Announcement

Collapse
No announcement yet.

Heads up on great changes coming to FamilyTreeDNA.com

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by KATM View Post

    I can only hope that FTDNA is still computing the data, and such disappearances as my 2C1R and those that others have mentioned will turn up after the dust settles. As for how long that will take. . . we shall see.
    I've no doubt the problem is caused by the imputation used since the re-launch. MyHeritage have been using it for years and you get similar problems. E.g. the same DNA matches who validly matched at GEDmatch and the previous version of FamilyFinder, failed to match on MyHeritage, regardless whether it was an actual MyHeritage test or a transfer. But even there it typically affected only more distant matches such as those sharing just 1 segment. Losing matches as close as what you're saying shows there are SERIOUS problems with their methodology.

    Comment


    • #32
      Does anyone know if FTDNA is looking at the potential issues with the matching algorithm? Reading other postings there definitely seems to be something going on. One of the kits I manage is for someone looking for her biological family. She had several cousins sharing in the 110-150cM range. Now all of them disappeared. She only has one other cousin with 167cM that remained with about the same amount of cM. What does this mean? Either the missing cousins prior to the upgrade were all false matches or the algorithm is wrong.

      Comment


      • #33
        In lieu of what I was going to post (about how we should all contact Customer Support to let FTDNA know of the problems we've been having), I will instead post what I just got in an email to group project administrators, from FTDNA. It seems that they are aware of the issues:
        Last week, we announced exciting updates to the platform and our services. These recent updates include improvements to our autosomal (Family Finder) matching algorithm. As we continue to refine the recent update to Family Finder matches, some customers may experience fluctuations in their match list as we synchronize matches. We expect to have this sync completed by July 23rd.

        In addition, we are aware of the issues with viewing family trees for your members’ matches and working to resolve this, along with the one that causes some members’ matches information not to be displayed when clicking their name on the match list.

        We are looking into a few items with the Illumina OmniExpress tools in GAP as well.

        We apologize for the inconvenience but are excited for you to experience the first of many planned improvements!



        Sincerely,
        FamilyTreeDNA
        The link to the "recent update to Family Finder matches" is to an FTDNA blog post. At the bottom of that post is a graphic of the Chromosome Painting which shows an example with more "Overarching" (or as now labeled in Chromosome Painter, "Super") populations (i.e. continental regions), if anyone would like to view another example. It uses eight "Super Populations" to give the effect of more colors, but does not include any myOrigins 3.0 Population Clusters (such as Ireland, Magyar, Eastern India, Northern Levant, Southern Ethiopia, etc.). The ultimate released version should include Population Clusters.

        I also just this minute got another email to me as an FTDNA customer, which has the first and fourth parts of the email quoted above.
        Last edited by KATM; 7 July 2021, 02:57 PM.

        Comment


        • #34
          I logged in to my account, and see that there is now a message posted at the top of the Family Finder Matches page:
          Please be aware, as we continue to refine the recent update to Family Finder matches, some customers may experience fluctuations in their match list as we synchronize matches. We expect this to complete by July 23rd. We apologize for the inconvenience.
          Guess we'll see how it looks in 16 days.

          Comment


          • #35
            I wonder if "synchronize matches" means that they don't know what they are doing?

            Comment


            • #36
              Scratching my head, trying to understand how a match with total 95 and longest 41 cM can turn into total 62 and longest 46 overnight.
              I always ranked FTDNA higher than other sites, but not now.

              Comment


              • #37
                Waited about 25 minutes for matches to be "generated" on my old FamilyFinder kit -- nothing happened except for the icon going around in circles. I'm guessing the "generation" program is actually comparing my kit with every other kit, and of course that it likely to take a long time! While we wait for results, it would be extremely helpful if FTDNA could provide some clues about exactly how the segment matching algorithm has changed, and the technical/statistical reasons for the changes. It is understandable that the industry and FTDNA have learned a lot in the past few years about how each segment matching algorithm performs and how they can be improved -- but we need to know too, in order to make intelligent use of the new information.

                Comment


                • #38
                  I'm hoping someone can tell me how I can compare matches with this new version.
                  I used to click on a new match and could compare matches in common.
                  Maybe I'm going batty, but I cant seem to find how to do this on the new version.

                  Comment


                  • #39
                    Originally posted by rivergirl View Post
                    I'm hoping someone can tell me how I can compare matches with this new version.
                    I used to click on a new match and could compare matches in common.
                    Maybe I'm going batty, but I cant seem to find how to do this on the new version.
                    Not clear, eh! Took me a bit to find it.

                    Look at the third icon from the right on the line with the kit's name. The icon looks like a double-profile. Click on it and you will get the choice of 'in common with' and 'not in common with'.

                    Comment


                    • #40
                      Originally posted by georgian1950 View Post

                      Not clear, eh! Took me a bit to find it.

                      Look at the third icon from the right on the line with the kit's name. The icon looks like a double-profile. Click on it and you will get the choice of 'in common with' and 'not in common with'.
                      Thank you Georgian, I shall give that a go.


                      Just tried it out.
                      What an "arse about" change.
                      Why do they bother changing something just to make it make it more complicated.
                      What was the point of that change.. just frustrating....

                      and then having to wait for all the matches to reload, again. Arrggghhhh.

                      Last edited by rivergirl; 10 July 2021, 04:56 AM.

