A computational genealogy · 1914

The Great War was a family feud

In August 1914 nineteen monarchs reigned in Europe. We traced eight generations of their ancestry through 1,835 people on Wikidata and measured how closely each pair was related. Then we followed the gene that Queen Victoria passed to Russia and Spain.

120/120

Every pair of monarchs whose ancestry Wikidata records well was related within eight generations. That includes all 15 pairs who ended up on opposite sides of the war.

Questions this project answers

A continent of cousins

Each square is one pair of monarchs. The darker it is, the more ancestry they share. That is measured by the kinship coefficient, which genealogists and geneticists both use. It adds up every line of descent the two have in common, not only the nearest one. Tap a square to see how the pair connects.

Allies Central Powers Neutral Shade: fifth cousins → siblings * hatched: ancestry sparse on Wikidata, so blank means unknown
φ (phi) is the chance that a gene picked at random from each person is the same copy, inherited from a shared ancestor. Siblings have φ = 1/4, first cousins 1/16, second cousins 1/64. Ancestry is eight generations deep for every monarch, so the values compare fairly. Unknown parents count as unrelated, so every value is a lower bound.

The three emperors

The war's best-known family story involves King George V, Kaiser Wilhelm II and Tsar Nicholas II. Wilhelm and Nicholas wrote to each other as "Willy" and "Nicky". The nearest lines are the ones history books give. The count shows how many more lines there were.

Look at the Kaiser and the Tsar. Their nearest line makes them second cousins once removed. But they share 118 lines of descent within eight generations, and together those make them more related than plain second cousins. Royal pedigrees fold back on themselves so often that the nearest line understates how related two people are.

How the matching works

  1. A family tree is a graph, not a tree. Each person points to two parents. When cousins marry, two branches join again lower down (pedigree collapse). Across eight generations the 19 monarchs have 9,690 ancestor slots. Wikidata fills 7,591 of them, using only 1,816 different people, so the average ancestor fills four slots.

  2. A relationship is a pair of paths. Climb from each person to a shared ancestor. The climbs must pass through two different children of that ancestor and not meet again lower down. The number of generations on each side gives the name: cousin degree is the shorter climb minus one, and "removed" is the difference between the two climbs.

  3. Kinship adds up every path. Each shared line contributes ½ for every generation it spans. The computation runs from older to younger, so 1,835 people take well under a second.

  4. Checked against history before use. Elizabeth II and Philip come out as third cousins and second cousins once removed. Victoria and Albert come out as first cousins. Isabella II of Spain comes out with an inbreeding coefficient of exactly 1/8, because her parents were uncle and niece.

Victoria's gene

Queen Victoria carried haemophilia B, a bleeding disorder on the X chromosome. A carrier mother passes it to half her children on average. Sons who inherit it bleed, and daughters who inherit it become carriers in turn. Nine of her male descendants are documented as affected, including Tsarevich Alexei and two sons of the King of Spain.

Mendel alone says each of Victoria's daughters had an even chance of being a carrier. The family's sons then changed those odds. One affected son proves his mother was a carrier. Each son who grew up healthy makes it less likely that she was. The chart puts both numbers side by side, worked out exactly with Bayes' rule over the whole line of descent.

Mendel alone after seeing which sons were affected
A son counts as healthy if he lived past 15 and isn't on the documented list. Severe haemophilia shows in childhood. Prince Maurice of Battenberg's case is disputed, so he counts as unknown. People marrying into the family are assumed not to carry it.

Limits: Wikidata stops where its records stop. Most lines still open end with women whose children aren't recorded, and the answer for real people today would come from a genetic test, not from a family tree. Living descendants are counted but never named.