College Basketball Computer Rankings 2007-08

RPI
This RPI uses only a team's winning percentage, its opponents' combined winning percentage, and its opponents' opponents' combined winning percentage to determine a numerical score (the formula the NCAA started using for 2004-05 also takes road wins and home losses into account). The weights given to each component of the score are 25%, 50%, and 25% respectively. Various published RPIs may differ from the NCAA's (and each other) because of differences in computing opponents' and opponents' opponents' winning percentage. For example, if team X beat team Y, does team Y's loss to X count against X in X's opponents' winning percentage? The NCAA's answer (and mine) is supposedly "no."

Dr. Glenn's Power Index
Dr. Glenn's Power Index uses an iterated method to determine each team's power index. Before each iteration, each team is assigned the same power index (perhaps 60.0). During each iteration, each team is given a grade for each game based on the opponent's power index and the margin of victory (with some consideration given to not having blowouts affect the final power index). So a team may be graded higher for beating a good team by just 1 point than it would be for beating a bad team by 20. Once the grade for each game is determined, the average grade is determined for each team, and that value is used to recompute grades on the next iteration. All teams start with the same power index, and as we progress through the iterations, good teams will have their power indices rise while bad teams will have theirs decline. After enough iterations, things generally settle down and the final numbers are output. (One problem with my current method is that the numbers do not always converge -- some teams' power indices may rise and fall a significant fraction of a point over a period of ten or so iterations. In these cases, we have to stop after a large number of iterations and take average power indices over the last several iterations.)

The above is the basic idea for many other computer rankings; these will differ primarily in how the grade for each game is computed.

All in one(Power, RPI, Strength of Schedule)

Data Quality

Beware that I at the mercy of my source as far as the accuracy of my data is concerned. My source is known to report late west coast games twice or not at all, misreport games against non-Division I opponents, and occasionally transpose the winner and loser of a game.

Previous Years' Rankings

2007: rpi power 2006: rpi power 2005: rpi power (data for 2004-05 was lost and will be reconstructed) 2004: rpi power 2003: rpi power 2002: rpi power