Family Finder Update Comparative Results

The new Family Finder update at Family Tree DNA is live this morning. I wrote about what to expect here and here.

Family Tree DNA’s dual matching algorithm leaves most of the current matches in place, with the newly added portion providing matches for people who didn’t match the old total threshold. The two matching thresholds currently in effect are:

  • 20cM total with a minimum longest single segment of 7.69cM
  • No total cM requirement but a minimum longest single segment of 9cM

There are lots of happy people reporting new matches on social media today!

You can check your new matches by signing on to your account and clicking on Match Date to sort by match date. A second click sorts in the opposite direction.

Ff match date

However, I noticed that not all of my new matches carry the match date of 5-23-2016, which is the latest match date.

Comparing Your Match Files

The best way to find all of your new matches is to download your matches list and compare to a previous match list. At the bottom of your match page, there is a link to download matches.

FF match download

I downloaded my match list on 5-23-2016 and again this morning, 5-26-2016, so that I could compare old and new.

My old number of total matches is 1392, and my new is 1447, so my gain would be 55. However, that’s not the entire story.

I compared the two files, and I lost a total of 20 matches. Three of those were in the 4th to remote category and the rest were in the 5th to remote category.  These must have been casualties of the tweeking of the algorithm.

In reality, I didn’t gain 55, I gained 75 new matches.

That’s about a 5% increase in matches.

The technique I use to compare the files from the two days is to color code one and then combine them into one spreadsheet file. I sorted by full match name, and that made it easy to look for any two white rows or any two green rows, which indicate a loss or a gain.  People appearing in both files will show as a whole and green identical row one after the other forming a green, white, green, white pattern on the spreadsheet, so it’s easy to spot an aberration.

Green is the 5-26-2016 file, so two white rows together means that one of those two white rows from the 5-23-2016 file doesn’t have a corresponding green row, so it was lost in the update. You can see that happened with Jerry, colored red, in my spreadsheet below.

FF Update Label

Two green rows together means that one of those green rows didn’t have a 5-23-2016 white row, so it was a gain. You can see that with Helen, colored purple, above.  I colored by loss rows red in the Full Name column and my gain rows purple.

Then, I sorted the spreadsheet again by cell color in the Full Name column. All the red and all of the purple appear together, so it was easy to see gains and losses.

Ff sort

Of course, in my example, there is only one gain and one loss, but the concept still holds when there are more.

The one unusual aspect that I found is that the match dates for all of my new matches are not 5-23-2016. They are primarily 5-4-2016 with a few 5-23-2016 and one 5-21-2016.  It looks like this rematching process was completed in steps on three different dates.

This means that you can’t rely entirely on the match date on your match page at Family Tree DNA by looking for today’s date. Your best bet is to compare a current matches file with a previous matches file.  If you don’t have a previous matches file, it looks like anything in the month of May is likely a new match.

Evaluating Matches

When I evaluated my new matches, I was surprised to find that only 7 were the result of the new algorithm change. I’m not sure if the bottom two would have appeared before, as there were 20.xx cM and I’m not sure if the threshold was exactly 20.0 or 20.99.  In any case, they are new now, and the top 5 are definitely a result of the new algorithm.

FF new matches

The balance of my new matches seem to be a result of truly new matches that have recently tested (I received 74 last month alone) or a result of the algorithm tweeking, because they clearly qualified as matches previously, under the old algorithm.

These 67 matches longest block ranges from 7.70 to 12.04 and the total shared cMs range from 21 to 54.

Now, of course, it’s time to see who matches whom, who triangulates, who phases against my mother and which of these matches have trees sharing a common ancestor.

I have a lot of new cousins to meet and genealogy work to do, and I hope you do too.  Let me know how the Family Finder update affected you!

49 thoughts on “Family Finder Update Comparative Results

  1. On GEDmatch, with the filters set at 500 SNPs and 5 cM, I have one match totaling 21.1 cM with segments of 8.6 cM, 6.8 cM, and 5.7 cM. With the filters at 500/3, the total is 31.8 cM, and at 500/1, the total is 63.8 cM. Yet this person still does not show as a match at FTDNA. Am I missing something?

    • Obviously I can’t give you an answer, but what I can tell you is that each company has a different way of measuring your DNA. Each one has a different way of “healing” bad segments or no reads. How big those “forgive” segments are can make a significant difference. How they measure can. I often see differences at GedMatch when compared to the other vendors, and not just in one direction or the other. I suspect the processing is just different enough that you’re under the threshold at FTDNA for some reason.

