AncestryDNA cousin estimates

[Matt62]

I've noticed something bizarre with the AncestryDNA cousin prediction. I have a confirmed 3rd cousin once removed who is estimated to be my 5th-8th cousin with a "good" as opposed to "very high" degree of confidence. By contrast I have a multitude of predicted cousins in the 2nd-4th and 3rd-6th levels with a "very high" degree of confidence who based on our family trees could not possibly be so close. The common ancestor has to be at least in the 17th - early 18th centuries, if not a century earlier.

How do they work out these estimates? Do they use longest block or total cM?

[davidft]

This should help

http://dna-explained.com/2012/08/06/what-does-mcra-really-mean/

and this (although a little out of date I think ie I believe you can post Ancest5ry results to Gedmatch now)

http://legalgenealogist.com/blog/2012/08/12/gedmatch-a-dna-geeks-dream-site/