Ancestry longest segment "anomaly"

[PhilGee]

I have a match with “xyz” of 24cM across 1 segment and a longest length of 49cM. Looking at all my shared matches with “xyz”, I have 49cM as the longest length for each of these shared matches with “xyz”, the shared match ranges from 44cM to 22 cM. This is the case with all my matches with 1 segment shared. Triangulation would also yield the same results.

Thinking about Excess IBD/pile-ups and how Ancestry address the issue, I have just been looking at the Ancestry white paper on DNA Matching and came across "Section 4" on adjusting IBD values, including reducing match values by removing the effect of a (partial) Excess IBD/pile-up segment within the match.

From the description in the paper and using the values above, my interpretation is that a 49cM segment "match" (longest length) containing an excess IBD region could be reduced by the Timber algorithm to "Match" values such as the 44cM and 22cM quoted.

Without a definitive answer from Ancestry, this is speculation!

Phil

 

[peter]

Without a definitive answer from Ancestry, this is speculation!

I'll be speaking to the DNA experts at Ancestry towards the end of next week - they are well aware of the anomalies that we're seeing.

 

[Charles]

From the description in the paper and using the values above, my interpretation is that a 49cM segment "match" (longest length) containing an excess IBD region could be reduced by the Timber algorithm to "Match" values such as the 44cM and 22cM quoted.

Phil thank you for this interpretation of the White Paper - even though good at maths I am not a great programmer. If correct this may question the value of the longest segment being provided. It would be useful to get the clarification from Ancestry that Peter will hopefully get next week as the only real benefit of the new algorithm being presented by Ancestry was the longest segment.

 

[peter]

the only real benefit of the new algorithm being presented by Ancestry was the longest segment

As I read the White Paper the main benefits were the more accurate delineation of segments and an improvement in accuracy.

 

[JoyNor]

Good to find I am not the only one finding the anomaly of a 23cM/1 segment match with 29cM being the longest segment. I am getting pretty cheesed off with Ancestry DNA, or more precisely with the people who take a test for seemingly no good reason. Are Ancestry just manipulating figures to draw us in and do their work for them? I have pages and pages of new matches with 95% having less than 5 people on their or no trees at all. Even many private trees which, when you click to go to the tree owner, only have less than 10 names. I have better things to do than write to all these people and help them build a tree so I can find out why we share up to 50cM.

Once again Ancestry presents us with quantity over quality. And the potential ancestors it gives me via ThruLines is to say the least ridiculous. I am pretty certain who my mother was! And as I have well documented evidence gathered over 20 years for 61 of my 64 5 x Great Grandparents I don't need adopting by spurious relatives that the Ancestry algorithm finds on Disneyworld trees because we share DNA that might mean we inherited brown eyes or webbed feet!

 

[Pauline]

And the potential ancestors it gives me via ThruLines is to say the least ridiculous. I am pretty certain who my mother was! And as I have well documented evidence gathered over 20 years for 61 of my 64 5 x Great Grandparents I don't need adopting by spurious relatives that the Ancestry algorithm finds on Disneyworld trees

I don't quite follow this. That is, I agree that ThruLines can sometimes show some ridiculous potential ancestors, but they are no longer replacing existing ancestors in our trees with potential ancestors as they did initially. Potential ancestors now only appear where no ancestor is shown in the tree we have linked to our DNA results.

 

[PhilGee]

As I read the White Paper the main benefits were the more accurate delineation of segments and an improvement in accuracy.

Agreed!

....
Once again Ancestry presents us with quantity over quality. And the potential ancestors it gives me via ThruLines is to say the least ridiculous. ...

I think that everyone agrees that many trees on Ancestry are badly researched and/or copies of parts of trees stitched together. That means that ThruLines has a major flaw as it is based on "opinions" and not "facts" using possibly inaccurate data. That is not to say it is useless, provided it treated with due caution.

Others may (will) disagree, but I would like to see segment data for matches as it should be as close to "factual" as it is possible to get - either it matches or it does not - though it still needs due diligence. Whether this can be done without compromising Ancestry's data on excess IBD regions used by their algorithms is another matter.

Phil

 

[Charles]

As I read the White Paper the main benefits were the more accurate delineation of segments and an improvement in accuracy.

Peter - may be you can ask Ancestry - accuracy of what - we all know that 95% of those with low cM (6-8) are not traceable as no significant trees or false matches but so far all of my top matches have the same shared amount of DNA with me, share DNA with the same people and only a few (with very high cM) have reduced segments. My biggest gripe with Ancestry is their 20cM limit for shared DNA. I have used DNAGedcom Client to download all of my matches so that I can get all of the low cM that share DNA with those above 20cM but this is a lengthy process each time I use it and I feel guilty as it slows down Ancestry searcher for others.

 

[JoyNor]

I don't quite follow this. That is, I agree that ThruLines can sometimes show some ridiculous potential ancestors, but they are no longer replacing existing ancestors in our trees with potential ancestors as they did initially. Potential ancestors now only appear where no ancestor is shown in the tree we have linked to our DNA results.