      • Understood. This person otherwise appears to be a critically important 3rd cousin, and absent the Bettinger crowd-sourcing project, I would have written him off as considerably more distant than that.

        (Roberta: Regarding this apparent 3rd cousin, I would very much appreciate your feedback on some interesting AncestryDNA matching results. Shouldn’t take more than 5 minutes of your time.)

      • Since you didn’t edit out my parenthetical, Roberta, I’ll assume this longer follow-up post is OK. As I mentioned recently on another thread, I’ve spent much of the past 4 years trying to identify my French-Canadian paternal grandfather. Through DNA testing, triangulation, and traditional document research, I believe I’m very close to solving the mystery.

        Assuming I had correctly identified my 3rd cousin (per the above) I added my suspected grandfather and all of his New World ancestors to my Ancestry.ca tree. I then did an AncestryDNA test, having previously tested only with FTDNA. I was hoping that Ancestry’s “leaf plus DNA match” feature would reveal something new.

        I understand the limitations of AncestryDNA results, but it is the pattern of those results which intrigues me and may interest you. I have 192 matches with leafy public trees, including 52 trees in which the ONLY common ancestors are ancestors of my assumed grandfather, with a roughly equal split between his paternal and maternal sides. I have independently researched all 52 of those lines and eliminated 9 as insufficiently-documented. So there are 43 trees well-documented in which the only common ancestors are ancestors of my assumed grandfather.

        What’s your assessment of all that?

  2. The age old question (prob not so old) but do you set more importance to one or the other? To total cMs shared? Or to longest segment shared? Im not seeing any significant difference in number of matches but suppose they will change all day. I havent had a dozen new matches in the last month and a half.

  3. FWIW, a quick check of one of my family kits gained about 150 matches, or growth of roughly 2%, to ~7150 matches. However, being Ashkenazi, I don’t usually look at matches that far down the match list.

    • Steve: You’re in my uncle’s match list (look for Harry Braunstein in your matches) with a 76.46 cM match. But as expected, I see no names or places and nothing in your tree that looks familiar.

      See my other comment below.

  4. Another easy way to see most new matches is to sort on cM shared. Anything under 20 will be a new match with the 9 cM min.

  5. Most of new matches for today, appear to be dated 05/04/2016. On two kits, there is about a 5% gain. In my kit, most of the 05/04/2016 new match are dispersed in the bottom 20% of my 866 total matches. My mother’s kit has the 05/04/2016 matches dispersed over the bottom 40% of 1069 matches. Being one generation closer to the MRCA’s might be the reason that my mother has more matches than me and that her matches using the new algorithm are more dispersed.

  6. I can’t see why people are getting excited about their new matches on social media, Roberta, as you report. I’ve looked at my new matches, and as with the old ones, the names don’t mean a thing to me and on the few occasions that people give their family-tree details, none of that makes any sense either. Some matches are supposed to be my 3rd-5th cousins, but I have paper-trail links to known distant cousins – including 5th cousins I know personally! – but no known connection whatsoever to so-called 3rd-5th cousins who match me on Family Finder. Especially those with recent ancestry, unlike me, from continental Europe (Netherlands, Belgium, Germany …).
    Harry

      • I don’t know how many matches I had just before the change, but I have 16 matches dated 26 May 2016. Total shared cM is not less than 20 for any of them.

        I was anticipating one old match who shows up at 23andMe and Gedmatch but not at FTDNA because of the 20 cM minimum — but she is still not on my match list at FTDNA.

        I don’t see any clear evidence of a change.

  7. I have several new matches and tried to filter them by any new ones after 5/23/2016 because I have a couple of those. Unfortunately, instead of seeing all 15-20 new matches I can only see about 6 (when filtering by date). This may be glitchy for a while.

  8. Thanks for telling us about this Roberta,

    So I checked today as you recommended. Also Ashkenazi like Steve Adelson (see above).

    My results used to have 7,053 matches. Now there’s 7,527. Lost 91 including a suggested 3rd cousin and people with shared cM up to 113.3.. Gained 565 including 11 suggested 2nd cousins with shared cM up to 138.5. You are quite correct that all the new match dates are spread out from May 4 up to today. I’ll keep those lost 91 in my Excel file with a special coloring. Also the notifications over the past few weeks only notified me of 10 new matches. They obviously did not notify people about the new matches from this change, or at least they haven’t yet.

