Just to add to what Michelle said, here's a link to the most recent (last August) update of WikiTreer Blaine Bettinger's Shared cM Project: https://thegeneticgenealogist.com/2017/08/26/august-2017-update-to-the-shared-cm-project/.
This is a crowd-sourced effort where people who have confirmed both genealogical and autosomal DNA relationships report the connection and the DNA sharing. It gives a good real-world range in the expected amounts of shared DNA by relationship.
What you'll notice is that zero amount shared can begin as soon as 2nd cousin 1x removed, and that the cluster chart (Table 1) doesn't extend beyond Half 4C/4C1R. That's because without significant pedigree collapse (cousins marrying cousins) there is often too little shared DNA at the 5th cousin level to be able to be validated as a match. Figure 1 indicates an average sharing among 5th cousins to be a total of 25 cM, but the theoretical average--with no endogamy or pedigree collapse--is only 3.32 cM. Fifth cousins share the same 4g-grandparents, so a relationship distance of 12...or separated by 12 birth events.
With the current state of the technology, small segments are as likely to be false positives--or simply noise--as not. An article by Blaine essentially categorizes single-segment matches of greater than 15 cM as a green light; 15 to 10 cM as a yellow light; and below 10 cM as a red light. Some feel they can work with segments smaller than 10 cM by extensive segment mapping and triangulation among multiple matches, but most agree that for anything below 7 cM it becomes effectively impossible, right now, to accurately determine which are real IBD segments and which are not.
If you carry on with the theoretical average of 3.32 cM or 5th cousins, you get 0.83 cM for 6th, 0.21 cM for 7th, 0.05 for 8th, and so on. So the net message with autosomal DNA is:
- You're solid through 3rd cousins (2g-grandparents).
- 3g-grandparents can likely be validated via triangulation with three or more 4th cousins.
- 4g-grandparents can only be validated with triangulation through multiple 5th cousins, and only if sufficient pedigree collapse occurred upstream to afford greater-than-average amounts of shared DNA.
- Beyond 4g-grandparents, any autosomal DNA validation will range from extremely unlikely to impossible.
Meiosis and DNA mixing/dilution is a generational thing. That's why the adage, "test the oldest generation first," is so important.