AncestryDNA has just updated some of their DNA-based relationship predictions!
Previously, the most likely relationship(s) for a DNA match was determined mainly by how much DNA (in centimorgans) that person shared with you. The graph offers a visual synopsis of which relationship groups (represented by colored lines) are most likely (have the highest probability on the horizontal axis) for different centimorgan amounts (vertical axis).
For example, a match of 1,800 cM is represented by the green line on the graph, and could be a half-sibling, a nibling (aunt/uncle/niece/nephew), or a grandparent/grandchild relationship. If the ages were very close, Ancestry’s algorithm might be more likely to predict the match as a half-sibling or nibling, but otherwise, the predictions were based solely on shared centimorgans.
This update allows AncestryDNA to predict more specific relationships by incorporating information from your shared matches. To extend our example, if your 1800-cM match is your nephew and their parent has also tested, Ancestry can follow the DNA trail from you to your full sibling to your nephew and label the relationship more accurately.
Not all DNA relationships are affected. Parent versus child labels won’t change, because a match of ≈3500 can only be one of the two, and they can be distinguished by age alone. The update also does not apply to matches who are more distantly related, like first cousins once removed and beyond.
Returning to the graph above, the update will affect matches in the green group (half-sibling, nibling, grandparent/grandchild) and the red group (1st cousin, great-nibling, and great-grandparent/grandchild). Overall, it appears to affect matches between about 650 cM and 2,200 cM. In theory, they should also be able to refine their half-nibling predictions, although they don’t list that in their announcement.
Could you have figured out the relationships out on your own? Probably, but this will save time and make genetic genealogy more accessible to newcomers.
Could the updated labels be wrong? Sure. Genetic genealogy has always been a game of trust-but-verify. Unfortunately, the new relationship labels are already in effect, so I am unable to compare the new predictions to the previous ones to see how they might have changed.
Have you noticed new labels in your matches? Are they more accurate? Have you found mistakes? Share your observations in the comments!
Mine are less accurate. Half-aunts on maternal and paternal sides now show as just a 2% probability of being that relationship. Has these estimates been in place at the time I was looking for unknown paternal grandparents, at only 2% I would have discounted them as close relatives.
Interesting. I noticed that they didn’t include half-aunt/uncle relationships in the update. Do you recall what the probabilities before?
The new DNA predictions says my grandmother’s half brother is 100% her nephew and he is not she was born in 1939 him in 1955 same mom! I’ve seen some other wild ones glad I’ve already marked them accordingly.
Are you looking at the probabilities or the named relationship prediction?
At the top it says predicted half brother but then below it says 100% nephew. Imagine being new to this and that shows up I think it just leads to confusion.
I agree. I’m trying to get more intel from Ancestry’s team.
I am afraid no improvement woul beat ditching Timber. Ancestry trims 35 cM from unweighted sharing of 81 cM, when I share with the same person 216 cM. She is a 1st cousin’s grandchild, so she’s my daughter’s 2nd cousin once removed. Just how is the 35 cM discounted by Timber “excess”?
The latest update wouldn’t apply to the situation you describe, as it only affects closer matches.
Timber is only applied below 90 cM, so there may be shared DNA in a pileup (excess IBD) for both you and your daughter, but you’re above the 90-cM threshold so you’re not affected. For a 2C1R, 81 cM and 46 cM are both in range.
I should have noted that it’s my daughter who shares a real 81 cM with this cousin, falsified by Timber to 46 cM. If Ancestry did this sort of thing in my college chem lab, they’d at least fail the course for altering data, and potentially face expulsion.
I wish they would use a different colour for their estimates vs the relationship that you confirm. They could also go to sixth cousin relationships being selectable as these are still in the realms of DNA. Currently have to use Distant relation, and add a note 🙁
Different colors would be a great way to distinguish estimated relationship from confirmed relationship!
Seems like a simple move looking from the outside 🙂