My question is what is a good sample size to know whether you are winning at certain stake or not and/or to know an edge in terms of bb. 

What is the variance in this, for example if I have a sample size of 50k at NLHE 6-max how likely does it distort my real edge say 50%? Or is the normal variance say 20% with a sample size of 50k hands? Obviously it gets closer the truth when you have more hands to analyze. There must be some formulas for this :)

Appreciate all the answers.