A New Cluster on the Block

Posted on by

AncestryDNA has rolled out an “auto” Cluster program. I tried it and got 8 Clusters, ranging from 3 to 9 Matches in each one. A total of 40 of my 60 Matches above 65cM. The other 20 Matches were not included because they didn’t form a Cluster of at least 3 Matches. I know the Common Ancestors for each of the 40 Matches and the program clustered them 100% correctly. I’d give AncestryDNA an A+ for this new program. I’m impressed and anxious to have the ability to adjust the cM ranges downward to get more Clusters.

Some additional input on auto-Clustering.

It began in late 2018, with Genetic Affairs (by EJ Blom), and soon we also had Shared Clustering (by Jonathan Brecher) and DNAGedcom Client (by Rob Warthen). I tried all three. I had already done segment Triangulation on all my Matches at FamilyTreeDNA, and I worked with Johathan Brecher and we Clustered those same Matches. There was over 90% concurrence between the hundreds of Clusters and the hundreds of Triangulated Groups. Not enough to say the two processes were equivalent (they are not), but certainly this analysis showed a strong tendency of Clusters to point to a Common Ancestor between me and all the Matches in each Cluster. A very strong clue in each case.

I then Clustered all of my Matches at AncestryDNA – down to about 18cM. Many of the Clusters had a Common Ancestor consensus (easily seen in the Match Notes I had previously entered – many from ThruLines). So, I imputed that Common Ancestor to the rest of the Matches in each Cluster. I used Ahnentafel numbers to represent my Ancestors and developed a tagging code: e.g. #A0020. The #A means a confirmed Common Ancestor with a Match, and 20 is Ahnentafel for William MITCHELL 1824-1895. This code is the first thing in the Notes field. When I impute a Common Ancestor to a Match from a Cluster consensus, I use #L0020 – which means the Match is highly Likely to have that Common Ancestor with me. With a #A or a #L, I tagged almost all my Ancestry Matches over 20cM and many below that. This was in the 2019-21 time frame.

Recently, with ProTools, I’ve been able to determine how many more Matches fit into my Tree – and thus our Common Ancestor. For well over 90% of all these new Match cousins, the #L tag turned out to be correct – I only needed to change the L to A.

Bottom line 1: I am a big fan of Clustering at AncestryDNA and really look forward to expanding the coverage to more Matches.

Bottom line 2: Use ProTools with Clustered Matches to really nail down Common Ancestors to Matches.

[22DI] Segment-ology: A New Cluster on the Block by Jim Bartlett 20250725

10 thoughts on “A New Cluster on the Block”

  1. Hi Jim,

    Although I initially resisted its use for quite some time, your coding system has indeed been very helpful!

    I thought I would share how I have adapted your coding system for my use.

    Here are some example tags:

    #A0028/3C-101
    As with your system, this specifies a common ancestral couple consisting of A28 & A29.  “/3C” denotes the relationship, and “-101” denotes the cM shared. The tag appears at the beginning of my notes like this:
    #A0028/3C-101 cM across 3 segments

    #A0008H/2C1R-100
    In this case, there is not a common ancestral couple but only a common ancestor – so a half relationship.  Here it was A8 since the match was from A8 & his second wife.

    #A0022/L-100
    Here I do not know the common ancestor but have imputed one based on the clusters or shared match gestalt.  So the relationship is replaced with “L”

    #A0028-30D/2C1R-206
    In the case of double cousins I add both relationships, and also the designator “D”.  So this is a double cousin 2C1R with two common ancestor couples, the first couple is A28 (representing A28 & A29) and the other couple is A30 (representing A30+A31).

    A0010X/2C-180
    The “X” designator indicates that this tester is a direct descendant of another tester (Typically child or grandchild).  I can then quickly edit the database to remove these testers before clustering.

    When coding, I have a copy of my color coded DNA Painter Ancestral Tree open on the desktop for quick reference, as I do not have all the A numbers or even all of the ancestors memorized :).

    JM

    Liked by 1 person

    Reply ↓

  2. Pingback: Friday’s Family History Finds | Empty Branches on the Family Tree

  3. Yes! I played with the auto clusters and they seem to be spot on! I am really excited about going through and verifying/ discovering with my matches and through lines. One of the best working tools. I agree, A+ on the tool!

