Actually, there are 4 averages for Serbia, 3 for Bosnian Serbs, 1 Montenegrin, and 2 for Serbs from Croatia.
I would merge Serb_Croatia_south and Serb_Croatia_north into one average.
So 9 averages for Serbs is not so much.
Printable View
Clusters for Serbs are more complicated than for Bosniaks and Croatians, because Serbs are more numerous and historically live on larger territory.
Ethnic map of Yugoslavia 1981 (blue Serbs, green Bosniaks, orange Croatians)
https://i.ibb.co/bscVKD4/J3-Ds-Zp672...Xddt-ROI-3.jpg
Well, those averages show only minor differences. That's what I'm judging by, of course I don't know about historical divisions of Serbia. But IMO if a country doesn't have massive genetic variation, there shouldn't be a dozen of pops for that country.
In my favorite calculator, Dodecad K12b, we have simply Serbian, Montenegrin, Croatian and Bosnian averages and I think they're decent.
In Serb_Herzegovina cluster vbn included western part of Montenegro known as Old Herzegovina https://en.wikipedia.org/wiki/Old_Herzegovina
Most of the Old Herzegovina is part of Montenegro since 1878 and smaller part since 1912. I think samples from there should be in Serb_Montenegro cluster. People from Old Herzegovina are mixed with people from other parts of Montenegro for the long time, linguistically they are "montenegrized" and historically Old Herzegovina is tribal region like most of the rest of Montenegro.
In the last over 100 years Herzergovinian Serbs are only from part of Herzegovina which belong to BiH (https://en.wikipedia.org/wiki/Herzegovina), they are not tribal neither montenegrized like people from Old (Montenegrin) Herzegovina.
I removed Zaza / Kurd-Zaza ref from Dodecad. Some peoople asked to add Kurd, now other on FB ask me to remove because they are not Kurds. I don't want to have such problematic ref anymore.