If you have a normal distribution, the z-score for 1% (1 in 100) is 2.326. If your mean is 10 and 1% is 18, then your standard deviation is 8 / 2.326 = 3.4394, then 20 is approximately 18 in 10,000; 22 is approximately 24 in 100,000; and 24 is approximately 2 in 100,000.

If we go with 18 being 0.46 (3d6 being normal people, 4d6dl being exceptional people), then z is 2.605, and sigma is 3.071. Then 20 is 56 in 100,000, 22 is 5 in 100,000, and 24 is too low for the web calculator I found to calculate (so less than 1 in 100,000). But if you're assuming 400 million population there's going to be at least 400 people walking around with 24's.