So, get this. My system that came from Pro Football Reference to project wins for college football teams was something I was unsatisfied. Mainly because no matter what, a team would start out at 4.07 wins, even if a team was winless. The result was a system that kind of looked like a joke. It didn’t reward teams that performed well enough, and it didn’t punish teams who performed badly.

And then, while reading Pro Basketball Reference, I found something that I knew I could tweak and make usable for college football.

It’s name is SPS, and it’s very easy to use, and easier to calculate. Regular SPS just uses weights of 6, 3, and 1 to calculate a players performance by using the last 3 years of data.

However, when I applied it to college football, I saw a few problems.

- It’s not detailed enough. I know the name is simple projection system, but still, just using the last 3 years of a teams wins-losses wasn’t giving me the results I wanted.
- It wasn’t taking Pythagorean wins-losses into effect. Pythag W-L is a better gauge of how good a team will do from year to year than actual wins.

So, with that in mind, I decided to kind of mesh up the old system with the new system, and I like the way it came out. It’s still mostly the same system, with some additions.

First off, I kept the weights of 6, 3, and 1. It still keeps it simple, and makes things easier on me to calculate. However, I practically added a second equation to original SPS to create SPS 2.0, which I believe will give a more detailed account of how many games a team will actually win next year.

The formula for SPS 2.0 is…

**((A * 6) + (B * 3) + (C * 1)/10 * .80) + ((A * 6) + (B * 3) + (C * 1)/10 * .20) = SPS 2.0**

The first part of the equation is for Pythagorean W/L, the second is for actual W/L. A is past year 1, B is past year 2, and C is past year 3. First off, the old equation gave weights of roughly .76 for pythag wins, and .24 for actual wins. I upped the importance of pythag W/L because the actual wins seemed to bog down the predictive record a bit too much with the .76 as a integer.

The yield for the raw equation is winning percentage, which I have deemed SPS Factor. This can be multiplied by 100 to give a power rating, or, you can multiply the raw sum by 12 to give predictive wins for next year.

I shall post the new results a little later.