To explain the columns below, Median is a median point spread, and can be used to get a feel for how good a team is without overly weighting a blowout win or blowout loss. HS is Brian Burke’s Homemade Sagarin, as implemented in Maggie Xiong’s PDL::Stats. Pred is the predicted Pythagorean expectation. The exponent for this measure is fitted to the data set itself. SOS, SRS, and MOV are the simple ranking components. MOV is margin of victory, or point spread divided by games played. SOS is strength of schedule. SRS is the simple ranking.
In median point spreads, the top three are Green Bay, Houston, and New England. Pythagoreans favor Green Bay, San Francisco, and Houston. On top of SRS are Green Bay and San Francisco, no other teams are even close. The third highest is now Chicago, still sporting the highest strength of schedule of them all.
November 19, 2011 at 1:42 pm
First time I’ve been to your site, thanks for the post. One thing to mention, Tampa Bay actual has a higher strength of schedule.
Also, you have homefield advantage as 2.78…I assume that is just for this season, correct? Do you believe that’s the “correct” long term number?
November 19, 2011 at 1:50 pm
Ryan,
Home field advantage in this instance is calculated via the “Home made Sagarin” technique, which makes it data set dependent and sensitive to the specific implementation of the least squares code. If you go through the calculated season stats, you’ll see it varying quite a bit.
I prefer personally, to look at total games and home wins, and think of the home field advantage in terms of a percentage chance to win at home. That’s far less model dependent and it is easy to calculate.
It also works better with logistic regression techniques, and probabilistic analysis.
SOS, in my case, is what I get from calculating simple rankings. And no matter what people think of it, my logistic regressions show it’s a provably real factor in predicting NFL playoff wins.
Hence my interest, and why I calculate it.
D-
November 20, 2011 at 1:22 pm
Another instance it would be excellent to have a link to the perl script you used to generate this. Thanks!
November 20, 2011 at 7:41 pm
As components of the code used to generate these stats have already been published (medians in my old NFL Stats 101 in Perl article, and Pythagoreans, and SrS is now a CPAN module), probably not.
The Homemade Sagarin has implementation issues. The ols implementation in Perl/PDL tends to blow up, and the fall back is the Nelder Mead nonlinear simplex. The simplex has its own set of problems. SciPy, interestingly, has access to Powell’s conjugate gradient method, which would be interesting to implement, if life gave me enough time to explore Python.
D-
November 21, 2011 at 2:14 pm
I’ll refer to your previous article and CPAN module. I’d much like the Homemade Sagarin in a repeatable script, but can understand the challenges. No experience in Perl/PDL, so that might be an issue.
I run a simulated football game called http://deeproute.com and I’d love to take the Homemade Sagarin and use that for each of the simulated leagues to generate rankings. I already use Pythagorean for win totals.
Perhaps the SrS will be useful instead.
Thank you so much for the great work you do!