A small sample size would just imply greater deviation. There's no reason to assume the programmers outside your sample would be x10 better than those in your sample.
That only works when you can assume your sample is representative of the global population. Since people are not evenly distributed according to their abilities, that assumption will not hold.