Every Global25 admixture result is only as good as the source panel it was fitted against — and every published calculator was built for a population, not for you. Most of the time that is fine: a well-curated panel for a region describes most people from that region well. Some of the time it is not, and the result is a breakdown that fits with a large distance, leans on sources that make no sense for the person, or spreads a real component across three proxies because the right one was never offered.
The Personalized Calculator is our answer to that case: an analyst builds a source panel around your own coordinates, by hand, and publishes the result as a second version of your report beside the standard one. This article explains why it is needed, how it is built, and what it changes. We are not aware of another service that offers it.
Why a standard calculator can fail one person#
A G25 admixture fit searches for the weighted average of source rows that lands closest to your row. Three properties of that search explain most disappointing results:
- A missing source is redistributed, silently. If the population that actually explains part of your ancestry is not in the panel, its share is absorbed by whichever offered sources are nearest — plausibly, and wrongly.
- Panels are built for the typical member of a group. A calculator curated for a country optimises for its most common ancestry profiles. Someone with an unusual combination — two distant regions, a minority background, recent migration — sits at the edge of what the panel was built to cover.
- Fit distance hides the problem. Offer enough sources and the distance falls regardless. A fit can look acceptable while assigning ancestry to populations that are only standing in for the absent one.
Reading those signs is covered in How to read a Global25 report. The Personalized Calculator is what to do when you see them.
How the analyst builds it#
The process is the same one we use to compose every published calculator, applied to one row:
Start from the distances. Your row's closest populations across the eras, and the individual samples nearest it in PCA space, set the neighbourhood the panel has to cover.
Compose an era-coherent panel. Sources are drawn from one era at a time, from the same curated reference of 30,386 samples in 1,535 populations the standard calculators use. Candidates that plausibly precede or constitute your ancestry are offered; ones that merely sit nearby are not.
Solve and test. The mixture is solved with the same deterministic engine as the standard calculator. Then the panel is stressed: every source is removed in turn and the fit re-solved. A source whose share collapses when a neighbour is present is dropped; a source that survives removal of its neighbours has earned its place. The search is exhaustive — every relevant era source tried, every axis of the panel varied — and it stops when the panel is a defensible local optimum, not when the distance is lowest.
Apply the bar. The published panel must be small enough to be honest — typically three to eight non-collinear, era-coherent sources — with every source earning a real share and the fit distance within the band we accept for the era. Lower distance is not better past that bar: distance falls with panel size, so the search optimises defensibility, not the number.
Write the rationale. The report states what the panel is, why each source is there and what was tried and rejected, so the result can be argued with rather than believed.
What changes in your report#
The Personalized Calculator is published as a second version of your admixture result, beside the standard one. Both stay. You switch between them in the report, each with its full source list — the populations the model used, and the ones it was offered and did not use — its fit distance and its rationale. Nothing is replaced and nothing is hidden; the standard calculator remains the reference point for comparing yourself with everyone else modelled on it, and the personalised one is the panel that describes you best.
The distance and PCA lenses, and the free Notable Matches lens against 172 published ancient individuals, are unaffected — they never depended on a calculator.
What it costs and when to add it#
The Personalized Calculator is a €10 add-on to the Global25 analysis (€29.99), chosen at checkout or added later to a finished report. The product description is at /g25#personalized. It is a good fit when:
- your standard breakdown fits with a distance outside the era's usual band;
- the sources it used are historically implausible for your background;
- a component you have reason to expect is split across several proxies or absent;
- your ancestry combines regions a single curated calculator was never built to cover.
It is not needed when the standard result fits well and reads sensibly — and the report says which case you are in. If you are unsure, order the standard analysis first: the add-on can be attached to a finished report at the same €10, and the distance and fit figures in the standard result are exactly the evidence you need to decide. Nothing about the report changes when the personalised version is added except that a second, better-fitting panel appears beside the first.
A worked example#
Take an imaginary customer with one parent from the western Balkans and one from the Levant. The standard calculator for either region fits with a distance well outside its usual band and assigns the "other" half of the ancestry to whatever sources sit nearest — for the Balkan panel, an Aegean and an Anatolian population standing in for the Levant. The percentages look tidy and describe the panel's limits, not the person.
A personalised panel offers both regions' era-coherent sources side by side: Balkan Iron Age and Roman-period populations, Levantine Bronze and Iron Age populations, and the deeper components each shares. Solved and stressed, the panel that survives typically has five or six sources, a fit distance inside the band, and a breakdown that reads as a Balkan–Levantine mixture because that is what it is. The standard result stays in the report beside it, so the difference is visible rather than asserted.
How it differs from the Calculator Explorer#
Two add-ons, two purposes. The Calculator Explorer (€10) re-solves your row against the other published curated calculators of an era — pure exploration of existing panels, with the delivered report unchanged. The Personalized Calculator builds a new panel around your row and publishes it. Explorer answers "what do the other standard panels say?"; Personalized answers "what panel actually describes me?"
What it does not do#
It does not change your coordinates, which come only from the independent Eurogenes service — see What are Global25 coordinates?. It does not add a p-value: G25 remains a coordinate fit, and the version that can reject a model is qpAdm, compared in qpAdm vs Global25. And it does not make any source population your ancestor — a source is a reference the model tests against, never a statement about who anyone's ancestors were.
If you want to see how much a panel changes an answer before buying anything, the free admixture calculator lets you fit your row against curated panels and your own pasted sources in the browser. The terms are in the glossary.
