[SoftwareMaven]

The problem is you are using a rounding of the balance across the million tries to decide on game B. The choice of which game B to use need to not be dependent on the other experiments.