* Ancestry Changes

Got general Family History research questions - this is the place
User avatar
gwilym'smum
Superstar
Posts: 302
Joined: 01 Feb 2016 16:28
Family Historian: V6.2
Location: South Cheshire
Contact:

Ancestry Changes

Post by gwilym'smum »

Hi
From the Lost Cousins' newsletter, next month Ancestry are updating their DNA matches. This will lead to matches less than 8cms being lost. Matches at this level are not usually used in isolation as they can give false matches but if they appear in a group could help indicate a line of inquiry. In order to stop them being deleted add the matches to a group (a special group could be created just for these matches), or make a note on them or if you have contacted that match it will be retained. Also Ancestry will change how they list segments.
Ann
Researching Mayer, Parr/Parr, Simcock, Beech and all related families
User avatar
Jane
Site Admin
Posts: 8508
Joined: 01 Nov 2002 15:00
Family Historian: V7
Location: Somerset, England
Contact:

Re: Ancestry Changes

Post by Jane »

I have 40,000 matches in the 6-8cm range, somehow I don't think I will be checking them all.
Jane
My Family History : My Photography "Knowledge is knowing that a tomato is a fruit. Wisdom is not putting it in a fruit salad."
User avatar
gwilym'smum
Superstar
Posts: 302
Joined: 01 Feb 2016 16:28
Family Historian: V6.2
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by gwilym'smum »

Burning the midnight oil Jane :lol:
Ann
Researching Mayer, Parr/Parr, Simcock, Beech and all related families
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

When Peter Calver says "matches where users share less than 8cM will be removed", I'm hoping that he means "strictly less than 8". Or in non-mathematical terms, since the current threshold is 6, I'm therefore hoping that only 6cM and 7cM matches are lost.

What I found useful was to run the DNA matches for 6-7cM and filter it for "Common Ancestors" - i.e. filter for where Ancestry's ThruLines is hinting that we have a common ancestor. I've then added all those with common ancestors into one of several groups - a couple actually suggest links beyond my brickwalls. They might be something or nothing - I discarded one such suggestion the other day since the parents in the "source" tree were having children over a period of some 50 or 60 years. (Thinks: Maybe I need a "Wot a Load of Rubbish!" group as well!)
Adrian
avatar
AnneEast
Superstar
Posts: 306
Joined: 20 Jul 2005 23:39
Family Historian: V6.2
Location: Cumbria

Re: Ancestry Changes

Post by AnneEast »

I did a run through of just 6 to 8cM matches today. I searched for surname and area for the majority of my family names. In fact there were not too many results for less common names and I left a note of the result on each of them. I didn't bother with common names ... searching for Brown in Lancashire was just silly! I do have quite a number of valuable, proven links at only 6 to 8cM but, obviously many thousands that are not so I'm reasonably happy to let them go.
Anne
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

Re the value of shared cM that you need to act on - if you want to.

Ancestry says: "... you’ll no longer see matches (or be matched to people) that share less than 8 cM with you - unless you have added a note about them, added them to a custom group or have messaged them."

Now I thought that meant that I only needed to think about maybe saving matches currently rated at 6 and 7 cM - because 6 and 7 are less than 8, but 8 isn't less than 8**. However, that's not true, I now understand, because of rounding.

The actual shared cM value is not, it seems, a genuine whole number but can have a decimal part (e.g. 7.9 cM). Ancestry round that shared cM before we see it on screen. However, the looming cut-off works on the original, UNrounded value. Hence a match at 7.9 cM will vanish unless it's in a group, etc. BUT that 7.9 cM match is currently presented as an 8 cM match.

So, if you want to keep them you need to look at your 8 cM matches, as well as 6 and 7 cM matches.