    I should say that of my matches, with no-one closer than about 130 suggested 2nd cousins, there is only 1 person on my list that I know that really is a cousin, and he is the 3rd person listed out of the 7,000. Of all the other submitters and surnames and research places, there are just a handful that look even the slightest bit promising.

    So that leads to a question I have for you. FamilyTreeDNA does state here https://www.familytreedna.com/learn/jewish-dna-testing/autosomal-dna-research-challenging/ that they adjust for Ashkenazi endogamy. They say this: “Beginning on April 21, 2011, we have modified our Family Finder matching algorithm to address this. The changes affect the match list for Ashkenazi Jews. The outcome is calculated Family Finder relationships that more accurately reflect relationships to other Ashkenazi Jews.”

    So my question is, how do I know if my results have been adjusted? There does not seem to be any indicator in my account or on the Family Finder results that indicates whether they adjusted mine. And is there anywhere that says how the adjustment works? Are they just changing 1st cousins to 2nd cousins, and 2nd cousins to 4th cousins, or what?

    Louis

    • I think surnames are harder to base our research on given Jews did not take surnames as early as other populations and when they did different people within a family may have taken different surnames. There are also Americanized Jewish names that may impede looking for surname matches. One of my Schwarzbaum’s (black or dark tree or wood) changed his name to Sherwood. And, female testers without an ancestor list or tree aren’t at all helpful in determining where a match comes from.
      That also makes Y-DNA testing more difficult, especially since it goes back so far in time. One reason I did the test on my brother was to try to find other possible surnames. My father was a member of the Odd Fellows.- my brother’s test proved it!!

      • No, I don’t believe so.

        p.s. You are in my uncle’s matches as well, as a distant cousin. Before yesterday’s FamilyTreeDNA upgrade at 70.0 cM match, but now we’re closer at 71.2 cM.

      • Louis what name should I look for in matches? my siblings might also match – look for Schwarzbaum surname.

      • Carolyn: It’s my uncle’s test under his name: Harry Braunstein. But at only 71.2 cM, it’s a distant relationship, and I don’t see any familiar surnames or places in your Ancestral Surname list. That combined with endogamy makes me pretty sure we won’t be able to find what the connection is.

  9. I first checked my mom’s kit as it is the largest I manage with now 1739 matches. She gained 72 and lost 14. The vast majority of the gains were dated 5/4. Six of Mom’s new matches had previously matched me only. From where and how they fit in my map I had guessed they “should have” matched Mom and now they do. Four of those matches are definitely due to the new algorithm. In my quick survey I noticed that the total and longest segment cM values had changed slightly for many of the kits, which may explain how the other two now fall on the match side rather than the non-match side.

  10. I also noticed all the Shared cM values and Largest block values have changed for every match.

    For my matches, the Shared cM is on average 2.8 cM less than it was before. The biggest drop was 21.5 cM. The largest gain was 4.6 cM. The Largest block is on average 0.1 cM less than it was before. The biggest drop was 10.1 cM. The largest gain was 2.8 cM.

  11. GAINED 41
    LOST 8
    And I saw one person who I had at one time matched me, then recently I discovered didn’t match me, and now she’s back. That may not at all be related to the change.

    The lost matches are 5th-remote. Here are the lost matches total cM and longest block in cM, respectively:

    Shared cM Longest Block
    35.84466 7.91360
    24.71920 7.98712
    34.17097 7.98712
    38.80461 8.13629
    44.37876 8.13629
    29.91797 7.77332
    29.06434 7.77332
    35.32247 9.75859

    When I compare the total cM on many of the matches from 5/25 to 5/26, it looks like they recalculated some numbers. Otherwise, I can’t see why the 8 above would be dropped. Roberta, do you know if there was a recalculation of total shared cM while the algorithm was being run?

    • Here is just one of many examples of what I mean about the “recalculation.” The top numbers are from yesterday, 5/25, and the bottom pair are from today. Same kit.

      Shared cM Longest Block
      36.61492 12.63062
      38.30226 14.10714

      • Ah… unless they had introduced a bug this month during their series of releases, perhaps causing some people to “disappear,” like the person I mention above. And then they fixed it in the last release. I pulled up an export from 5/8, and she was in it. Point being that some of the “new” matches may be “returned from the dead” matches.

  12. The “New Since” option is definitely not working with these new matches. If I do New Since 5/1/2016, the new matches do not show up. But if I sort all matches by shared cM and look at the lowest matches, there are a few new ones with a date of 5/4/2016.
    I do not know if this is related to the previously existing “New Since” problem.

