Starting high to low also means not getting sucked into exploring small segments unless you have an immediate reason to do so. There have been zero, zilch, nada scientific, peer-reviewed studies validating or quantifying the use of small segments in genealogical research. Or even what a "small segment" means. Some not-scientifically-tested but empirical data suggests that 6cM segments are false 74% of the time; 7cM segments are false 58% of the time; and 8cM segments are false 38% of the time. Not "difficult to validate"; false. At the level of 4th cousins, your theoretical amount of shared DNA is 13.3cM; for 5th cousins it's 3.3cM. Moreover, only 14.9% (17:3 odds) of your 5th cousins are going to share any detectable DNA with you at all, so it takes a lot of luck as well as work to use autosomal DNA back to 3g- or 4g-grandparents.
Too, often overlooked, is that what we see when GEDmatch or FTDNA or 23andMe reports a segment length is not only computed via a form of linear equation originated by an Indian mathematician named Kosambi in 1944 (refining work by a gentleman named Haldane who coined the term centiMorgan way back in 1919)--meaning that it isn't a physical measurement at all, it's an estimated one; there is actually no centiMorgan "length"--but the value in cMs is what is called "sex averaged." All us men know that females are more complex than we are, and the female genome does see about 30% more DNA crossover at gamete production than males. The result is that the cM computation for a male and a female at precisely the same range of DNA base pairs can be very different, sometimes as much as 10cM or 15cM. Working with segments as low as 7cM means we're using a sex-averaged value that may see the actual gender value above the threshold for males, but below the threshold for females.
For example, here is an actual match between a cousin and me on chromosome 18 starting at base pair 72,323,427. The three different centiMorgan computations by gender are:
- Male: 11.1cM
- Female: 4.4cM
- Sex-averaged: 7.9cM
We have to use the sex-averaged value, in this case because I'm male and she's female, but also because that's all that GEDmatch and the testing companies can realistically report to us. But is using the 7.9cM average value valid if we're talking about two female cousins who descend from two different daughters of a shared grandmother? The gray area is simply too broad when trying to work with small segments, IMHO.
The Shared cM Project's crowd-sourced data shows 35cM average sharing among 4th cousins; the mathematical average is 13cM. I personally would split the difference and not bother trying to investigate matches that show less than about 20cM or 25cM unless you absolutely feel you know what you're doing.
Last, RJ said, "There's a process misnamed triangulation..." And he's correct. That Latin root "triangulus" is an artifact that causes a lot of trouble. Three legs on the stool is a minimum and most often isn't enough. With autosomal DNA--I'm now speaking as a method (which is not proven or standardized, BTW, in the genetic genealogy community), not of WikiTree guidelines--using just three people can work fine at the 2nd cousin level, but probably needs an additional tester or two to validate at 3rd cousins, or 2g-grandparents. When you get to 4th cousins, it rapidly becomes much more complex. In fact, one of the researchers who is often pointed to as saying 7cM segments are likely valid is Jim Bartlett. What Jim actually wrote is this:
"However, my triangulation process involves a lot of work... I highly recommend starting with 15cM as a threshold (or even higher, if you don't have the time or inclination). Setting a personal threshold is a good way control the amount of work you are willing to put in."
Jim's point is also that working to establish valid evidence using small segments is very complex and difficult...if it can even be done. When he says his process involves a lot of work, Jim has said that it's typical for him to work with triangulation groups that have upwards of 20 people in them before he's willing to accept a 7cM segment as valid: "If 1 or 2 of those triangulated shared segments turns out to be IBS [identical by state, or not able to be validated to a specific ancestor...think, "identical by chance"], it's not harmful in the grand scheme."
Lindis, I know this does little to address the five matches you've found on GEDmatch and Ancestry. But it's refreshing to see someone asking, "How does this work?" rather than "How do I mark these 3g-grandparents 'confirmed with DNA?'" It's about understanding the process and the fundamentals, not about leapfrogging straight to an end result that may or may not be valid. I'm now putting my soapbox away for the day.... :-)