You may have noticed a bit more movement in today's selection and seeding results than you would expect after a Friday with few games involving top teams. Let me explain.
As I've described before (and others have unofficially confirmed), a team's performance in conference (both raw games +/- .500 and overall conference record factoring schedule strength) are important to the committee. I include conference play explicitly in our model using the RPI rank for conference games only and games +/- .500.
Unfortunately the conference RPI is not always indicative of reality early in conference play, much like regular RPI is not early in the regular season. For example, UT Arlington and North Texas were in the RPI top 15 on December 15 this season. Hardly indicative of reality. As a result, I have not included either of the conference attributes in our results this season.
But that all stops today.
It's always a struggle to decide when to flip the switch on the conference data. Ignoring it causes some odd results. (Vanderbilt and St. Mary's are probably seeded too high, and Notre Dame and Kansas State are probably too low.) However, using it before it settles can cause equally odd results, both because of small sample size and RPI's nature as being a better "blunt instrument" than a precise tool.
Let's look at an example of that. Texas lost their first conference game to Missouri (ranked around RPI #100 at the time), so they were 1 game under .500 (which at the end of the season is not a positive for at-large selection) and had a conference RPI rank of #193. Since 2000, no teams with sub-100 conference RPI ranking have been selected at-large. Obviously, that would have introduced a false measure (two, actually) of the Longhorns' performance to the CTD magic box. I would much prefer too few good attributes and possibly incomplete information over several bad attributes and misinformation.
My rule of thumb in previous seasons has been to wait until most teams, or at least those with a realistic chance at an at-large bid, have played half of their conference games. This weekend marks that milestone (or it's close enough, anyway), so I am now feeding conference games +/- .500 and conference RPI data to our magic box.
Hopefully that won't open Pandora's box.