Talk:Kled/@comment-27004415-20160828011105/@comment-28977071-20160901140156

Well, firstly, yes, the pick rates are practically equal, players tend to opt for in cca. 7% more cases than.

Secondly however, I don't think the pick rate influences the results in the way you suggested. Higher pickrate causes no one-sided statistical bias for winrate. Although not really exact, the fact that each game starts anew with "same" conditions makes it possible to consider it a single quasi-replication. That implies that a larger sample size increases the reliability of obtained winrate numbers.

In our case the relative ratio of pickrates is of little importance, but the fact that the pick numbers are only 60 and 64, is relevant. Although the sample is considerably decent, it introduces still a significant level of uncertainty regarding the winrates. The winrates become more reliabe once more high-ranked players play such games.

Plus there is the statistical bias I mentioned- technically those games will never be true replications of a single scenario, hence unseen variables might secretly shift results towards one extreme (for example a smurf player's preferred build path'd shift results in favor of his build). But those are the unomittable limits of applying statistics on the game. One never gets perfect replications.