For anything short-handed I would agree that win rates come together at no less than 100,000 hands. Obviously a 10,000 hand sample is not enough to determine this.

Yes, variance is a b****.