-timing metrics. Missing eligible data are unavailable, not zero. Wrong-click distance is great-circle distance to the target country’s curated centroid. Capital-placement scores average every point’s distance to the capital (nearest of South Africa’s three); lower is better. These modes are excluded from correctness-based geography statistics.
• Knowledge scores. Chance adjustment: (accuracy − chance) ÷ (1 − chance). Geography uses 4% for 25 choices, zero for map clicks, and equally averages modes with at least three answers per country; scores can be negative. Language recognition requires five answers per language and mode, uses 4% for 25 choices or 10% for 10 choices, and clamps each mode score to 0–100 before equally averaging. It measures quiz recognition, not fluency.
• Comparisons. Rankings and comparisons require three completed games in the same mode. Rounded score percentages determine ranking; ties count ahead for top tiers and are not beaten. Distance comparisons weight games equally; only strictly larger mean distances count as beaten. Country highlights generally require five answers; detailed tables retain smaller samples. Language games are excluded from country-based geography aggregates.
• Updates. Cached statistics can lag play by about 30 seconds; click overlays refresh separately. Shared challenges use separate reads, updating on completion, manual refresh or every 30 seconds while visible.

Self-selected players, repeat attempts, different modes, question selection and small samples affect results. These are not representative population surveys or evidence that one population knows more. When reporting results, include the metric, mode, sample size and access date.

Explore the statistics or the separate language dashboard for recognition and country-language measures.

Data & privacy

Game records. No accounts or site-wide player/device identifiers are stored in the game database. A random ID groups each attempt: mode, question order, targets, answers, correctness, timing, click coordinates, viewport size, optional confidence and starting interface language. Capital placements also store the scoring location and distance.

Origin and public data. Origin is optional, self-reported, never inferred from IP, and remembered locally only when selected. Origin comparisons require three games, not three people; this cannot guarantee anonymity. Public statistics aggregate answers and show recent map-click coordinates, not players’ physical locations. Country profiles exclude challenge nicknames and credentials.

Shared challenges. Optional nicknames are public; browser credentials are challenge-scoped and hashed on the server. Completed participants see nicknames, scores, ordered outcomes and active progress, never individual choices, coordinates, origins or confidence. Templates and participant records have no automatic expiry; shared results have no minimum-participant threshold.

Browser storage. The browser stores preferences and challenge credentials; language also uses a one-year same-site cookie. Clearing storage removes preferences and challenge access, not recorded games.

Analytics. We use self-hosted Umami for traffic, performance and limited product events. Umami analytics and individual game records are kept separate and are not linked. Events contain settings, aggregate outcomes and navigation; home setup can include an origin code, “declined” or “not selected”. Nicknames, credentials, personal identifiers and search text are excluded. Challenge events also exclude challenge codes, game IDs, origins and locale.

IP addresses and logs. Temporary in-memory IP rate limits protect requests; IPs are not stored in the game database or application logs. Hosts and proxies may process connection metadata under their own policies. Rejection logs exclude submitted bodies, nicknames, answers, origins, click coordinates, credentials and query strings.

Privacy questions: [email protected].