How many data are enough to determine a probability distribution?

The probability distribution is more correct if my dataset encompass 200 data than with 20 points.

So the answer may be the more the better. But is there any conventional thought?

Thanks