Remember using this very simple talent eval you can get to about a 70% prediction rate. Some teams / coaches fall within the 30% exception, obviously.
Attached is Baylor and Stanford's history to date. As you can see, 1) they haven't played any "top" teams and, 2) they don't recruit exceptionally well. But neither 1 or 2 matters because they don't play in conferences that recruit much better than them anyway. Baylor will play Oklahoma and Texas who out recruit them. Texas is a team who has a coach who has a drastic under performing effect. Similarly, Stanford will play Oregon and USC. Like Baylor only one of those teams has a modern history of using talent to their level of recruiting, or higher.
Either way, these teams sit having both won one game that talent predicted they should lose. BUT neither of their wins came over teams with exceptional talent gaps.