The average win rate for the sample is 53.54%.
Assuming in each game, one civ wins and one civ loses, this is an impossible result. The winrate aggregate among all civs should be 50%, not 53%, so we can safely assume every data point is inflated in winrate significantly.
When people say “there are lies, damned lies, and statistics,” they don’t usually mean “the statistics are clearly, unmistakably false.” Usually they mean someone’s decided to interpret the statistics in a different way and used the statistics to tell an incomplete story. These statistics are damned lies.
The reason is all these results are counted for players above 2k but not for players who played against 2k + players. So, we’re seeing a general inflation of winrates coming from players at < 2k vs players at > 2k where only the player with the higher rating is being counted.
Thus, 3.54% more civs win than civs lose in the sample, therefore every civ gets a winrate spike. odd that you chose to ignore the Indians and the Chinese, both above 59% winrate in the sample to pick on the Mayans, but again, this data is bunk, so it doesn’t matter in the slightest.