Probability of being a long-term winner

[Albert Silver]

I'm not a mathematician just to be clear (high school algebra is as good as it gets). My question is this: suppose I am playing SNGs, 500 of them, all of the same value. My ROI is a fantastic 20% (this is all 100% hypothetical). How certain is this? What if it were only 10%? or 5%? Would I need a larger sample? 1000? 2000?

I am familiar with this via chess software ratings (ex: here) but not in poker. Since obviously this will vary according to the game, let us suppose 9-10 player SNGs.

Any light shed would be enormously appreciated.

Albert