Some of my female ancestors do not have surnames (maiden names) because whilst I might have their death, and find them on census returns the births of their children being before 1837 I cannot ascertain maiden names if the marriage is not online, or they were not actually married. I was leaving the surname blank - and assume this is what threw up lots of spurious people who were definitely not mine. I got around it eventually by giving them honorary surnames/random words or their married name in brackets or quotes. Since I did this some time back it sounds like something has changed and perhaps if I now remove those made up surname replacements Ancestry will now stop telling me names which I can detemine are just plain rubbish. A good example was being told a name and dates of a supposed ancestor - and she had died before several of the children my real ancestor was mother to to had been born.

Of course we all know that a lot of fault lies with copy and pasted trees with no research. I was seeking an 1843 burial record last week - I'll call her Mary Smith, who lived in Honley - out of 30 trees with her place of death 28 had the same misinterpretation of the place name entering it as Hornby. One person unfamiliar with this town had misread it and 27 others have copied it blindly. Not one of them had a link to the actual image from the burial record. I did a proper search, found the record and I had my Mary, died at Honley and buried in the local church, not 90 miles away on the east coast of Yorkshire! It is inevitable that Ancestry is going to spread this erroneous information even wider.

 

[Charles]

I cannot ascertain maiden names if the marriage is not online, or they were not actually married. I was leaving the surname blank - and assume this is what threw up lots of spurious people who were definitely not mine.

When I started back in 1997, I used the surname UNKNOWN when I did not know the maiden name and there are still some of them left in my tree to this day. None of my current UNKNOWNs are direct line ancestors that could be within ThruLines (i.e. further back in time or in branch trees) and I have not come across anyone who has copied an UNKNOWN into their tree though the occasional Shared DNA cousin may also have a group of UNKNOWNs.

 

[peter]

Peter - may be you can ask Ancestry - accuracy of what

I gave you my interpretation of the White Paper.

 

[Chris from Stansted]

Ancestry Timber algorithm is the reason that the longest segment shown can be significantly longer then the shared match according to the following:

https://whoareyoumadeof.com/blog/2020/08/21/what-is-longest-segment-on-ancestry-dna/

 

[Chris from Stansted]

I have also now noticed that the longest segment length reported has changed and is much smaller for some matches then originally reported. With the Timber explanation and the new values of the longest segment, this now makes sense.

 

[Charles]

Ancestry Timber algorithm is the reason that the longest segment shown can be significantly longer then the shared match according to the following:

https://whoareyoumadeof.com/blog/2020/08/21/what-is-longest-segment-on-ancestry-dna/

What the blog actually says is The size of the longest segment that we currently see on our Ancestry DNA match profile is the pre-Timber length. The total amount of shared DNA, reported in centimorgans, is the total after Timber has been applied.

Later in the blog it indicates that longer largest segments should suggest closer relationships though I am not sure as my top 4 matches are as follows

• Person A (2nd cousin once removed) 117cM over 7 segments, longest segment 35cM
• Person B (half 5th cousin) 106cM over 3 segments, longest segment 61cM
• Person C (3rd cousin once removed) 104cM over 6 segments, longest segment 32cM (this on is a bit of an oddity as the common ancestors were themselves 1st cousins)
• Person D (3rd cousin) 103cM over 5 segment, longest segment 47cM

I also match with two siblings of Person A

• Person A1 98cM across 7 segments, longest segment 31cM
• Person A2 83cM across 7 segments, longest segment 17cM

I also match with the son of Person D

• Person D1 67cM across 5 segments, longest segment 27cM

Persons A to D are from different parts of my tree and therefore not related to each other.

From this small sample, the longest segment comes from a half 5th cousin. I am certain that this is correct connection as this family is well researched with two brothers in Birmingham in late 1700s - one goes to Canada and then USA and the other remains in Birmingham before my branch goes south to London in mid 1850s. There are letters from 1840-1850 between the families in UK and USA. I have over 20 DNA cousins in this branch of my family tree.

Charles

 

[peter]

Yes, in general longer segments indicate closer relationships. The more meiosis events there are, the more likely it is that a long segment will be broken up into two shorter segments.

But as mentioned previously I'll be on a conference call with one of the DNA experts at Ancestry next week, and then everything should become much clearer. There's really no point speculating at this stage (and, by the way, it's best to stick to highly-experienced bloggers like Debbie Kennett, Blaine Bettinger, Judy Russell, and Kitty Cooper).

 

[Andrew Lloyd]

Is it just me, or has the longest segment gone. Wonder why?

 

[Charles]

Mine gone as well
Possible reasons
1) it was incorrect or did not match up with total cMs - will not return
2) it is wrong and will be corrected in a week/month or so
3) it was causing too much confusion and messages that it will come back in an form

Any news Peter

 

[Pauline]

Is it just me, or has the longest segment gone. Wonder why?

It is still there, but now you have to click on the link for 'Shared DNA: X cM across Y segments'. That brings up a pop-up and you then need to click on the down arrow alongside 'How do we estimate DNA relationships?' The longest segment information is there plus the 'unweighted' shared DNA - which I presume is the pre-Timber amount.

 

[peter]

It is displayed intermittently. It seems to be more reliable at Ancestry.com than at Ancestry.co.uk, probably because many North Americans are still asleep.

 

[jorghes]

For all the ones I checked it popped up once you clicked the shared DNA link, at the bottom of each list of facts - "Shared DNA x cM across x segments"; "unweighted shared DNA" and then "longest shared segment"