**I'm a mathematician and programmer - 8 is definitely not less than 8! :)
Adrian
User avatar
Mark1834
Megastar
Posts: 2458
Joined: 27 Oct 2017 19:33
Family Historian: V7
Location: South Cheshire, UK

Re: Ancestry Changes

Post by Mark1834 »

Memories of my old day job - a precision mismatch between spec limit and measurement, and that’s before we’ve even considered sampling error (DNA differences between full siblings) and measurement error (repeat test on fresh sample from same person)... 🙂
Mark Draper
avatar
E Wilcock
Megastar
Posts: 1181
Joined: 11 Oct 2014 07:59
Family Historian: V7
Location: London
Contact:

Re: Ancestry Changes

Post by E Wilcock »

I dont know the cm, cut off point for other DNA websites.
There are some sites to which one can upload one's raw data (which you can download from ancestry) . The best known is Gedmatch.
Mine was uploaded also to Familytree DNA because a one name study DNA site was being hosted there.
I am not too happy about losing those remote matches.
User avatar
gwilym'smum
Superstar
Posts: 302
Joined: 01 Feb 2016 16:28
Family Historian: V6.2
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by gwilym'smum »

Hi Evelyn (I think)
According to Peter Calver from Lost Cousins the date of the Ancestry change has been moved to the end of August. This will give you time to either add low matches to a group, add a note or contact the match.
A quick way to add the low matches might be to look at your main names. Then take the high matches connected to those names. Take each match and look at their shared matches. Some of the low matches will probably then show up and you can then add them to a group. If you don't have time to look at every low match at least you will capture some who may be connected to known names.
Keep safe Ann
Researching Mayer, Parr/Parr, Simcock, Beech and all related families
User avatar
Mark1834
Megastar
Posts: 2458
Joined: 27 Oct 2017 19:33
Family Historian: V7
Location: South Cheshire, UK

Re: Ancestry Changes

Post by Mark1834 »

Ancestry themselves say the change has been postponed until the end of August as a result of user feedback. The stated reason for the change, as Ann pointed out in the first posting, is to "improve the likelihood you’re actually related to very distant matches." That got me thinking about the numbers involved....

It's not difficult to calculate the theoretical number of nth cousins you might have, assuming an average number of breeding children per family. This is commonly taken as about 2.5 in western societies over the 18th-20th centuries, giving around half a million theoretical 8th cousins (plus or minus quite a bit depending on the exact value). It will, of course, be reduced significantly by shared lines of descent, common in most communities, where people generally mate with somebody a bit like themselves (socially or geographically).

From what I've read, the chances of any two 8th cousins having an identifiable DNA link are actually quite low, but what seems to be missing is any reliable estimate of how many of the tens of thousands of "distant matches" in the typical DNA profile are actually related. I suspect we just don't have the data, as relatively few relationships that far back would have been reliably documented.
Mark Draper
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

What disturbs me about chasing these low cM matches is whether the matches are Identical by Descent (i.e. the same bit of DNA has descended through the generations from a common ancestor) or whether they are Identical by Chance. There are, after all, a limited
number of letters in the DNA code and if it's only a 6cM stretch (say) the chance changes in DNA might very well result in 2 chunks being the same. It's less of a risk for the bigger matches because the length makes it less likely that something that long will repeat.

I have no idea what the chance of false positives are on these short lengths - I've been looking today but can't find anything yet.

And before anyone says that they've found a short length match that has been justified by paper - well, yes, if you generate enough random pairs of contacts, you will find a genuine relative sooner or later... Not saying the Ancestry matching is random - but how much better is it?
Adrian
User avatar
gwilym'smum
Superstar
Posts: 302
Joined: 01 Feb 2016 16:28
Family Historian: V6.2
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by gwilym'smum »

For information regarding statistics on DNA https://isogg.org/wiki/Autosomal_DNA_statistics.
When you get further back the amount on average of DNA shared with say 3xgt grand parents is 3.125%, 4x gt grandparents 1.563% and 5xgt grand parents 0.781%

