Imperfect Triangulation

While examining the autosomal DNA results of Ira and Steve on GEDmatch, a curious thing turned up. Ira and Steve are second cousins, great-grandsons if Isak Pikholz of Rozdol and his wife Toby Blum.

In the course of my analysis, I saw the following on chromosome 7.

Ira and Steve have a match of just over 66 cM, as represented by the orange bars.

Below that, each has a green bar representing a match of about 13 cM that each has with Esther. But Esther's matches are not identical. Her match with Steve is 13.2114 cM, while her match with Ira is 13.3094 cM. Close but no cigar. It fails the triangulation test.

Let's look at an analogy. Ira and Steve sit down to dinner together for sixty-six minutes. Esther comes to join them for part of the meal, about thirteen minutes worth. But since Steve and Ira are together, Esther cannot spend even one second more with one than with the other. Her matches with them must be identical to the last decimal.

It is possible in theory that she matches both of them on their other sides - not the Pikholz side where they match each other, but Ira's mother and Steve's father. But it would be too much of a coincidence for those matches to be so similar.

In order to confirm the GEDmatch numbers, I did a chromosome browser for Esther.

Esther's matches with Steve and Ira are the same as theirs with her, so we see no internal inconsistency.

So just for fun, I did the same chromosome browser on FTDNA.

The bar graphs look just like the GEDmatch chromosome browser. But the numbers are slightly different. Particularly the 13 cM segments in blue. Esther matches both Steve and Ira from position 67,711,326 to 80,123,786, for a total of 13.04 cM.

GEDmatch is much easier to work with when comparing more than five kits. Some say that it is more accurate. It may be, much of the time. But not this time. This time FTDNA's numbers work and the GEDmatch numbers do not. Sometimes these tiny variations do not matter. In this case, it's the difference between possible and impossible.