G25 distance calculator

The calculation is Euclidean distance across all 25 dimensions: for each dimension, take the difference between your value and the reference population's, square it, add the 25 squares together and take the square root. Every population in the era you pick is scored that way and the list is sorted, closest first. There is no weighting, no model and nothing fitted — it is one arithmetic operation repeated across the panel, which is why it returns instantly and why the same input always gives the same ranking.

Era matters more than most people expect. A reference panel scoped to the Late Bronze Age and one scoped to the Modern Era contain different populations, so 'closest' means something different in each — being nearest to a modern national average says where you sit among people alive today, while being nearest to an Iron Age group says where you sit among the samples excavated from that window. Comparing across eras is the interesting part; treating one era's ranking as the answer is the common mistake.

What a small distance is not. It is not a measure of descent, and it is not a family connection: a population can sit close to you because you genuinely share ancestry with it, or because it is itself a mixture that happens to average out near your point. Reference populations are averages of individuals, so matching an average closely is not the same as matching anyone who lived — and two populations only a little further down the list are often statistically indistinguishable from the top one. Rank order is not significance, and a distance-based ranking cannot reject anything. If you need a method that can test a model and refuse it, that is qpAdm.

Everything runs in your browser: the coordinate you paste is not uploaded and nothing is stored. The reference panels are the same curated ones the paid Global25 report is computed against — the era populations are public, read-only data.

Suppose the Modern Era panel returns: Population A 0.0134, Population B 0.0151, Population C 0.0158, Population D 0.0223, Population E 0.0241. The scale first: our calibration puts a typical individual around 0.011 from their own country's average, so 0.0134 reads as 'this average describes me well'. Distances between roughly 0.02 and 0.04 read as same broad region without being this exact group, and beyond about 0.05 to 0.06 you are outside the coordinate's neighbourhood. One diagnostic before any interpretation: if the nearest population on Earth sits past roughly 0.08, suspect an unscaled or corrupted paste rather than exotic ancestry.

Now the gaps. A minus B is 0.0017 — noise territory. The honest reading of this list is a top cluster of three near-tied populations, then a step, then a second tier: the cluster is the finding, the exact winner is not. Reference entries are averages of individuals, so beating an average by a hair says nothing a re-projection of the same genome could not flip.

The cross-era read is where the tool earns its place. Run the same row against the ancient eras and watch whether the story holds: modern neighbours whose regions were populated by the ancient-era leaders form a coherent transect; a ranking that jumps regions between adjacent eras usually indicts the paste or the expectation, not the ancestry. Ancient-era distances run systematically wider than modern ones — every living person is far from every Bronze Age average — so compare ancient eras with each other, never with your modern numbers.

Vahaduo's distance mode computes exactly this arithmetic — paste sources, paste target, sort — with the sheet management left to you. The strengths and hazards mirror the admixture case: any panel you can assemble is runnable, including panels that mix scaled with unscaled rows or eras with each other, and nothing in the tool will notice. This tool's trade is curation: six fixed, era-scoped panels, the same ones the paid report uses, at the cost of bring-your-own-source freedom.

GEDmatch's Oracle answers a related question through a different object: it ranks reference populations by how closely their calculator percentages match yours — a distance between component profiles, not between genomes' coordinates. Oracle rankings are one step further from the DNA and inherit whichever 2012-era calculator ran upstream. The habits transfer (read neighbourhoods, distrust hair-thin winners); the numbers do not.

The limit shared by every distance tool, this one included: closeness is resemblance, not descent, and no ranking can test whether a population belongs in your ancestry at all. The distance list is triage — the neighbourhood a model must explain. The model itself is the admixture calculator's job, and the tested version of the model is qpAdm's.