Exactly, it's just some dude having fun with sabermetrics.
Sabermetrics works for things like baseball with enormous amounts of data to work with over long timescales and a very well-defined and limited domain with explicit rules and constraints. This hardly applies to war, especially over millenia. If anything, this guy's analysis just proves the point. Reminds me of the tongue-in-cheek "power rankings" people make for granola bars or hot dogs.
The author is just looking at purely combat performance as some sort of way to rank them. Anything deeper wouldn't really work from a data perspective because we really don't have the best records from back then.