MyHeritage Matching

MyHeritage is getting some well-deserved positive press lately as one of the newest entries in the genealogical DNA testing market. After more than a decade hosting family trees and providing subscription-based access to genealogy records, the company launched its own autosomal DNA testing and matching service in November 2016. The test in normally US $99 and is on sale for $69 through 24 July 2017. (Note: I earn a small commission if you purchase through the links in this post. The cost is the same for you. Click here for more information.)

 

The MyHeritage Database

Their database of tested individuals is small—approximately 50,000 according to the ISOGG wiki—but growing rapidly. I had 95 matches on 30 May, 2017, and 129 by 17 July. Based on those numbers, their database is expanding by about 22% per month. That’s faster than AncestryDNA, 23andMe, or Family Tree DNA!

One reason for this rapid growth is their free upload program:  if you’ve already done an autosomal DNA test at any of the Big Three companies (AncestryDNA, 23andMe, or Family Tree DNA), you can transfer your raw data file into MyHeritage’s database without charge. My parents recently agreed to let me upload their data into MyHeritage’s system; they each had ethnicity estimates within 3 days and lists of matching relatives within 9 days. My mother has 233 matches and my father has 13. You can find out more about the upload program here.

The free transfer program isn’t the only explanation for the rapid growth at MyHeritage. Their database also contains a substantial percentage of people who haven’t tested anywhere else. Of the 37 matches who have responded to my messages, 13 say that they have tested only at MyHeritage. That’s 35%!

Although the majority of matches to my parents and I are from the USA, we also have quite a few from Canada and a smattering from Australia, Germany, Great Britain, and Ireland.

 

Communication at MyHeritage

I’m also impressed by the response rate of my matches at MyHeritage.  Of 125 matches whom I had messaged by 20 July, 37 have replied, or nearly 3 in 10. And an astounding 71% of my matches have trees of two or more people. The engagement rate of MyHeritage’s customers right now is quite high.

A downside of MyHeritage’s messaging system is that you can contact to 20 new matches per day without a subscription. And after 5 outgoing messages, you must enter a “captcha” code for each additional message to send it.

These limits do not seem to apply to matches with whom you’ve already communicated.

 

MyHeritage Matching Tools

MyHeritage has clearly paid attention to what the genetic genealogy community wants in a DNA testing company. As with their competition, they offer “ethnicity estimates”, which I reviewed here when they were first made available to users who had transferred raw data into their database. Like the other companies, their ethnicity estimates are hit-or-miss.

For DNA matches, the listing includes name, age bracket, country, estimated relationship, whether they have a family tree and how big it is, and three measures of DNA match quality: amount of shared DNA (in percent and cM), the number of shared segments, and the size in centimorgans of the longest segment. The latter are all important when gauging our matches, and at MyHeritage you don’t have to go looking for them.

 

My favorite feature of MyHeritage’s matching is their presentation of shared matches. (Shared matches are people who match both you and a selected relatives. In the screenshot below, I’m looking at the match information for Arlene; Philip, Gordon, and Roger match both me and Arlene, so they are shared matches.) What I love about the shared matches at MyHeritage is that they show us not only the estimated relationship and shared DNA amount to ourselves but also to the match we’re reviewing (in this case, Arlene).

 

This shared match information allows you to cluster your DNA relatives into family groups and narrow down the connection. It is similar to the Relatives in Common tool at 23andMe and much better than the Shared Matches tool at AncestryDNA and the In Common With tool at Family Tree DNA, neither of which tells you how much DNA your shared matches have in common with the selected person.

Similarly, MyHeritage shows you a direct comparison of your estimated ethnicity percentages versus those of your DNA match. 23andMe also provides this level of information. AncestryDNA only lists the ethnicities of your DNA matches (without percentages), and Family Tree DNA does not report ethnicity information for your matches.

 

MyHeritage does not currently have a chromosome browser, although one is expected in future updates.

 

MyHeritage’s Matching Algorithm

The biggest weakness of MyHeritage’s current DNA offering is their matching algorithm. What I mean by that is the computer program they use to decide how much DNA someone shares with you. I gauged their matching two different ways. First, I looked at how their matching compares with matching at other sites. Second, I compared my matches at MyHeritage with those of my parents.

 

MyHeritage Matching Compared with Other Databases

My top two matches at MyHeritage (other than my parents) are both genealogy buffs who are in all of the other databases. One is a 1C1R to myself and the other is a 2C1R. The table below compares their matching statistics to me across the different genealogical sites.

