The How of a Computer Ranking System

I cannot divulge every secret of my ranking system, or then anybody could just copy what I did. So, I am just going show you basically how it works.

First, I had to develop how the teams would be compared. I had some ideas like, compare rankings, compare wins and loses, compare score, compare defense, and etc. So, I tried to combine many of these comparisons into the current version of my rankings. Most of the major team statistical categories are compared.

• Scoring offense and defense
• Running offense and defense
• Passing offense and defense
• Time of possession
• Fumbles
• Interceptions
• First Downs

To compare these categories, the current game against an opponent is compared to the average of that opponent. Lets say, Michigan St. usually gives up 14 points a game, over 5 games. Then on their sixth game, Purdue scores 27 points on them Michigan St. will receive (14/27)*weight for that game. Then to find Michigan state's rank all the statistical weighted averages are summed.

Also, where a game is played (location) and when a game is played (date) will determine the win loss record as well. If a team wins more on the road then they are given more points than winning at home. Just as losing at home costs more points than losing on the road. Similarly, late season wins and loses help or hurt more than early season wins and loses.

The weights are a work in progress as well. Right now I feel that they are set at a good level. The category with they highest weight is wins and loses. It really doesn't matter how good your stats are if you never win any games. This accounts for almost half of the weighting. Next is scoring defense and then offense, followed by run defense and offense. The weights trail off after that.

That is pretty much how the rankings work. Later I will show you an example of using last years statistics.