What you share on average with cousins: 1st = 850cms, and 12.5%
2nd cousins = 212.5 cms with 3.125%
3rd cousins = 26.56 cms with 0.781%
4th cousins = 13.28 cms with 0.195 %
5th cousins = 3.32 cms with 0.0488%

This of course is all on average as the match with after 2nd cousins could be 0. My match with my 1st cousin is way above the average number of cms and despite being 1st cousins we do not share a match from a descendant of a gt aunt. Recombination makes many matches quite random.
The number of cousins to investigate as the generations pass grows greatly and some research suggests the following numbers.
1st cousin with 2 children per generation is 4, with 3 children is 12
2nd cousin with 2 children per generation is 16 with 3 children is 72
3rd cousins with 2 children per generation is 64 with 3 children 432
4th cousins with 2 children per generation is 256 with 3 children 2,592
Just puts things in perspective and the use of very low matches can really best be used in conjunction with larger matches.
Another consideration is the company who the test is with. Family TreeDNA may have larger matches but they use every small amount of DNA making a match having 80 cms on FTDNA but only 45cms on Ancestry and perhaps 60 on My Heritage
Just a few random thoughts to throw into the mix as you study your DNA conundrums! :?
Take care and stay safe
Ann
Researching Mayer, Parr/Parr, Simcock, Beech and all related families
avatar
Gowermick
Megastar
Posts: 1702
Joined: 13 Oct 2015 07:22
Family Historian: V7
Location: Swansea

Re: Ancestry Changes

Post by Gowermick »

gwilym'smum wrote: 30 Jul 2020 22:13 My match with my 1st cousin is way above the average number of cms and despite being 1st cousins we do not share a match from a descendant of a gt aunt. Recombination makes many matches quite random.
It suggests a bit of jiggery pokery going on, what’s the expression, non-paternal event? :D
Mike Loney

Website http://www.loney.tribalpages.com
http://www.mickloney.tribalpages.com
User avatar
Mark1834
Megastar
Posts: 2458
Joined: 27 Oct 2017 19:33
Family Historian: V7
Location: South Cheshire, UK

Re: Ancestry Changes

Post by Mark1834 »

The link posted by Ann makes for some more interesting calculations. Using their model of 2.5 breeding children per family, the average person today will have nearly three quarters of a million 5th - 8th cousins. If we multiply the individual groups by the claimed Ancestry detection rate for that group (which is generally on the high side compared with other providers), we get a total of 13,243 detectable 5th - 8th cousins.

Remember that is the maximum, assuming no shared descent, and all living relatives taking a DNA test. Ancestry don't seem to give a regional breakdown for their testing, but the claimed 15 million total is going to be very heavily concentrated in the developed western world. Let's say 5% of all living relatives are listed - probably a bit high, but we'll give them the benefit of the doubt.

5% of 13,243 is 662 detected 5th - 8th cousin relationships. My DNA profile, which appears to be entirely typical, claims 32,365 matches at the 5th - 8th cousin level. Therefore 98% of the matches are false positives.

I suspect a lot of the customer resistance to removing these is driven by "it's what the computer says, so it must be true"...
Mark Draper
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

Mark1834 wrote: 31 Jul 2020 13:33...
I suspect a lot of the customer resistance to removing these is driven by "it's what the computer says, so it must be true"...
Exactly. I'm sure that the belief is that something is better than nothing. But the something that you get is a downright lie in many cases. Does it make sense to follow a lie?

I tried to get a handle on false positives but most of the articles lost me after the introduction - I wanted to say, "Just reverse the polarity of the neutron flux and it'll be fine..."

I managed to access some Ancestry help files but they don't quite seem to match the current algorithms that they use. The files refer to various high confidence matches dropping down to:
Good Confidence
About 16-30 cM
Above 50% chance of genealogically useful ancestor

Moderate Confidence
6-16cM
No more than 15-50% chance of genealogically useful ancestor

I think "genealogically useful ancestor" was my summary as the converse of that seemed to include errors, identical by chance, and "identical by descent but so far back it's pointless".

