Look, lemme try to explain statistics on a more basic format.
In order to give unbiased statistical data you have to take into account more factors other than x games played divide by x factions. Yes that is how you obtain a basic average, we all know this.
Basic statistics states that data cannot be represented in a skewed manner, meaning the matches played between each faction are completely uneven and not necessarily representing a good experiment.
For example, let's create a hypothetical experiment to put this into perspective. I'm a scientist, and i'm doing research on 3 regions to see which regions have the best long distance runners. Rather than lumping all runners into each region into a category, to remain unbiased I'd choose 10 runners from each region with olympic medals out of the 100 runners (age20-30) randomly, opposed to just picking all runners in each region and basing my study off raw data. This way the data is unbiased, otherwise you would run into the fact that there's an uneven playing field (i.e old runners, younger runners, non olypmic, olympic, in good health, not in good health)...
My basic point being the average (http://coh2chart.com/) is based off raw data... Which doesn't really represent which factions are better than the other.
This guy knows his stuff.