So I did some experimentation to demonstrate that as the sample size increases the percentage moves closer to 50%.
I used a computer to randomly generate a number, either 1 or 2 (representing male or female).
Results by sample size (ss):
ss 10 = 70%
ss 100 = 55%
ss 1000 = 48%
ss 10000 = 50.22%
ss 100000 = 49.808%
ss 1000000 = 49.9077%
ss 10000000 = crashed the computer
In terms of distance from 50%:
ss 10 = 20
ss 100 = 5
ss 1000 = 2
ss 10000 = .22
ss 100000 = .192
ss 1000000 = .0923
ss 10000000 = crashed the computer
It is clearly getting significantly closer to 50% as the sample size gets larger. In fact even a sample size as small as 100 was pretty close to 50%.