[WindyLakeReturn]

If we assume they are two normal distributed characteristics that aren't related, then it is a natural consequence once you select for those who are at the top of either distribution.

Someone in the top 10% of competency or charisma will have average distribution of the other. If you consider the full population, you'll find no relationship, but because you are only look at the top 10% of each, those in the top 10% of one will be average in the other and vice versa. While some (~10% of each top 10% group) will be represented in the other group, you'll still find that they are more likely to be at the bottom of the group in the second group.

The end result is that if you only plot the data for the top 10% of each group, there will be an inverse correlation.

This can hold true even if there is a positive relationship between the two characteristics depending upon the strength of the relationship and how selective you are in selecting the top of each group.

This leads to some interesting results, such as when fruits are grown for look and shelf life risk losing taste and flavor despite these factors being unrelated.