What exactly is he basing this on?
His own thoughts.
So far in the top 25 there have been:
Week 1: 2 win by an unranked over ranked team; 10 FSU to Ga Tech (10.5pt dog), FSU to BC (16.5pt dog)
Week 2: 3 wins; NIU over 5 ND (28.5pt dog), Ill over 19 Kansas (4.5 pt dog), Syracuse over 23 Ga Tech (2.5pt dog)
Week 3: 0 wins
Week 4: 2 wins; Buffalo over 23 NIU (12.5pt dog), BYU over 13 K State (7.5pt dog)
Week 5: 2 wins; Kentucky over 6 Ole Miss (15.5pt dog), AZ over 10 Utah (7.5pt dog)
Week 6: 6 wins; Syracuse over 25 UNLV (5.5pt dog), Vandy over 1 Bama (22.5pt dog), Ark over 4 TN (14.5pt dog), Wash over 10 Mich (favored by 1.5), SMU over 22 LVille (6.5pt dog), MN over 11 USC (9.5pt dog)
Week 7: 1 win; ASU over 16 Utah (5.5pt dog)
Total: 16 wins
Teams who were these favorites who have lost
FSU: 1-5
ND: 5-1
Kansas: 1-5
Ga Tech: 5-2
NIU: 4-2
K State: 5-1
Ole Miss: 5-2
Utah: 4-2
UNLV: 5-1
Bama: 5-1
TN: 5-1
Mich: 4-2
LVille: 4-2
USC: 3-3
Those who pulled the upset who are now ranked: SMU, BYU, Illinois
Only 6 of these teams are still ranked (ND, K State, Bama, Ole Miss, TN, Michigan). There have been a couple of highlight games, same as there are every year. In 2021 there were 16. In 2022 there were 13, in 2023 there were 6. It's a relative thing where it felt so chalky last year with the top tier teams and this year it's been back to the normal variance we usually see with much of it being driven by that people are typically horrible at prognosticating which teams should be ranked and missing on teams like Illinois, BYU, and SMU who were actually better than the teams they beat (in addition to Ga Tech and BC who are better than a very mediocre FSU team)
To me his data doesn't have much actual support aside from a couple memorable ones (NIU, Vandy, Ark) that we see in typical years.
Further, 15 teams have made the CFP since its inception which is kind of nuts in 10 years. 7 of them are in the top 10 (top 10 teams who haven't made a CFP--> PSU, Miami, Iowa State)