Funny. If one penalty doesn’t get called late in the Iowa game, it probably would have projected us to go 9-3. A lot of these things are so reactionary.
Those projections usually account for points differential and/or per play net efficiencies, not just simple W-L records. Crushing Alabama by 30 like Clemson on a neutral field earlier this calendar year is more impressive than squeaking by them by 1 at home when it was raining, somebody was hurt, and other weird stuff happened with the game.
Those "quality of play" not just W-L statistics pick up on that better. Winning a close game versus losing a close one should not radically alter the forecast. We needed a very small amount of points or net yards per play to beat Iowa, so our projections would not change much.
The countervailing effect, however, is the college football season is so short. One game is 8.3% of your total sample size of 12. The Drake game last year, for instance, absolutely
killed us in the advanced metric statistics once the season was over. So you have to make pretty radical revisions pretty quickly because only limited data comes in because N=12.
This is not like the NBA where if, for instance, LA was to lose to a bad team like the NYKs one night. The season is 82 games long. Unless LeBron and AD struggle in the long-term, one game like that is kind of meaningless. One college football game is the equivalent to 6.8 NBA games or 13.5 MLB games, so you have to put more into each of them.