By the Numbers Football Ratings
Welcome to By the Numbers, a college football rating site. If I were younger, I would call this my blog, because much of the reason for its existence is to have fun writing about college football. After a brief word about how the rating system works, you will find links to postings from various years.
The rating system used here is amazingly simple. I want to assign each team a "power rating" that can predict games well. If Clemson is rated at 114 and Auburn is rated at 107, I want that to mean that Clemson is 7 points better than Auburn (coincidentally, that's the margin of victory for the 2012 game). The home field advantage averages about 4 points, so my prediction would add or subtract 4 depending on where the game is played. Though the ratings are designed to predict games, they can also serve as a ranking system, since the higher the rating the better the team is.
Where do the ratings come from? I only use the teams' schedules, points for and against for the season and net wins (wins minus losses) for the season. That's all. No adjustments for when the games were played, where the games were played, who was injured, or anything else. I find it surprising that such a simple system actually predicts future games well (about 75% winners and a little over 50% against the spread). As a bonus, the derived rankings match the polls to a large extent.
There are different aspects of the ratings that I publish (starting in 2013). The offensive rating measures how many points each team should score against an average defense. The defensive rating measures how many points the opponents would score on this defense compared to an average defense. The points rating is the sum of the offensive and defensive ratings. The wins rating, described below in more detail, is the version of the ratings that ignores points and only uses wins and losses (and schedule). BCS computer systems were required to ignore points. The overall rating is a weighted average of points and wins ratings, with slightly more emphasis on the wins ratings.
You should view this rating system, or any other rating system, as a piece of evidence that might be useful to your evaluation of the college football system. Since I am telling you how this system works, you can correct for all of the missing details. This just tells you how the teams stack up .. by the numbers.
Here are some more details, if you are interested in the math. Let's say that Clemson has played 5 games in its season and has outscored its opponents by 60 points. Since Clemson averages 12 more points per game than its opponents, the most basic form of the ratings sets the Tigers' rating as being 12 higher than the average of its opponents' ratings. Mathematically, this can be expressed as a simple equation (T = (a+b+c+d+e)/5+12). Each team gets its own equation. Solving this system of equations can create a vicious circle, since to know Clemson's ratings I need to know the rating of the first opponent, but to know the rating of the opponent I will need to know Clemson's. Fortunately, mathematicians came up with techniques (linear algebra) to solve such messes. The downfall of this system is it gives no particular credit to winning. (Clemson could pound somebody by 80 points and then lose 4 in a row by a point each, finishing 1-4 but +76 points.) So rerun the system using only wins and losses, effectively counting each game score as 34-0. Then average the two ratings; 60% of the wins ratings and 40% of the points ratings has given the most accurate predictions (as determined by regression analysis, which is also where the 4-point home field came from). And that's it!
Here are past postings. I hope you enjoy the games!2015-16: 10-24