So, on this original scale, 6 cM matches had no more than a 15% chance of being any use. Which means that a lot of them are downright dangerous... From memory, I think I remember reading somewhere that some adepts won't touch matches below 15 cM with a barge-pole, so liable to error they are. This presumes that you have bigger matches.
Adrian
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

On a slightly different aspect, when I look at my DNA matches on Ancestry, right down at the 6 & 7 cM match level, I have a number of matches who, judging by their profile pictures, are African American. (So there must be others with that ancestry who don't have profile pictures who I don't know about). Now, speaking personally, I'd love to have African American relatives - I'd be intrigued to trace where the common ancestry is and delighted to be part of the wider world. Given the way that such racial mixing occured, my delight might not be mirrored on their side, of course.

The problem for the idea, however, is the possibility that the matches are simply false positives. Or even if they are genuine Identical by Descent Matches, they might refer to some fossil DNA sequence that has come down unchanged since (making something up) my untraceable Civil War ancestor ended up on the losing side and got shipped out to Virginia as an indentured labourer... Almost as misleading / useless.
Adrian
User avatar
gwilym'smum
Superstar
Posts: 302
Joined: 01 Feb 2016 16:28
Family Historian: V6.2
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by gwilym'smum »

Mike
I haven't found any "jiggery pokery" in our line and I have researched it very thoroughly. Our match is just within the 1st cousin range but at the highest part. It is the randomness of recombination that can throw a spanner in the works1 ;)
Ann
Researching Mayer, Parr/Parr, Simcock, Beech and all related families
avatar
AnneEast
Superstar
Posts: 306
Joined: 20 Jul 2005 23:39
Family Historian: V6.2
Location: Cumbria

Re: Ancestry Changes

Post by AnneEast »

Although I am content to let the 6 and 7cM matches go, I have found some that have proved useful. I have worked through and established 'paper' connections which I would never otherwise have found and some have been 'brick wall breakers'.

The paper records establish that there is a connection BUT is that the same connection that the tiny DNA match is flagging? I guess I'll never know but it doesn't really matter, the person is successfully linked to my tree.
Anne
User avatar
trevorrix
Famous
Posts: 242
Joined: 17 Nov 2002 20:27
Family Historian: V7
Location: Suffolk, England

Re: Ancestry Changes

Post by trevorrix »

The consensus of several blog posts by experts is that around 50% of 6 and 7cM matches are false positives, which means that around 50% are postive positives.

The same blog posts calculate that on average we will lose around 50-60% of our AncestryDNA match lists.

Speculation is that the real reason for the drastic haircut is economic in that AncestryDNA tests are a one off per person, therefore no repeat business. Ancestry however have to pay for server infrastructure/cloud space forever to provide us with ongoing continuous access to our match lists.

I recommend reading these blog posts by Roberta Estes of DNAeXplained.

https://dna-explained.com/2020/07/16/an ... structions

https://dna-explained.com/2020/07/19/pl ... nealogists

https://dna-explained.com/2020/07/30/an ... rge-update
Trevor Rix
avatar
Sue036
Platinum
Posts: 39
Joined: 13 Aug 2015 17:04
Family Historian: V6.2

Re: Ancestry Changes

Post by Sue036 »

AdrianBruce wrote: 24 Jul 2020 10:17 Re the value of shared cM that you need to act on - if you want to.

Ancestry says: "... you’ll no longer see matches (or be matched to people) that share less than 8 cM with you - unless you have added a note about them, added them to a custom group or have messaged them."

Now I thought that meant that I only needed to think about maybe saving matches currently rated at 6 and 7 cM - because 6 and 7 are less than 8, but 8 isn't less than 8**. However, that's not true, I now understand, because of rounding.

