Hi! If you're reading this, you've found my bracketology website. The goal of this site and the accompanying program is to procedurally generate a bracket every day based on a resume factors I've identified. The "Men's Bracket" and "Women's Bracket" pages show the bracket as it would be if the season ended today, while the "Men's Future Bracket" and "Women's Future Bracket" pages show my program's estimate of where each team will end up when all is said and done with the season, simulating the remainder of the games using KenPom.
The factors that I am using to generate the men's bracket, along with their weight in the program:
I arrived at the values for these weights through my own research and iterative testing with different combinations. Obviously, there is some overlap, so a win against a top 10 team will boost a team's score for both the Top 10 wins factor and the Top 25 wins factor, etc. The nitty-gritty of how this works is that a team's score for each factor is on a scale of 0 to 1, where 0 is near the lowest value a tournament-viable team would have on average (for example, 0 wins against top 25 teams translates to 0.0 for this factor) and 1 is the value a top-end team's resume would have (5 wins against top 25 teams translates to 1.0). The score for each factor is then multiplied by the factor's weight. Adding all those values up produces a team's total resume score from 0 to 1.
The quadrant records are a bit more complex than simply a team's record against only that quadrant. (The goal of the NET-based quadrant system used by the selection committee is to group a team's wins based on opponent quality. Quadrant 1 roughly corresponds to the NET top 50, Quadrant 2 the NET #51-100, Quadrant 3 the NET #101-200, and Quadrant 4 #201-362. Games played at home contain a penalty, meaning wins and losses look worse than they would on a neutral court, and games played on the road contain a bonus.) A team's wins count for all quadrants equalling and below the one it lands in. So, for example, Illinois defeating the #40 team on a neutral court would be a Quadrant 1 win, and my system would also count it as a Quadrant 2, 3, and 4 win. Conversely, losses count for all quadrants equalling and above the one it lands in. Illinois losing to the #40 team at home would be a Quadrant 2 loss, and my system would also count it as a Quadrant 1 loss. The point of this is both to stabilize team's records for quadrants (meaning one game will make up less of a team's record in each quadrant) and to more accurately reflect the value of wins and losses. Illinois shouldn't be able to say they have a 3-0 record against Q1 if they have 3 losses to Q3, and my system would see that as 3-3 for Q1.
The corresponding factors for the women's bracket:
Unfortunately, power rating stats are less readily available on the women's side, and I'm unclear on which the NCAA considers, if any. The NET is a decent enough approximation, but in general, the weights here are not as finely tuned as they are on the men's side, mostly due to lack of data.