I've taken both the v1 and v2 versions of Ancestry's autosomal DNA test. Until yesterday, I'd have said there wasn't much difference in the reported DNA shared between any of my Ancestry cousins and my v1 or v2 "selves".

For example, on both lists I have five individuals predicted to be 2nd cousins. They're the same people on both lists, and in the same order. Two of them have slightly different numbers of cM shared, and two of them have a different number of segments shared -- by one. In each case, I know the actual relationship. The first four are 1st cousins once removed, and the fifth really is a 2nd cousin.

Until yesterday, I also had twenty individuals predicted to be 3rd cousins. The order is the same in both lists for the first twelve, but after that there's a little bit of a difference. The person who was 13th in the v1 list is 17th in the v2 list. That's because he shares 9 cM less on the v2 list than on the v1 list.

However, yesterday a 21st person was listed as a 3rd cousin -- but only on the v2 list. Sharing is shown as 95 cM shared across 6 segments. This actually causes her to rank 18th among my 3rd cousins, in terms of cM shared.

The thing is ... on v1 she isn't listed as a 3rd cousin at all, or even as one of the 18 highest-sharing 4th cousins. She shows up as 4th cousin #19, with just 54 cM shared across 6 segments. That's a pretty huge difference.

Incidentally, this cousin had enough of a tree that I could figure out how we're probably related. She seems to be around my father's age, and is apparently a 2nd cousin twice removed. According to the "Shared cM Project" (https://thegeneticgenealogist.com/20...ed-cm-project/), the average sharing for 2nd cousins twice removed is 81 cM -- with a range of 0-201 cM. So either 95 cM or 54 cM is possible for 2nd cousins twice removed.

However, the average for 4th cousins is 31 cM, and the range is 0-90 cM. So if there's 54 cM shared, then 4th cousin seems quite plausible. But if there's actually 95 cM shared, 4th cousin would not seem all that plausible. So you can see that if I couldn't determine the relationship from trees, the v1 list would have been much more misleading that the v2.