                      Comment


                      • #41
                        Originally posted by DavidB View Post
                        I am Ashkenazi and my Family Matching now shows several known relatives suddenly related to me both paternally and maternally, although that is unlikely (except very distantly). Two of these instances involve sibling pairs (in both cases first cousins of mine) where one sibling is shown as paternal or maternal and the other both. I hope this is just a bug and will be gradually worked out.

                        Anyone else notice this?
                        I chased relatives on all sides to get tested, and over two dozen have. My mother's sister matches almost every single one of my known relatives - father's side as well as Mother's side, I linked her, FTDNA showed her as Red. I linked a few more, FTDNA showed one of my second cousins as Purple. I linked her as 2nd cousin on my Father's side, FTDNA switched to showing her as just blue.

                        Now -- FTDNA shows them and many more of my known relatives as Purple. I have no doubt they are correct about that.
                        Very distantly counts, so yours are probably also correct.
                        That is the, or at least a, major joy of endogamy.

                        As for "Two of these instances involve sibling pairs (in both cases first cousins of mine) where one sibling is shown as paternal or maternal and the other both. " -- siblings do not inherit the exact same DNA. So yes, it is totally possible that is correct.
                        Last edited by loobster; 11 July 2021, 05:09 PM.

                        Comment


                        • #42
                          Whether the "chromosome painting tool" turns out to be helpful depends on how its internal algorithms work. If it depends solely on the "ethnic origins" probabilities of individual SNP's (without taking into account the probabilities of the scores for neighboring SNP's, for example), I don't expect it to be uniformly successful, and it might not even be useful in identifying maternal versus paternal segments. LIke other aspects of the genetic genealogy toolkit, "chromosome painting" needs to be subjected to objective testing -- in this case, perhaps, by comparison with "long read" sequencing, to verify that the "painting" is really inferring the correct sequences from individual SNP scores. In fact, as more "long read" data becomes available, one of the important features that might set one vendor apart from the others is the degree to which their proprietary algorithms can be shown to deliver the best approximation to the more expensive "long read" data. Each vendor will need to convince us that we should prefer their algorithms and tools.

                          Comment


                          • #43
                            Is Advanced Matches working for anyone?

                            Comment


                            • #44
                              I agree totally with the gist of the comments by John McCoy above, but from the standpoint of being far more interested in the narrow but "deep" focus of Y testing.
                              Our surname project is heavily dependent on Y STR/chromo testing, and while I understand the usefulness of autosomal testing and encourage posting across as many services as possible, it seems to fallen into the non-scientific abyss of "ethnic" and "origin" theories for the sake of marketing to those who do not understand that nation-states have not existed for most of human history or that "tribal identity" might have nothing to do with genes but adoption and/or absorption by yet another "tribe," frequently identified by a name assigned to a language group having nothing to do with genetics.
                              The idea that change is always an exciting, positive improvement has the ring of political sophistry. It sidesteps the issue of the learning curve for everyone: casual samplers and genetic genealogists such as Roberta Estes. I would have expected FTDNA to provide an explanatory slideshow or video for customers (and especially for Group Administrators) when so many terms are altered, but it would probably not answer the question of why so many useful items keep disappearing from the interface as tools. After the loss of the very useful FAQ (for instance, explaining the probabilities of relationships for various matching levels at various testing level for Y -STR), we are now being presented by decisions made by "that [person] behind the curtain." It is difficult for me to explain what happened to the match lists for my yDNA/FF members overnight. when I do not understand them myself. It also makes it difficult to justify any "upgrade" from a 37- to 111-marker Y test. It seems odd that 23andme provided me with an accurate Y SNP, admittedly upstream from my Big Y finding, but way downstream from my 67 and 111 tests at FTDNA, at significantly less cost.
                              As the only game in town for integration of Y results into a prebuilt Project grid, FTDNA is making it harder to explain the usefulness of the Y tests to prospective samplers with each iteration of "improvements" at FTDNA. I understand that each service uses proprietary algorithms, but as Mr McCoy points out, how are we customers to evaluate the decisions made to tinker with our kit pages?

                              Comment


                              • #45
                                Ok, so the message on FamilyFinder about 'synchronization' of matches (expected to be completed by July 23rd) has now gone. Has anyone noticed any drastic change or improvement to their matches?

                                One thing is undeniable, namely that the total number of matches is higher than last month (before the upgrade). I used to have 335 matches, now it's 597. By comparison, the match numbers of my mother increased from 1640 to 2258 whereas for her sibling it went from 1439 to 1785. Now at first glance that looks like progress. But the problem is the chromosome location of these new 'matches.' E.g. in the case of my mother they're mostly matches in the archaic HLA region of chromosome 6 (29 to 33 million). For my uncle they're mostly in the well-known 'excess IBD' region of chromosome 9 (38 to 72 million). For me it's a combination of such locations. I find it exceedingly odd that such problematic chromosome regions had far fewer matches before the update and now they're flooded by them. Wasn't the point of the upgrade to deliver more relevant matches?

                                Also, has anyone noticed the huge impact of the new 'anti-endogamy' restrictions? I lost dozens of matches on a segment shared by Ashkenazi Jews, even though I knew from investigations at GEDmatch that this segment was triangulated and valid. Very annoying.
                                Last edited by StefG; 25 July 2021, 12:14 PM.

                                Comment

                                Working...
                                X