About

The world contains over 5,000 recognized ethnic groups distributed across 7 continents. Selecting representative samples for research, worldbuilding, educational exercises, or creative writing requires unbiased randomization - not gut-feel picks that cluster around familiar Western groups. This generator draws from a curated dataset of 180+ ethnic groups using the Fisher-Yates shuffle algorithm seeded by crypto.getRandomValues, producing uniform distribution across the selection pool. Each result includes continent, region, approximate population, and primary language family. The tool approximates global ethnic distribution; micro-ethnic subdivisions and contested classifications (e.g., caste-based vs. linguistic groupings) are simplified for practical use.

Filtering by continent narrows the pool before randomization, so output diversity depends directly on filter breadth. Population figures are order-of-magnitude estimates from ethnographic sources and should not be cited in academic work without cross-referencing primary census data. Pro tip: for character creation or demographic simulation, run multiple generations with different continent filters to avoid geographic clustering bias.

Formulas

The generator applies a two-stage pipeline: continental filtering followed by unbiased random sampling via the Fisher-Yates (Knuth) shuffle.

Pool = { e ∈ D | e.continent ∈ F }

where e = an ethnic group entry, D = full dataset of 180+ groups, and F = set of selected continent filters (default: all).

Fisher-Yates: for i = n − 1 down to 1: swap arr[i] with arr[j] where j = floor(random() × (i + 1))

Randomness source: crypto.getRandomValues provides a uniform 32-bit unsigned integer, converted to the range [0, 1) via division by 2³². This avoids the modulo bias present in naive Math.random implementations. The first k elements of the shuffled pool become the output, where k = user-specified count clamped to [1, min(50, |Pool|)].

Reference Data

Ethnic Group	Continent	Region	Est. Population	Language Family
Han Chinese	Asia	East Asia	~1.3 billion	Sino-Tibetan
Yoruba	Africa	West Africa	~45 million	Niger-Congo
Bengali	Asia	South Asia	~230 million	Indo-European
Quechua	South America	Andes	~11 million	Quechuan
Zulu	Africa	Southern Africa	~12 million	Niger-Congo (Bantu)
Basque	Europe	Western Europe	~3 million	Language Isolate
Māori	Oceania	Polynesia	~900,000	Austronesian
Tamil	Asia	South Asia	~77 million	Dravidian
Navajo	North America	Southwest USA	~400,000	Na-Dené
Sami	Europe	Northern Europe	~100,000	Uralic
Hausa	Africa	West Africa	~80 million	Afroasiatic (Chadic)
Kurdish	Asia	Middle East	~40 million	Indo-European (Iranian)
Ainu	Asia	East Asia (Japan)	~25,000	Language Isolate
Guaraní	South America	Paraguay/Brazil	~8 million	Tupian
Romani	Europe	Pan-European	~12 million	Indo-European (Indo-Aryan)
Berber (Amazigh)	Africa	North Africa	~36 million	Afroasiatic (Berber)
Inuit	North America	Arctic	~180,000	Eskimo-Aleut
Javanese	Asia	Southeast Asia	~100 million	Austronesian
Tibetan	Asia	Central Asia	~7 million	Sino-Tibetan
Maasai	Africa	East Africa	~2 million	Nilo-Saharan
Aboriginal Australians	Oceania	Australia	~800,000	Australian (diverse)
Catalan	Europe	Western Europe	~10 million	Indo-European (Romance)
Hmong	Asia	Southeast Asia	~6 million	Hmong-Mien
Mapuche	South America	Chile/Argentina	~1.7 million	Araucanian
Tuareg	Africa	Sahara	~2.5 million	Afroasiatic (Berber)
Hawaiian	Oceania	Polynesia	~300,000	Austronesian
Pashtun	Asia	Central/South Asia	~50 million	Indo-European (Iranian)
Igbo	Africa	West Africa	~45 million	Niger-Congo
Sorbian	Europe	Central Europe	~60,000	Indo-European (Slavic)
Seminole	North America	Southeast USA	~35,000	Muskogean

Frequently Asked Questions

The Fisher-Yates shuffle algorithm produces a uniformly random permutation of the filtered pool. Combined with crypto.getRandomValues() as the entropy source, each ethnic group in the active pool has exactly a 1/n probability of appearing in any given position, where n is the pool size. No group is weighted higher unless the optional "weighted by population" mode is enabled.

The dataset contains 180+ groups selected to represent global diversity across all continents and language families. Some nationally-defined groups (e.g., "American", 'Brazilian') are civic identities rather than ethnic classifications and are excluded. Sub-ethnic divisions (e.g., specific Han Chinese regional groups) are consolidated under their parent group to avoid over-representing large populations.

Uniform mode gives every ethnic group in the pool identical selection probability regardless of population size - the Ainu (~25,000 people) have the same chance as the Han Chinese (~1.3 billion). Population-weighted mode assigns probability proportional to estimated population, making larger groups appear more frequently. Use uniform for maximum diversity; use weighted for demographic realism.

The population figures are order-of-magnitude estimates and should not be cited as primary data. Ethnic classification itself is contested in anthropology - this tool uses practical groupings common in ethnographic literature. For academic work, cross-reference results with sources such as the Ethnologue, Joshua Project, or national census data from relevant countries.

Filtering narrows the pool before shuffling. A filter selecting only "Oceania" reduces the pool to approximately 12 groups, so requesting 10 results from that pool yields near-exhaustive output with limited randomization value. The tool displays the active pool size so you can gauge diversity potential. For maximum randomness, keep all continents selected or use broad multi-continent filters.

No. The Fisher-Yates shuffle produces a permutation (sampling without replacement). Each result set contains unique entries. If you request more results than the filtered pool contains, the count is automatically clamped to the pool size and a notification is displayed.