Ancestry Assignment to Both Sides

[ikas]

Ancestry originally assigned two of my matches to both sides. I last checked about two months ago and it was still only two matches. Checked this morning and they have now assigned 78 to both sides. Has anyone else noticed a sharp increase in assignment to both sides? If so, do you think the asssignments are largely correct? I have reservations about mine. Gedmatch suggests my parents were not related. Of the 78 now assigned to both sides, some I know are on my paternal side and some are on my maternal side. Seems unlikely to me.

[Pheno]

They appear to have done an August update and I now have an additional 5 matches assigned to both sides - none of which I know.

However 78 seems a high number to assign to both sides.

Pheno

[phil57]

Parental assignment uses a computer algorithm involving known DNA matches and a trawl of Ancestry family trees to recognise lines of descent and match them between trees and hence to families. It may be subject to the same errors as Thrulines where user trees are incorrect and I think it is still in beta. It is not derived purely from an analysis of your DNA. Autosomal tests are not phased and so the phasing is estimated.

Ancestry do say that it will be at least 90% correct for around 9 out of 10 users, but less than 90% correct for the other 1 in 10.

There is always the possibility that a new update to the algorithm has introduced an error...

Edit: Just looked and my allocation for both sides has gone from 1 (my brother) to 15. The other 14 are all below 10 cM matches and I suspect an IBC error as a result of their updated algorithm.

[ikas]

Thanks for your replies. So, likely the update in August was the cause.

Phil57. Surprised by your comment that assignments are made based on trees as well as segments. The original statement said it was wholly dependent on shared segments. Do you know if they have changed the way they make these assignments?

[LizzieL]

Ancestry originally assigned two of my matches to both sides. I last checked about two months ago and it was still only two matches. Checked this morning and they have now assigned 78 to both sides. Has anyone else noticed a sharp increase in assignment to both sides? If so, do you think the asssignments are largely correct? I have reservations about mine. Gedmatch suggests my parents were not related. Of the 78 now assigned to both sides, some I know are on my paternal side and some are on my maternal side. Seems unlikely to me.

Gedmatch also suggests my parents are not related, which according to my paper trail I would agree with. They came from different parts of the UK. However they do not need to be related for you to have a "Both Sides" match. I have one which I found myself. Ancestry only assigns it to paternal side (the closest side). The person is my third cousin on paternal side. And 5C1R on maternal side.
His paternal grandmother was the granddaughter of my 2 x great-grandparents on my father's side.
His paternal grandfather was the 2 x great grandson of my 5 x great grandparents on my mother's side.
So if this match's father had done a DNA  test, Gedmatch would also say his parents were not related to each other.

[LizzieL]

I had 1 "both sides" match that Ancestry chose before, today I have 11, the highest match is 15cm.
It has to be done on DNA and not trees, because only 5 of those 11 have a linked tree.

[phil57]

Phil57. Surprised by your comment that assignments are made based on trees as well as segments. The original statement said it was wholly dependent on shared segments. Do you know if they have changed the way they make these assignments?

Sorry, yes I've just checked and they are now saying that the assignments are being made solely from DNA, although originally they were utilising an algorithm to refine parental sides via trees. Probably, as they've increased the database and refined their processes, they've removed that aspect.

But they can't phase your DNA. Most autosomal tests including Ancestry's are unphased, which is partly the reason that lower length matches are less accurate. They are calculating the phasing of base pairs by comparison between matches who share the same string lengths. As the size of the database and number of comparisons increases, the accuracy should also increase, but the ultimate results are still governed by how they calculate and produce them.

https://support.ancestry.co.uk/s/article/SideView-Technology

I still suspect an issue with the latest update. As I say, my both sides allocations are all at low match lengths (apart from my brother) where the probability of IBC matches is relatively high.

[Pheno]

I think there is something odd about this update.

My full sibling sister has an additional 27 'both side' matches but none of these are the ones that show up for me.  I know that we won't necessarily have the same DNA but for none of these 'both side' matches not to be the same seems odd.

Pheno

[Biggles50]

I too have some both side.

The are 16 DNA matches who are in this class, highest is a mere 21cM with Zero shared match’s.

My Paternal and Maternal lines are from very different parts of England so the probability of this happening in my case in recent times is in any case very, very low.

Using the Gedmatch “Are My Parents Related” tool produces a NO.