Pages

New Seed Projections

3/03/2006

I tweaked the methodology used to come up with my latest seed projections. After doing some analysis of previous seasons, it's clear that the selection committee values quality wins more than they penalize quality losses. In other words, if the JCI correctly equates 2 teams, one with a couple more quality wins but some ugly losses, and another without as many quality wins but with no ugly losses, the Committee will likely pick the team with the greater number of quality wins. I don't know if I necessarily agree with the logic, but I wanted to give a boost for quality wins in order to have a way to systematically adjust for this known bias.

So the overall JCI still measures the quality of one team's complete season of wins and losses versus another's, similar to what the Committee would do if they strictly adhered to their 'body-of-work' principal. But since we know that this isn't true, my projected seeds are based on a combination of the three following components:

1. Raw JCI
2. Performance over the last 10 games
3. Number of Quality Wins

A fourth factor could easily be added -- inaccuracy of the RPI numbers -- in order to try to capture the fact that the Committee will be looking at numbers that will artificially make some teams look better than others. I didn't want to go down this path because a.) it's difficult to predict how much the Committee will be able to account for this and b.) i rather stick with ranking the teams on whether they 'should' get in, rather than whether I think they 'will' get in. Whether they should be included or not, factors #2 and #3 above are at least tied to performance. The RPI error is artificial. I'll save the Selection Sunday predictions for Joe Lunardi.

The quality wins factor will hurt teams such as George Washington, Gonzaga and Bucknell, all teams with strong records but lacking in big wins. Even with the penalty, George Washington currently isn't dinged enough to move off my latest '1' line, although Ohio State is very close to passing them for the last one seed. Again, I doubt GW will get anything higher than a '2' or '3' seed, whether they deserve that much of a penalty or not. It is interesting to note that despite most bracket projections having Gonzaga a '2' or '3' and GW a '3' or '4', you'd have a difficult time making a case that GW should deserve to be any lower than Gonzaga. My measure of quality wins (each victory get a fraction of a win assigned to it based on the expected probability of victory given the quality of opponent and location) has Gonzaga 36th overall and George Washington 39th, nowhere near enough difference to offset Gonzaga's 3 losses (vs UConn, at Washington, at Memphis) versus GW's 1 loss (at NC St).

1: Connecticut, Duke, Villanova, George Washington
2: Ohio State, Memphis, Illinois, Texas
3: Gonzaga, North Carolina, West Virginia, Pittsburgh
4: Iowa, LSU, Georgetown, Tennessee
5: Boston College, North Carolina St, Michigan St, Florida
6: Marquette, UCLA, Oklahoma, Kansas
7: Michigan, Wisconsin, Kentucky, Arkansas
8: Indiana, Wichita St, Bucknell, Cincinnati
9: Washington, Nevada, Missouri St, Alabama
10: Northern Iowa, Florida St, Syracuse, George Mason
11: Texas A&M, Creighton, Seton Hall, Hofstra
12: UAB, Arizona, UNC Wilmington, Kent St
13: Western Kentucky, San Diego St, Northwestern St, Wis-Milwaukee
14: Winthrop, Murray St, Pacific, Manhattan
15: Pennsylvania, Georgia So, Siena, Oral Roberts
16: IUPUI, Wis-Green Bay, Denver, Lipscomb, Southern U

Last 4 In: Seton Hall, Hofstra, UAB, Arizona
First 4 Out: Vanderbilt, Louisville, Bradley, Maryland
Next 4 Out: Southern Illinois, California, Colorado, Notre Dame

0 comments:

 
Wegoblogger #31 © 2011 | Designed by Bingo Cash, in collaboration with Modern Warfare 3, VPS Hosting and Compare Web Hosting