 

MyHeritage does a pretty poor job of matching to my 1C1R, failing to find eight segments and 81.5 cM relative to GEDmatch. That amount of shared DNA (260 cM) would strongly favor a prediction of 2C over 1C1R.  Their algorithm worked better for my 2C1R, with one extra segment than GEDmatch and nearly identical centimorgan estimates. For both relatives, MyHeritage’s estimate of largest segment size was similar to the other companies.

For more distant matches, MyHeritage has similar woes. Excluding my parents and the two cousins mentioned above, I have 33 matches at MyHeritage whom I know are in at least one other database. On average, MyHeritage predicts that they share 27.2 cM more than the same people at GEDmatch (n = 31 comparisons), 32.4 cM more than at 23andMe (n = 21), 18.1 cM more than at Family Tree DNA (n = 24), and 32.8 cM more than at AncestryDNA (n = 25). Worse, of those 33 matches, 11 of them do not match me at all in the other database(s), although MyHeritage predicts those 11 as sharing between 16.7 cM and 46.9 cM. The 46.9-cM match tells me that he is at GEDmatch, 23andMe, Family Tree DNA, and AncestryDNA, yet he matches me only at MyHeritage.

 

MyHeritage Matching Comparing Parents and Child

I have 129 matches at MyHeritage other than my parents. Because all of my DNA was inherited from one of them or the other, all of my DNA matches should also match at least one of my parents. Of the 129, two match my father, 49 match my mother, and 78 do not match either parent. In other words, at least 60% of my matches at MyHeritage are either false positives for me or false negatives for my parents. Of course, it’s possible that even more of my matches are false positives, but without a chromosome browser, I can’t tell.

Some of my false matches share substantial amounts of DNA with me. Arlene, the example I used above, is estimated to match me on 85.0 cM, with a largest segment of 19.6 cM. If you scroll up and look at the screenshot again, you’ll notice that her shared matches do not include one of my parents. And of the three shared matches shown in that earlier screenshot (Philip, Gordon, and Roger), only one matches a parent, my mother.

At the moment, even largest segment size is not a good metric for filtering the real matches from the false ones. Philip (a shared match with Arlene), shares a total of 78.5 cM with me and a largest segment of 45.7 cM, yet he does not match my mother at all.

On the other hand, these false matches aren’t completely random; based on their surnames or those in their trees, they are all Acadian, as is my mother. And all Acadians are related.  All of us, even if we don’t share measurable DNA. MyHeritage is clearly on to something, although their matching algorithm needs a lot of refinement. For that reason, it would be unwise to write off MyHeritage completely.

 

Recommendation

The best strategy (still) would be to test at either AncestryDNA (if you prioritize genealogical tools) or 23andMe (if you prioritize health information), then upload your raw data into MyHeritage’s database while the transfers are still free. (You can transfer those data into GEDmatch and Family Tree DNA for free, as well.) Given how much effort MyHeritage has put into providing tools that genealogists want in a DNA test, I am confident that they will refine their matching algorithm. Hopefully, the more people in their database, the sooner they will update their algorithm and the more accurate those changes will be.

8 thoughts on “MyHeritage Matching”

  1. I did a comparison of my matches with those of my parents. I have 25 matches at MyHeritage. I share one match with my mum and five matches with my dad. That means that 76% of my matches at MyHeritage are either false positives or false negatives.

  2. Not loving MyHeritage either.

    I’ve uploaded my DNA and can only find matches to five relatives (total) on that web site, whereas the same DNA file uploaded to GEDMatch.com shows at least 30+ matches to MyHeritage.com tests, on the first screen alone.

    I can’t find any answers on-line to try to help solve this problem. Can anyone help?

    In addition, when I’ve tried to contact MyHeritage.com about this, and some other administrative issues on their site, they haven’t responded. I have no idea if my questions have been logged, or are just floating in the ether.

    1. That’s interesting. I have 165 matches at MyHeritage and only 10 “H” kits in my one-to-many at GEDmatch.

      As for help, I’m afraid there isn’t much we can do done right now. MyHeritage needs to improve their matching algorithm and grow their database. In the meantime, it’s worth considering the free upload on the off chance that we’ll find legitimate matches there who haven’t tested elsewhere.

Leave a Reply

Your email address will not be published. Required fields are marked *