In my system - simplied of course, ratings are assigned and based on the delta and the actual difference in the game adjustments are made.
example:
TeamA 150, teamB 100, based on the difference in ratings the game should be about a 50pt difference. They play the game and the final score is 30 - 20 thats a 10 pt difference so there needs to be adjustments made:
choices are:
TeamA was ranked to high but TeamB was ranked just right - adjustments would be TeamA loses 40 points team B stays the same OR
TeamA was ranked just right and TeamB was to low - adjustments would be TeamA stays the same and team B gains 40 points OR
somewhere in between.
My system adjusts based on how far the actual deviates from the projected so in this scenario A would move down but not to much as they still won and B would move up somewhat because they played within 1.4 scores so the adjustments might be TeamA - 5.5 to 144.5 and team B + 8.5 to 108.5. But each week the measure gets better as the teams approach an accurate rating. The amounts of change can change week to week because there is some curve fitting going on in the back ground and there are 6 possible formulas that can be applied. And twice a year/season I take the current power rating and go back to week 0 and calculate the deltas all the way forward again.