  13. I’m seeing new matches dated today (27th) so FTDNA apparently is still posting matches. My cousin has about 150 new matches so far!

  14. I have Ashkenazi ancestry on only one side (father). I am now up to 416 PAGES of matches. I was foolish to not download my results before the change – I was not expecting it to happen immediately. I do have it loaded to DNAGedCom.com but it is an older list. Some cousins in the 2-4th range seem to have moved down. At least one that I had written FTDNA about soon after my test (he had moved from 2-4th down to the next range which they could not explain) is now back in 2-4th. I will have to play with this more to see where people went.

    Being Ashkenazi I generally do not look beyond 3rd cousin range if surnames sound Jewish rather than German (mother’s side), although this can be difficult to differentiate.

  15. I am fortunate that I have tests for both my parents and find it interesting that some of our “new” matches are new to us individually but not to my combined spreadsheet. I have 82 new matches: 42 are truly new and match one of the other of my parents; 23 have previously matched a parent but not me; 17 match neither of them. Before the update I had 180 matches that did not match either parent. Now I have 183 but 24 of them are different kits. I had four losses but added seven I that a parent lost, and, ironically, 17 of the previous “neithers” now match a parent, the same number of new matches that don’t. So between our three kits I already had 40 names in the spreadsheet with the 23 that previously matched a parent but not me and the 17 that matched me but not a parent.

    I don’t think this is anything beyond a curiosity I noticed when updating my spreadsheet. The periphery is still the periphery where some kits just barely match and some just barely miss. But I will keep in mind that a “not Mom” match is not necessarily “Dad”, and just because someone doesn’t match Mom or Dad today it doesn’t mean they might not match tomorrow.

  16. I’ve only looked at one of the kits I manage and found something interesting. I looked at the new reported total cM’s and largest segment listed for a great aunt between her sister and 4 niece/nephew’s and compared to what previously reported. I found a significant change between my mother & her aunt, the “largest” segment changed from 105 cM’s to 139 cM’s. The total cM’s wasn’t as significant, dropping from 1596 to 1592. I then compared the old chromosome browser data with the new and found several slight changes, but the significant changes are found on 4 chromosomes where 2 or more segments are now combined into 1 larger segment. On the X it went from 5 segments ranging in size of 4 cM’s to 84 cM’s and is now 1 segment of 195 c’Ms. (The X segments must not be included in the combined totals). The new matching effects more than just gaining or losing matches 🙂

    • I’m finding the same thing. Discovered this one because this person matched Dad and me previously and now also matches Mom, which is unusual for us. Segments in both places where I match my parents were combined, and in Mom’s case put her kit over the threshold for matching, assuming her old numbers were the same as mine.

      Me old
      1 14340307 22180765 15.35 2576
      1 22449877 30418573 8.83 2000
      5 174545619 176763233 3.76 572
      5 177438258 180372094 6.47 875
      Me new
      1 14340307 30418573 24.44 4676
      5 174545619 180372094 10.75 1547
      Dad old
      1 14340307 22180765 15.35 2576
      1 22449877 30418573 8.83 2000
      Dad new
      1 14340307 30418573 24.44 4676
      Mom new
      5 174545619 180372094 10.75 1547

      Now I’m wondering if it matters enough that I will have to rework all my spreadsheets…

  17. Hi

    I gained 8 and lost none…………..one is noted as a 3rd cousin , but surname appears in my family tree in the mid 19th century.

    My wife got one and lost one

    regards

  18. As of my call to FTDNA today their customer service recording says problems with access are occurring because of weather in certain areas. Basically, you have to leave a message. I have been having ongoing issues with entries in the note block in Family Finder not saving my text. Annoying. There are other issues but maybe I’m the only one bothered by them.

  19. FTDNA message this morning… no graphics or anything showing on screen – “Our website is currently undergoing maintenance to introduce some exciting new updates.
    Please check back soon!”

  20. Tuesday June 7
    Interesting! I just signed on to ftdna, got in with no problem, no changes and no warning notices or heads up comments! Is it being done in cycles? What are we looking for?

    • They update the site periodically, about once a month, when they put up that banner for a few minutes early in the morning (US time). Most of the changes are behind the scenes.

      • I hadn’t seen the mrssages previously but the site came up. Thank you, Roberta!

  21. Pingback: Family Tree DNA Introduces Phased Family Finder Matches | DNAeXplained – Genetic Genealogy

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s