Well, theoretically, they do-- but the problem is that any tool for that purpose MUST use examples and problem-solving scenarios that may (and almost certainly DO) introduce bias into the tool itself.
How does one measure cognitive potential with NO reliance upon previous experience? Answer: this is a trick question, because there really isn't any good way to do that.
Someone who has seen a particular type of puzzle has a significant performance advantage over someone who has not.
Ergo-- enriched environments lead to children who 'test better.'