Tuesday, April 15, 2008

Strength of Schedule Bias

Since the NFL is scheduling up right now, it's worth mentioning that strength of schedule measures are inherently biased. This is nothing groundbreaking-- good teams beat their opponents more than bad teams, so if the 49ers played the exact same schedule as the Patriots last season, the strength of schedule would seem less for the Patriots, simply because they're winning against them more often.

I'd assume you simply look at the games that don't involve your participation in them-- so 16 wins would be taken from the composite won-loss total for the opponents of the Patriots from last year to compute their year-end strength of schedule. I suppose calculating this year's before-season strength of schedule upon last year's win-loss results would be unbiased. I can't think of a large reason why they would be...unless considering a) you don't play entirely new schedules from year to year due to divisional games and b) there is some mild persistence of team ability across seasons. Take together, the bias would be presumably small. Previous season strength of schedule doesn't seem to predict much in the NFL anyway.

Of course, this issue carries a bit more weight in college football, where strength of schedules matter to determine which teams play for the national championship. I'm sure they adjust for this...right? How do they do that? (Holub, that means you.)

My preliminary guess as to the determination of SOS for college is this:

1) Spots 1-12 are reserved for SEC teams.
2) The next spots are reserved for teams that play against SEC teams.
3) The rest.


rolub said...

You're spot-on, except for placing Oklahoma 13th, followed by the teams that play the SEC, followed by Adrian Peterson's unwashed practice jersey, followed by the rest of the teams.

Unknown said...

You're forgetting Notre Dame. Football Jesus will smite you.