When establishing dual power ratings, an offensive and defensive power rating is created for each team and adjustments are made when the final score is more than seven points away from the predicted final score in football and 10 points away from the predicted final score in basketball.
Early in the season the maximum adjustment can be up to three points down to a maximum of one point later in the year.
One of the great things about these power ratings is they can be started at any point in the season by using scoring stats, although starting before the beginning of the season by merely using last year's statistics is probably best and also incorporates another one of Barnes' concerns, which is people tend to move ratings too quickly.
Creating the Numbers
Creating an offensive number is probably the easiest of the two, as you can simply begin by using a team's scoring average. Don't worry about the competition level just yet, even though at some point it must be considered. Just because Duke and Lamar are both averaging 83 points a game does not make them equal offensively, as Duke is obviously superior, having averaged its 83 points against much tougher competition.
Defensive numbers are a bit trickier, as we first need to know the median points allowed by each individual conference, which will require a bit of work. The reason we're using each conference, as opposed to the entire NCAA as a whole, is that certain conferences tend to be higher or lower scoring than most others.
If we applied stats from the entire NCAA we would be predicting lower scores than would reasonably be expected in the lower-scoring conferences and higher scores than we would get in the higher-scoring conferences.
After calculating the median points allowed by a conference that number becomes the base of 0. A team allowing the exact number of points as the conference median would get a defensive rating of 0. A team allowing five more points than the conference median would receive a defensive rating of 5, while a team allowing five fewer points than the conference median would receive a defensive rating of -5.
Then you would simply add a team's offensive rating to its opponent's defense to come up with your predicted score for the game. If a team has an offensive rating of 72 and is playing a team with a defensive rating of 3, we would predict that team to score 75 points. If a team with an offensive rating of 72 is playing a team with a defensive rating of -3, we would predict that team to score 69 points.
Home Team Advantage
Home court/field advantage is factored in by the oddsmakers, so it is something that should factored in by us, as well. For simplicity, the easiest number to use is four, although there are certain teams which historically have a greater, or lesser, advantage when playing in front of the hometown fans.
Don't merely add four points to the home team, however, as that disrupts your predicted total points. Subtract two points from the road team and add two points to the home team so that your predicted total points remains the same.
Non-Conference Games
The method as described above will be sufficient for dealing with conference games. And to be honest, that is its greatest use. The obvious problem is when teams venture out of their conference, especially when playing an opponent who has played a greater or lesser strength of schedule.
What I would do is use Jeff Sagarin's USA Today conference rankings as a guideline and look at his conference rankings, where he lists all 33 conferences in order, based on mean power rating. Simply make an adjustment of one point for each ranking difference one conference has over the other.
As of this writing, Sagarin has the Big East ranked first, the Colonial Conference ranked 11th and the Atlantic Sun ranked 21st. If a Big East team was playing an Atlantic Sun team, I would do my calculations by using my current numbers and then make a 20-point adjustment in the same manner as factoring home court advantage by adding 10 points to the Big East team's predicted score and subtracting 10 from the Atlantic Sun's team. If a Colonial Conference team was playing an Atlantic Sun team, there would be a 10-point adjustment, which would call for adding five points to the Colonial team's predicted score and subtracting five from the Atlantic Sun representative.
Those who love to play with numbers will probably enjoy this method, although it will be a bit much for the majority of people. Still, it's something that can be considered by those who are looking for every advantage they can get.
