Phased family matches

So it’s DNA Sunday here at The Legal Genealogist, and finally there has been enough time and enough information available to really play with a new tool from Family Tree DNA.

It’s now possible, by adding even a skeletal family tree to a person tested, and then linking known cousins who’ve tested to the tree, to have the system automatically associate matches with one or both sides of the family — maternal and paternal.

Family Tree DNA calls this Phased Family Matching, and it’s basically dividing your matches into buckets, putting people into the buckets when those people match you and a specific relative with enough shared DNA on the same DNA segment.

It’ll use a number of different relatives to sort the matches: a parent, grandparent, great grandparent or 2nd great grandparent; aunts, uncles, grand aunts and uncles, and great-grand aunts and uncles; first, second or third cousins; and half siblings.

And the initial results I’ve seen from some initial testing shows this tool has an awful lot of promise.

For my paternal half brother, who has matches on both sides of his family, the tool was able to put roughly half of his matches into either the paternal or the maternal bucket. It seems to be correctly putting all of his obviously German matches into the paternal bucket of our German-born father and all of his obviously Swedish matches into the maternal bucket of his mother, whose grandparents were all born in Sweden.

In my match list, only the folks I share with my paternal half-brother are in my paternal bucket — and there’s only one person on my match list who unexpectedly ends up in both buckets (someone other than my siblings!), so I need to see where he matches my mother’s side as well as my father’s side.

You can play with this feature a little by adding and subtracting known cousins in known branches — and by limiting the branches you link to at any given time.

For one of my Shew cousins, for example, adding everybody in my direct line who was a known cousin and a match (three of my siblings, me, all of my aunts and uncles, my first cousin and my grandmother’s first cousin) produced 31 likely cousins in the line we share, his paternal line. Adding in a first cousin to my aunts and uncles, who’d be another third cousin to this cousin, brought the number up to 39 paternal line cousins.

Perhaps the best immediate use for my family was with one of my Cottrell cousins who is half-Ashkenazi Jewish. The endogamy on the Jewish side of his ancestry makes it very hard to separate out those few matches who might be in our shared line because there are simply so many matches that he has. (As of this writing, he has 6911 matches in the Family Tree DNA database.)

So I added in just enough to his family tree to start identifying only the matches that might be on his mother’s mother’s side — the side we share.

And that took us down to 186 matches we can now try to work with.

Now we could have gotten there, eventually, or at least to somewhere close to there, by working through all of those people who are in common with the known family members. But this is so much easier and so much faster. Enter the family tree information, link those known family members who have tested, and then open the match list. Watch for the “calculating family relationships” line to disappear and voila.

You have it right there in front of you.

The tool is a good one, and very helpful. But, like all tools, it’s not — it can’t be — perfect. It’s got a couple of odd glitches, like the need to include two parents for the person tested even if you’re only going to link matches to one side of that person’s family. If you don’t put in two parents, the system just won’t generate any paternal-versus-maternal side matches at all.

It can’t divide people into two buckets if you only have known matches on one side. If you only have known matches on the maternal side, it’ll only associate matches on that side. If you only have known matches on the paternal side, it’ll only associate matches on that side.

And it isn’t going to pick up every single person who’s tied to a particular side of the family, and sometimes it isn’t easy to figure out why not.

For example, I added a family tree for one of my Robertson cousins linking him back to his (and my) second great grandparents, Gustavus and Isabella Robertson. I then added in his matches to my immediate family — four of my siblings (his third cousins), four of my aunts and uncles (his seconds cousins once removed), and one of my first cousins (another third cousin). (Alas, he’s one of those third cousins who — sigh — doesn’t match me.)

After adding those people to the tree, he was matched to 94 people as paternal side matches. Interestingly, two of his closest matches overall are two people who are first cousins to each other and second cousins once removed to this cousin. The algorithm correctly detected one of them as a paternal side cousin. It didn’t identify the other one as belonging to that side. (We’ll call this known second cousin once removed A-2C1R for short.)

I added in one more known cousin — this one a first cousin to my aunts and uncles, so another second cousin once removed to this Robertson cousin. That bumped the numbers up to 103 paternal side matches — but A-2C1R wasn’t among them. This despite the fact that A-2C1R is in common with each of those who are close matches in the paternal side list.

Taking it back another generation, to add in a known third cousin once removed, didn’t change the numbers at all. It’s not supposed to, since it’s just a half-degree further removed than the third cousin level that’s supposed to trigger the matching, but hey… I can hope anyway!

It’s a very nice tool, very helpful, not perfect, but very very nice.

Kudos to Family Tree DNA for adding this one.