How reliable are matches

[g eli]

Mary severn born about 1811 Kimberley Nottinghamshire  has been a brick wall for several years, I have made several trees that could solve the problem. Recently one of those trees showed me with 35
matches, dating back to 4ggrandfather of these 3 are from relatives I know about, 9 I am not sure about, the rest seem to have the same ancestor.
How sure can I be that this is the right ancestor,

[Biggles50]

If you have a paper/citation etc trail to the person then you are part way there.

If you can validate with DNA the branch to the person and from them to the DNA match then that is as good as you are likely to get.

There are other DNA tools and techniques that can be used but given the likely low cM with a DNA Match then errors can exist.

Multiple DNA validated branches to the person will substantiate the Tree.

If you create a Gedcom of the DNA Match to the person concerned and include as many Shared Matches as you can in the Gedcom you can then upload the file into DNA Painters WATO tool.

Once uploaded and completed run the analysis and it will give you the Probability in numerical terms.

Good Luck.

[Glen in Tinsel Kni]

It should be possible to go back 5-6 generations with the standard Ancestry type tests but the more distant the lower the chance of actually matching. If that's the point being looked at then it's getting to the limit of the tests.
About 85% of full 3rd cousins appear as a match, around 50% of 4th cousins and 30% at 5c.

[g eli]

Thank you for replying to my request. I have done some more research. that is me creating family trees for the matches rather than relying on the matches trees.I haven't finished but they seem to be ok. I already knew 3 of the matches being close relatives. The remaining 32 matches are mostly in the 5 cousin range with matches from 8 to 48 ( 0ne of each) the rest between 10 and 29

[Glen in Tinsel Kni]

I get so many hints based on trees of dna and non dna matches that can be disproved by a couple of bmd searches but they still message to argue my tree is wrong and that a man who died in 1892 was still fathering children 1895-1903.

[g eli]

I agree,I only look at tree hints for ideas, I have been lucky most of the people I have contacted are glad to hear from me.

[4b2]

How sure can I be that this is the right ancestor,

You need to account for all of your matches back to about 20cM (or more) and all of their shared matches that seem to be real (share 3 matches+ in a cluster). By doing this you will be able to validate lines, and likely be left with other groups of matches who you can try and place in your tree.

I'm of the impression that (virtually) all DNA matches on Ancestry down to 20cM are real relatives. Beyond that it's not easy to tell.You will have clusters of matches below 20cM (many of them), but it becomes very difficult to figure out how they are related to you. This is owning to - the further back the relationship, the less likely they will overlap with enough matches to suggest which lines of our ancestry they relate to; and that they can relate to ancestors born more in the window 1670-1700. For example, for 7th cousin matches, we have 128 possible pairs of great-grandparents.

***

If the test is taken by someone born in the window 1950, then there are some shared matches showing ancestors back to window of 1670, while there will be other lines that only have shared matches back to the window of 1800.

Some lines just happen to be very faint. This could be due few descendants, few descendants who have taken tests, little DNA inherited on that line, or maybe issues relating to Ancestry stripping out chunks of DNA they think are not inhered (but are). The only way you can figure out which lines are faint and which are strong in your test is going through all the matches. Going through the matches of relatives' tests may yield earlier generations on weak lines. But in my expedience weak lines in one test seem to correlate to weak lines in relatives' tests. This could point to Ancestry stripping out too much inherited DNA.