    Now I will see if one of my ancestors really was my ancestor 🙂 …

    Ellen

    Liked by 1 person

    Reply ↓

  4. Hi Jim – have commented before. I agree it might be useful but I don’t use ProTools in Ancestry. Ancestry must be using AI to do this because it has wrecked my tree! Will have to complain to them because of the awful mistakes it is making. Putting in Ancestors as potential Father, Potential Mother, etc. That might be good but NOT If it is JUST WRONG. I have personally read the entire Volume 4 of the Maryland Archives Online. So I know it is wrong. Just sent my extended family members an invitation to follow my tree all the way back to 1634. BUT, in the LifeStory tab, it said that my older brother was unmarried and had NO children! He has been deceased since 2015, but if his wife AND children had looked, they would have thought I was crazy – that I didn’t know my own brother. We (brother and I) got married on the same day, in the same ceremony back in 1966! The AI must be what is ‘rating’ my tree, too. It says I have a couple of mistakes. Well, who doesn’t have some mistakes in a tree with over 6,ooo people. No ides how to correct anything when an AI has said I’m wrong but I KNOW I am NOT wrong. Any suggestions? Guess I better just get my tree off that website! Linda

    Like

    Reply ↓

    • Linda, It is frustrating when the online Trees (and even close family) can’t agree on the real Tree. I don’t understand how your tree got wrecked – you are the only one who can change your Tree at Ancestry (unless you’ve allowed specific people to change it. Ancestry may suggest some potential Ancestors but you should not accept them unless you agree (I have several branches where Ancestry suggests I change, but, like you, I’ve done the research, and I refuse to change: it’s my Tree.
      The LifeStory tab only draws on what you have in your Tree – so you may want to add the inro about your older brother being married with children. If you add living people, they will be kept private and should not be in the LifeStory.
      AI has found several mistakes I’ve made (almost all are typos), and I fix them as I have time. Most folks will have some errors.
      I’ve found ProTools hints to be pretty accurate – a 3,450cM Match will always show up as a parent/child relationship. Smaller cMs may have alternate relationships (Ancestry usually lists the two relationships that are most likely, but if you clich on it, Ancestry will show a prett full range. If there is still an issue, I’d use the interactive Shared cM Project at DNA Painter to see all of the possibilities.
      I’ve reviewed several thousand relationships with ProTools and find it is spot on well over 90% of the time.
      My recommendation is to edit your Tree so it is correct and leave it at Ancestry – this will help other get it right too. Jim

      Like

      Reply ↓

  5. Ciao jim ricordi kevi e da molto tempo quel gruppo triangolati ho avuto un altra corrispondenza sempre collegata e triangolazione con quel gruppo sul chromosoma16 anche se il nuova corrispondenza e di soli 8,5cM ma corrisponde ha ad altre 16 sempre con triangolazione tu dicesti una volta che potrebbero essere distanti perché i segmenti triangolati variano tra 7,1cM fino ha 9cM.

    Like

    Reply ↓

      • Kevin.Si ho visto su dna painter e la connessione può essere oltre 8trisavolo.però questo gruppo ti ricordi continuo ad avere corrispondenze che triangola con il gruppo tg già esistente.questo gruppo sono qualcuno cugini più vicini tra di loro ma tra di loro sono anche distanti.come si fa ha capire se discendono da parecchi figli?

        Like

  6. I really like your #A/#L system, so great to have that visual distinction within ‘colored dot’ match groups on Ancestry, gonna steal it! my hope is that clustering when fully operational w lower cms will let me crack some groups into subgroups descended from mysterious older ancestors and fold in folks from my huge ‘no matches’ group that surely match w me to a mysterious common ancestor way back in time.

    Like

    Reply ↓

    • Sari, Thanks for the feedback. The #A0080 method lets me have unlimited tags (vs only some dots that I have to remember). And now Ancestry lets us download a page of 50 Matches – so I can take a list of Shared Matches, download them, and then sort by the Notes field and all those tags come up together. And ProTools lets me find even more close relatives – who will have the same tags…
      With a “floating branh” (what I call a group with close shared Matches), I keep building out *their* common Tree. I work down the Shared Match list, finding more Matches who are more distantly related, which takes their Tree back another generation – hoping it will intersect with my Tree somewhere, Jim

      Like

      Reply ↓

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.