Disclaimer, data and sources

What this site is for

DNADojo is a genealogy and education product. It helps you interpret shared-DNA numbers in terms of family relationships, and it helps you look inside a raw genotype file you already own. That is the whole scope.

Not medical information

Not legal or forensic evidence

Output from this site must not be used as evidence, or as the basis for a decision, in legal proceedings, parentage or paternity determination, immigration or citizenship applications, criminal or forensic investigation, insurance, employment, credit or housing decisions, or to identify a person who has not consented to being identified. Those uses require accredited testing under a documented chain of custody. See section 4 of the Terms of Service.

The relationship dataset

Every cM figure on this site comes from one versioned dataset file. The current version is v1.0.0-simulated, built 2026-09-14, containing 33 relationship records.

How the numbers are derived

1. A published genetic map. Chromosome lengths come from Bhérer et al. 2017 refined sex-averaged genetic map (build 37). The import script reads the published files directly and records the SHA-256 of each one, so every figure here traces back to exact bytes rather than to numbers somebody retyped. The sex-averaged autosomal map totals 3,346.3 cM.
2. Simulated meiosis. For each relationship we build the pedigree and drop crossovers along each chromosome as a Poisson process at one per 100 cM (Haldane, no interference). Founders carry unique labels, so any interval two people share is identical by descent by construction — there is no matching heuristic to get wrong.
3. Reported the way a testing company reports it. Shared cM is the total length of half-identical regions, counting a fully identical region once. Segments below 7 cM are dropped, which mirrors real detection thresholds. This is why full siblings come out near 2,600 cM rather than half of the genome.
4. One disclosed rescaling. A genetic map spans only the region its markers cover, which is slightly shorter than the scale testing companies report against. Every simulated value is multiplied by ×1.0414 so that parent/child reads 3,485 cM — the number you would see on your own results. That is a single constant applied to the whole distribution, not a per-record adjustment: the shape of every range, and the relationship between relationships, is exactly what the simulation produced.
5. The displayed range is the widest and narrowest values seen across the simulated pairs that shared anything, which is how published tables are built too. Typical is the median of those pairs. How often a pair shares nothing at all is reported separately, as share any DNA.
6. Identical twins are the one record not simulated. Twins share the whole genome by definition rather than through recombination, so that figure is anchored to reported values.

How it compares to published averages

Nothing here is derived from these figures — they are an external check. Close relationships land within a few percent:

Every range on this site is conditional on the two people sharing detectable DNA. That is deliberate. A published shared-cM table can only contain pairs who matched — if two fourth cousins share nothing, neither of them is ever submitted to it — and a visitor holding a cM figure has, by definition, already matched. So the two are comparable, and the number answers the question actually being asked.

The unconditional fact is reported separately, as share any DNA on each relationship page, because for distant cousins it matters more than the range does. Two fourth cousins have roughly a 46% chance of sharing anything at all; sixth cousins almost never do. Not matching is not evidence that you are unrelated.

Our distant-cousin figures still run below published tables, and the gap widens with distance. We have not tuned that away, because the honest explanation is a bias we cannot remove: a published table contains relationships someone recognised and submitted, which skews toward larger matches, and real platforms apply matching thresholds and post-processing stricter than our flat 7 cM floor. Read a published distant-cousin average as the high end of what you might see, not the middle.

Sources

Changelog

Known limits

• Endogamy is not modelled. In populations with long histories of intermarriage, shared cM runs well above these ranges for the same stated relationship.
• Multiple relationships are not combined. If two people are related through more than one line, the true sharing is higher than any single label predicts.
• Platform differences are not modelled. Testing companies use different chips, genetic maps and matching thresholds.
• The X chromosome is not included. X-DNA has its own inheritance pattern and is out of scope for the current tools.
• Compatibility scores are not probabilities. The ranking shows relative fit across a shortlist; it is not a posterior probability and does not incorporate a prior.

Found something wrong? Corrections to the dataset are the most valuable contribution anyone can make to this site, and every change is recorded in the changelog above with its version.