The actual shared cM value is not, it seems, a genuine whole number but can have a decimal part (e.g. 7.9 cM). Ancestry round that shared cM before we see it on screen. However, the looming cut-off works on the original, UNrounded value. Hence a match at 7.9 cM will vanish unless it's in a group, etc. BUT that 7.9 cM match is currently presented as an 8 cM match.

So, if you want to keep them you need to look at your 8 cM matches, as well as 6 and 7 cM matches.

**I'm a mathematician and programmer - 8 is definitely not less than 8! :)
Hi Adrian,

I've just read this after having spent a lot of time finding potentially useful 6cM and 7cM matches (based on names in their trees) and adding them to groups. Incidentally I already have several matches in that range that I am sure are my 4th or 5th cousins, based on 3 decades of research.

The most recent message from Ancestry is more specific and says 'you’ll no longer see matches or be matched to people who share 7.9 cM or less DNA with you'.

Before I repeat all my efforts for my 8cM matches, can you please advise whether you still think I might lose some of these from my list of matches at the end of August?

Thanks,

Sue
Sue in County Durham, UK

Researching numerous families but my current focus is on LE(E/A)D(H)AMs in Burton Upon Trent prior to 1829, thereafter in Stourbridge and also from c. 1865 in Sunderland and Newcastle Upon Tyne.
User avatar
Mark1834
Megastar
Posts: 2458
Joined: 27 Oct 2017 19:33
Family Historian: V7
Location: South Cheshire, UK

Re: Ancestry Changes

Post by Mark1834 »

Trevor,

There’s quite a difference between the 3% that analysis of the numbers suggests and this claim of 50%, so it’s worth looking into in a bit more detail.

I must admit, I’m a little wary of the links you posted. They are not peer-reviewed science, but a blog by somebody who has no formal training in genetics and appears to earn their living from advising on DNA matching. Before anybody reaches for their phone to call their lawyer, I am not accusing Ms Estes of anything, just pointing out that she appears to have a vested interest, so needs to provide independent evidence. Another blog post by somebody with a science degree who shares the same opinion is not a scientific consensus – it needs to be published in the recognised academic press.

Let’s assume for the minute that the true figure is indeed 50%. That must mean one of two things – either the Ancestry tests are far better at identifying matches than they claim, or we have far more living relatives than previous studies have calculated. I can’t see any reason why Ancestry would deliberately understate the performance of their tests, and the conspiracy theories of it all being a plot to save money on processing and servers can just as easily be explained by “we spend far too much effort on recording matches that have no scientific justification”.

I’m not a geneticist either, but I’m just asking that we “follow the science” rather than rely on unsubstantiated blogs.
Mark Draper
avatar
Sue036
Platinum
Posts: 39
Joined: 13 Aug 2015 17:04
Family Historian: V6.2

Re: Ancestry Changes

Post by Sue036 »

Sue036 wrote: 05 Aug 2020 17:24
AdrianBruce wrote: 24 Jul 2020 10:17 Re the value of shared cM that you need to act on - if you want to.

Ancestry says: "... you’ll no longer see matches (or be matched to people) that share less than 8 cM with you - unless you have added a note about them, added them to a custom group or have messaged them."

Now I thought that meant that I only needed to think about maybe saving matches currently rated at 6 and 7 cM - because 6 and 7 are less than 8, but 8 isn't less than 8**. However, that's not true, I now understand, because of rounding.

The actual shared cM value is not, it seems, a genuine whole number but can have a decimal part (e.g. 7.9 cM). Ancestry round that shared cM before we see it on screen. However, the looming cut-off works on the original, UNrounded value. Hence a match at 7.9 cM will vanish unless it's in a group, etc. BUT that 7.9 cM match is currently presented as an 8 cM match.

So, if you want to keep them you need to look at your 8 cM matches, as well as 6 and 7 cM matches.

**I'm a mathematician and programmer - 8 is definitely not less than 8! :)
Hi Adrian,

I've just read this after having spent a lot of time finding potentially useful 6cM and 7cM matches (based on names in their trees) and adding them to groups. Incidentally I already have several matches in that range that I am sure are my 4th or 5th cousins, based on 3 decades of research.

