Every admixture calculator fits you against references — and the single most consequential fact about any calculator is when its references lived. A panel of modern populations and a panel of dated ancient samples produce different-looking percentages from the same genome, both correct, answering different questions. Most calculator arguments online are two people comparing answers to questions they did not notice were different.
The machinery itself is covered in what an admixture calculator is; this post is only about the reference axis.
Modern panels: resemblance among the living#
Fit against modern populations — as Eurogenes K13, K36 and the testing companies mostly do — and the output reads as affinity to present-day groups: how your allele frequencies or coordinates sit among people alive now. That is genuinely useful. It is also built on a moving target, because every modern population is itself a mixture of the same deep sources. When a modern-panel calculator says "32% Tuscan-like", it is describing your position relative to one blend, not naming an ingredient. Assigning ancestry to a modern country label — the most tempting reading — is exactly what this panel type cannot support: modern borders postdate the mixing by millennia.
Ancient panels: ingredients with dates#
Fit against ancient samples and the components stop being blends: a Neolithic Anatolian farmer reference is the ingredient, excavated and radiocarbon-dated. A European genome fitted against hunter-gatherer, farmer and steppe sources returns proportions of actual formation-era ancestries — the decomposition the 2012 GEDmatch calculators were groping toward with the labels of their day ("West Med", "Gedrosia") before the genomes existed to anchor them.
The date cuts the other way too: ancient panels cannot see anything recent. A person with an Irish parent and a Greek parent fits cleanly as farmer-plus-steppe-plus-forager proportions that resemble neither parent's country, because the panel's resolution ends where the deep ancestries stopped differing. Recent admixture is a modern-panel (or better, a matching-based) question.
One more axis: when the panel was frozen#
GEDmatch's ancient-flavoured calculators (Eurogenes Hunter-Gatherer vs Farmer, puntDNAL K12 Ancient) were built from the handful of ancient genomes available in 2012–2015 — before Caucasus hunter-gatherers were published, before the AADR existed. They are historically interesting but structurally incomplete: an ancestry stream absent from the panel gets silently reassigned to whatever is nearest. A current ancient panel, by contrast, draws on tens of thousands of published dated samples.
Era-scoping: the resolved version of the choice#
The clean modern form of this whole distinction is not "ancient versus modern" but era by era. Our free Global25 admixture calculators are scoped to dated windows — Late Bronze Age, Iron Age, Imperial Antiquity, Middle Ages, Early Modern, Modern — so "which sources explain me" is answered within one time slice at a time, and the distance rankings run across the same six eras. Reading your genome against 1200 BC and against the present are both legitimate; the era label keeps you honest about which question you asked. (Coordinates are the one prerequisite — here is how to get them.)
The matching table#
| Your question | Right panel |
|---|---|
| Who do I resemble among the living? | Modern — distance ranking, modern era |
| What deep ancestries formed my genome? | Ancient — era calculators |
| Was my great-grandparent from X? | Neither — recent genealogy needs matching, not admixture |
| Is a steppe source required for my genome? | Ancient + a test — qpAdm, which can reject the model without it |
The last row is the general rule in miniature: panels choose the question, and only a formal method turns an answer into a claim that can survive argument — the difference in kind is here.
Terms used here are defined in the glossary.
