A win is worth the weight of the defeated team, a loss worth the negative reciprocal (not exactly, but it's an inverse function). So a win over the median team is worth 1 a loss -1. A win over the #1 team is worth 2, a loss 1/2 (not exactly, but you get the idea).
A dominant win is worth more. A close loss is less negative. There is a floor, so a win over a really bad team is worth at least a minimum amount.
All the wins and losses, plus the initial weight, are averaged, and the schedule strengths are recalculated. That's it.
The purpose of the system is to provide an unbiased alternative to the AP poll. It is not a predictive system, and does not pick winners or point spreads. It gives a top 25 ranking, which does not reward running up the score (as the AP does), does not value late season games more than early season games (as the AP does), does not give equal credit for beating lousy teams as for beating moderately decent teams (as the AP does).
Hope you like it.
---------------- Roger Dendy Analex Corp. (216) 433-2186 firstname.lastname@example.org
Roger P. Dendy / email@example.com