I’m not sure of the validity of ranking groups of teams vs each other when there is so little interleague play. There aren’t even enough data points in Power 5 or FBS to draw highly confident predictions so I’m not sure why he includes FCS. From a story on Sagarin:
For all of the advances, however, Sagarin and Harrell still keep things pretty basic. The look of the websites hasn’t changed much, if at all, since their inception, and Sagarin still does all of his work off of a DOS prompt on his home computer — using a formula that he hasn’t changed all that much since the 1970s.
“You could put all the mathematical whistles and bells in, and in terms of predictive accuracy for future games, you’re just getting marginal improvement,” Sagarin said. “So it’s the law of diminishing returns.
“Basically, who did you play, where did you play and what was the result, including the score.”
Sagarin’s ratings have held up well over time — but he cautions that no system is going to accurately predict the winner every time out.
“Real life is under no obligation to conform to mathematical models made by human beings,” he said. “Reality simply happens.”
Since the final score and margin of victory are such a key factor in Sagarin’s ratings, accuracy is key — one that’s not always guaranteed.