The most recent message from Ancestry is more specific and says 'you’ll no longer see matches or be matched to people who share 7.9 cM or less DNA with you'.

Before I repeat all my efforts for my 8cM matches, can you please advise whether you still think I might lose some of these from my list of matches at the end of August?

Thanks,

Sue
I ought to have said, '... I already have several matches in that range that I am sure are my 4th or 5th cousins, based on 3 decades of research and, in most cases, the fact that they also match my first cousin, whose DNA I manage, and with whom they share more cM than with me.'
Sue in County Durham, UK

Researching numerous families but my current focus is on LE(E/A)D(H)AMs in Burton Upon Trent prior to 1829, thereafter in Stourbridge and also from c. 1865 in Sunderland and Newcastle Upon Tyne.
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

Sue036 wrote: 05 Aug 2020 17:24... Before I repeat all my efforts for my 8cM matches, can you please advise whether you still think I might lose some of these from my list of matches at the end of August? ...
Ancestry now shows the decimal place for low matches so you'll see 7.9cM if you look now where you previously saw 8cM. So any 8cM matches that you see now will be 8.0 or above so will remain in the system.

You need to recheck - depending on when you did your previous checks - because some of the 7.9cM and below matches may have been 8cM when you last looked.

Hope this makes sense.
Adrian
User avatar
AdrianBruce
Megastar
Posts: 2090
Joined: 09 Aug 2003 21:02
Family Historian: V7
Location: South Cheshire
Contact:

Re: Ancestry Changes

Post by AdrianBruce »

Mark1834 wrote: 05 Aug 2020 18:54 ... There’s quite a difference between the 3% that analysis of the numbers suggests and this claim of 50%, so it’s worth looking into in a bit more detail. ...
I have tried, as I said above, to find some simply understandable numbers on the rate of false positives - those that are Identical By Chance (IBC) - and failed. Maybe it's like Quantum Theory and my question makes no sense but I'd sure like to know.

The claimed rate of 50% false positives is, in my view, a misunderstanding. It comes from, I believe, serious statistical comparisons between the DNA matches that a child has versus the DNA matches that their parents have. Any genuine Identical By Descent (IBD) match must have come through one or t'other of their parents (ignoring Non-Parental Events). Thus the child should have no DNA matches that their parents don't have (between them). In fact, it appears that looking at these short segments, the typical child has - if I understand it correctly - a number of DNA matches that their parents don't have. These are Identical By Chance (IBC) and are the result of normal DNA changes between parent and child that, purely by chance, create a match where there was none in the previous generation.

The analyses give a rate that 50% (roughly, vaguely) of the matches are IBC (because they're not matches that the parents have) and 50% IBD (because they're matches that the parents do have).

But 50% is not the full story is it? On the previous generation, the transition from grand-parents to parents will have introduced its own normal DNA changes between generations that, purely by chance, create a match where there was none before. Repeat ad nauseum.

As you go up the tree, into the past, towards the common ancestors, I believe that the rate of generation of false positives will reduce because the relevant segments will be getting longer and therefore harder to match by chance change. This is, however, a bit of guesswork on my part.

The closest I've got to the sort of simple numbers I crave, is on https://genealogy.stackexchange.com/que ... s-evidence where cleaverkin states:
Diahan Southard (among others) suggests that the probabilities for a 7 cM match are about 80% IBC, 20% IBD.
If anyone can set me right in a simple clear manner, please do so...
Adrian
User avatar
Mark1834
Megastar
Posts: 2458
Joined: 27 Oct 2017 19:33
Family Historian: V7
Location: South Cheshire, UK

Re: Ancestry Changes

Post by Mark1834 »

I think it is exactly like Quantum Theory - it is both false and positive at the same time, and we can’t know until we look at it, and by then the cat is probably dead anyway... :D
Mark Draper
Post Reply