Well, first off, nobody believes in evolution. That would be like believing that the sun rises in the morning and sets in the evening. Even on a cloudy day, you can determine that the sun is still there from simple observation and a bit of deductive reasoning.

That being out of the way, what you really have in mind is Mendelian genetics. Sadly, that does not work the way you are thinking. Allow me to explain.

Let's say that there is a gene for Atheism and a gene for religion. We will call them R and A for the father and r and a for the mother. This will yield certain breeding possibilities, specifically, the children must be one of the four following combinations:

Rr, Ra, Ar or Aa.

OK, now the Rr and Aa children are called homozygous because they have matched pairs of genes. Then the Ra and Ar children are called heterozygous because they have the same mismatched genes as the parents did.

Now, if the heterozygous children beget a second generation with other heterozygous individuals, you should see that the matter just continues. Where it gets interesting is when you decide to encourage breeding between homozygous children and heterozygous ones. For simplicity, I will assume that we can detect the Rr children and prevent them from breeding as is done with plants and animals that have undesirable genetic traits. Now the possibilities for the second generation are as follows:

Ar, Aa, Ar and Aa. For cases where the mother is heterozygous and the father is homozygous atheist.

Obviously, if both parents are Homozygous atheists, then all of their children must be Aa. There is no other possibility.

However, if the goal is to over compete with the theists, then the answer is to find a way to prevent the homozygous theists from breeding in the first place. That will provide a mathematical advantage once we start seeing the second generation homozygous atheists breeding with the heterozygous general population.