There numbers used the F+ rankings as their base. They presume each individual team runs the table, including vs. the most likely conference championship game opponent:
Pythag is how good you are. A team with a Pythag of .750 would expect to beat an average team on a neutral field 75% of the time.
Percentile is a bit more complicated. I took a team with a Pythag of .900 and placed it against every contenders schedule. Let me give a simple example:
A team with a Pythag of .900 would go undefeated vs. University State’s schedule 10% of the time. They would go X-1 vs. the schedule 20% of the time.
With that information, if University State went undefeated, their Percentile would be 95% (right smack dab in the middle of that 10%). If they went X-1, their percentile would be 80%.
The big surprise to me is Notre Dame’s schedule. It’s quite brutal. I’ll see if I can calculate SOSMOV tomorrow to see how dominant they’ve been vs. it. I imagine SOSMOV is going to prefer Oklahoma.