I'm literally looking for an example of what people are talking about, because I don't know how you game these models. You always have to do better than the model predicts to move up, it's not enough to just beat some bad team by a lot. I looked at some mid B12 teams on Bart Torvik since that site gives you a score by game, and the B12 seems to be the conference where people bring this up the most.
Using Oklahoma as an example, one of the games that impacted their ranking most positively was beating #244 Texas St. by 39. 3 of the games that impacted their rating most negatively were beating #336 Central Arkansas by 16, beating #324 UT Rio Grand Valley by 24, and beating #330 Arkansas Pine Bluff by 21. Those are 3 of their worst 7 games of the year in the Bart Torvik model, even though they won all of them easily. The rest of their buy type games were scattered somewhere in between.
People act like all you have to do is play some crap teams and beat them by 20-25 and then the models will think you're amazing, when that's not close to true. Our 2 worst game this year on Torvik are losing to Creighton and Seton Hall. Our 3rd worst game is beating Mississippi Valley State by 34.