Someone who is adept at statistics could probably determine a confidence level for the hypothesis for example that 49.35% Greek are no less powerful than 50.42% Egyptian, given the # of games played. Maybe they could even say for example that evenso, there's a 95% chance Greek is not underpowered (UP) compared to Eggy.
I think win %'s are an aberration as each player who specializes in a particular civ tends to move towards a rating where he wins half his games.
You can't really look at the higher level players as there are so few of them and often there is 100-200 spread in ratings in matches, so the civ of the high-ranking player is almost certain to win, whatever it is. Small effect on overall stats probably.
As to the higher level players, if they all play Isis, there will be an exactly 50% win rate. We don't even need to count those games as they say nothing about OP or UP.
What I want to see is something like eX_Hope as Thor against MOR_Romio as Isis. Whatever the outcome of one game would anyone want to say one god was OP, one UP, or one player was really better than the other? G3_, maybe they'd both lose and the OP Eggy and Norse would both become UP
The work of science is to substitute facts for appearances, and demonstrations for impressions. John Ruskin