(12-10-2017, 09:22 PM)Goose link Wrote:[quote author=stupac2 link=topic=18293.msg217846#msg217846 date=1512963513]
The way to do this is with Z-scores (which normalize things to number of standard deviations away from the mean) by year, but compiling that data with college stats would probably be a royal pain. With cfbstats it would be easy, but of course that only goes back to 2008.
I also think that that normalization only works if the distributions are the same (like both Gaussian). I would strongly suspect that over a 46 year period, the points per game distribution in 2017 would have a much different shape than the one in 1971. As you say though, even determining that would be tons of work.
[/quote]
1. You are asking for too much here. It's absolutely true that no number can exactly tell you the relative strength of two offenses in two different years. That doesn't mean you throw up your hands and give up. It means you treat the numbers that result with some care.
2. College football hasn't changed
so much since 2000, though indeed offenses are inching ahead. I have read several places that offensive production numbers are up about 5% over 20 years. Annoyingly, it's very difficult to find Division 1 team-ranked stats before 2003, but 2003 was pretty close to 2001. Here are the medians on some relevant stats that are available on teamrankings.com.
Yds/game
2017 = 397.5;Â 2003 = 375.1;Â difference then vs now -5.6%
yds/play
2017 = 5.45;Â 2003 = 5.2;Â difference -4.6%
pts/game
2017 = 28.25; 2003 = 26.6, difference -5.8%
pts/play
2017 = 0.382; 2003 = 0.362;Â difference -5.2%
So, the 5% seems about right, particularly for rate stats.
I would conclude that if you care about Stanford's place versus other teams, you should throw in a cushion in the numbers I posted of about 5%, pro-Willingham.
BUT, the Willingham numbers also don't include Bowl games, because the NCAA used to say those didn't count in season stats, but now they do. Willingham's best two Stanford teams had very disappointing Bowl games in terms of offensive production - per play is too much work but per game the Rose Bowl dropped Stanford 1999 per game avg by 3.5% and the Seattle Bowl the 2001 per game average by 2%. So for those two years the productivity stats with bowls should only get about half that kicker.
-------------------
One shouldn't be trying to read these closer than a 5% range anyway, so none of these adjustments should make much of a difference for what conclusions you draw.