Ancestry ProTools Peculiarities

[familydar]

Hammering the DNA aspects of this having taken out a month's sub.

Several of my matches have other matches in common, and ProTools tells me how much DNA they share with one another.  What is weird is that the predicted relationship often varies, depending on which of the pair of shared matches I am looking at.

It's not a one-off, I've come across several, but for instance I have a match with a user I'll call SR and I am confident about where they sit in my tree.  They are female.  Amongst SR's shared matches is KR.  He is male.  SR shares 1739cM with him and he is predicted to be her nephew.  But when I look at KR's shared matches, the match size between him and SR is the same (phew!), but SR is predicted to be his GRANDMOTHER or aunt.

SR also has a match with CR, 3471cM and he is predicted to be her son.  Looking at CR, SR is predicted to be his mother, for which I'm thankful.  KR shares 3470cM with CR.  KR is predicted to be CR's son.  So because the three have tested, I can deduce that CR is in the middle, SR is his mother and KR his son.  If I can work this out just from the aspects of DNA that Ancestry are prepared to let me see, why do Ancestry, who surely can see so much more, muddy the waters?

Jane :-)

[Glen in Tinsel Kni]

The shared cM will be the same whichever end of the telescope you look through but a lot of the relationship predictions are never corrected manually by testers so you can have a protools suggestion that's based on a manual correction and another based on a relationship prediction. A half cousin and a once removed cousin can have very similar amounts for example but the issue of people thinking the child of a 1C is their 2c can result in wrong 'correction'. Many turn a blind eye to half relatives too whereas some of us have nothing but half sibs and half cousins to work with.   

Quite a few of my matches seem not to have accounts but have probably tested at the request of the their child/grandchild, the child/grandchild has then assigned the kit to their account. As the dna is effectively attached to the wrong person (often the wrong generation), that throws things off in protools and also in Thrulines.

I have many examples of seeming male usernames with female icons and vice versa, it's stupidly easy to overlook it too and as a result building the trees can be a nightmare until the penny drops.

[familydar]

Thanks for such a clear explanation Glen

Jane :-)

[David Nicoll]

Glen,

   Hi, it’s not even as complex as that, or more complex 😁.
   They are making predictions purely based on cM matching.
   Have a look at the cM relationship charts.
   It’s 3470 cM is on at the lower end, but the only relationship that could give this is parent - child.
   1739cM can be generated two ways
   Either half of half - two steps from grandparents.
   Or Half of parents sibling - aunt/uncle.
   The only thing Pro-Tools is giving is triangulation, between unknown relatives.

   I hope this is understandable.

[Glen in Tinsel Kni]

It does flag up a very important point in that taking the protools results of one person in comparison to many others can be somewhat misleading so checking them in each direction can be beneficial. That's how I eventually worked out the kits were assigned to the child/grandchild in a couple of cases but I still kick myself for ignoring the obvious male name and pink icon staring me in the face.  We live and learn eh?

I don't run protools constantly as it's not something I feel I can justify as an individual researcher (I probably see a 20+ cM match once every two or three months), but a a month here and there seems to work. I have a group I've called 'protools candidates' so I know which I want to check when the time comes. If I ran it constantly I'd probably burn out and not use it to the full potential but it probably means I haven't developed the techniques or gained the knowledge yo make best use of it yet as I've only had two runs with it. 

[Glen in Tinsel Kni]

Glen,

   Hi, it’s not even as complex as that, or more complex 😁.
   They are making predictions purely based on cM matching.
   Have a look at the cM relationship charts.
   It’s 3470 cM is on at the lower end, but the only relationship that could give this is parent - child.
   1739cM can be generated two ways
   Either half of half - two steps from grandparents.
   Or Half of parents sibling - aunt/uncle.
   The only thing Pro-Tools is giving is triangulation, between unknown relatives.

   I hope this is understandable.

I get what you are saying but it also relies on self reported ages, in the case of that child assigning their parents kit to their account for example it could be the dna of Mary who is 60 assigned as the dna of her son John who is 30.  One such example is a first cousin of my two closest mystery matches mentioned in the next paragraph, the 1C assigned their mothers' result to their account.

I had two matches (pre pro tools), that claimed to be full siblings to each other but had a 25 year age gap (for context one was born in 1935, one in 1960 and I was born in 1966),  the relationship predictions to me put them in different generations and whilst I did ask what cM they shared the surviving match wouldn't provide the answer. I had some reservations about them being sibs and considered a mother/daughter relationship as more likely but protools proved the full sib link.

I just feel it would have been easier and quicker for the match to answer the question but at least I didn't buy the month to answer that one question alone. I still don't know how we connect though so the standard relationship predictions for them are reversed in terms of probabilities as I share 168cM with one and 199cM with the other. For now I just have to be content with knowing they are safely in the full sib range and luckily not in the overlap of full/half. 

[David Nicoll]

Glen,

     Apologies, I wasn’t trying to be pointed, I was just trying to say to Jane that there was a straightforward explanation for her issue.
     That is before you get in to people recording their parent’s DNA as their own or similar.
     With regards your example it could be just that and a child being protective and a fake age.
     People have different reasons for doing DNA but from my point of view, the more the merrier.
     I have had a few cases of a group of no tree people linking to someone with a tree that gets me through a brick wall

David

[Glen in Tinsel Kni]

Many 'same generation' relatives to me are 30 years older than myself which puts them well into their 80's. I suspect a fair few are either not internet users and/or dna savvy and whilst Ancestry claim a kit can only be assigned to the account of the tester it's easy to overcome from an existing account without setting up a new one for someone who will never login. I'd much rather see a dormant account with accurate age/gender where the test is managed by someone else but not everyone is aware they can transfer management or share their result, if it was more widely known and used the need for protools would be greatly reduced but Ancestry wouldn't have the huge income stream it must generate.