Tuesday, March 10, 2009

Tweaking the rankings

I made some adjustments to the rankings, and consequently to the brackets:

-I increased the weight of opponents' opponents, so that each win or loss of an opponents' opponent counts for 0.4 wins, instead of 0.2. I also added a "quality win" bonus (win over team in top 50) and "bad loss" penalty (loss to team under 200). These changes had the affect of helping major conference teams and hurting teams from smaller conferences.
-I eliminated the scores of games against Division II or lower schools. For most schools this didn't make much of a difference, but it did help some schools who had played just one game against lower level teams because it helped their average. For some of the really small conferences, it hurt the teams because it lowered the overall number of wins in the conference, thereby rewarding fewer points for each win.
-Finally, I adjusted the victory and loss margin rewards/penalties so that they reflected game totals and win / loss totals more accurately. This generally widened the margin between the good and the bad teams.

-For the brackets, in conferences that had finished their champion as of Sunday night, I included them. If not, I used the highest rated team still in the tournament.

For our new readers who felt the need to offer some constructive criticism, please remember that the bracketology is not meant as a prediction of the tournament field, but rather is a projection of the field using only our ranking system, which admittedly is imperfect and is still a work in progress (as evidenced by the changes above). Clearly we don't think Utah State is worthy of such a high seed, or Niagara worthy of a bid at all. But what we are trying to do is create a mathematical model that will simulate the strength and weaknesses of all teams that we can all observe, and thereby predict future results.