The Impact Prediction System (IPS) is designed to project the score of a game based on each teams’ prior performances. Whenever you start comparing statistics, you always run into the objection that it depends on the competition. So, the IPS factors in the averages of the respective opponents. For instance, it determines that an offense scores 19 more points per game then its competition normally allows. This is called the Offensive Impact. Similarly, the IPS calculates the Defensive Impact (e.g. a defense may give up 12 points less than their opponents are used to scorring). The IPS then uses these Impact factors to predict the expected scores.
There are some caveats, limitations, and adjustments to the IPS. First of all, the IPS adjusts the score for a homefield advantage determined be 1-5 points based on the rankings of the team (figuring that higher ranked teams have an enthusiastic fan base to create the advantage). The IPS makes adjustments for the affects of playing Div II teams as well as BCS vs. non-BCS teams. Finally, a statement of the obvious, but the IPS predicts scores purely on prior performance and does not account for major changes (e.g. a season ending injury to a key player, weather, ….)
BTW, the first time I used the IPS was to predict the score of the UT-USC championship game. It predicted with UT = 44, USC =37. Pretty close and predicted the wrong team was favored outright. So, I’ve been toying with it ever since.