We're going to try a little experiment here.
I frequently have trouble posting on the blog because I try to do too much. So as we draw closer to Selection Sunday, I'm going to write smaller more frequent posts and lay off the essays.
I'll start by doing something I expect to repeat over the next three weeks and look at one team at a time. I'll shine some light on how the model is interpreting their profile, whether I think the model is right, and what the rest of the schedule has in store for that team.
We start this week with Wisconsin. Feel free to familiarize yourself with the Badgers' profile.
Several things jump out at me. First, Wisconsin is ranked in the top 30 in both RPI and Sagarin, and their RPI against conference (capturing imbalanced conference schedules) and non-conference opponents (capturing not only non-conference strength of schedule, but how they performed against that schedule) each rank in the top 40.
The nitty gritty data I use to train the model goes back to 2000, and of the 249 teams with an RPI rank of 40th or better and both conference and non-conference RPI ranks of 50th or better, only 6 (2.4%) did not make the field as at-large (or were automatic qualifiers and would not have been selected at-large based on their seeds).
|San Diego St.||2002||39||44||46|
As you may have noticed, all six are from non-BCS conferences, so as long as Wisconsin doesn't stray too far from their current ranks, they should be in.
The other is that the Badgers are 2-8 against RPI Top 50 teams. Two wins is pretty good (only about 10% of teams since 2000 ranked 50th or better with two or more RPI top 50 wins weren't at-large selections), but they also have eight losses (about 20% of teams since 2000 ranked 50th or better with two or more RPI top 50 wins and seven or more RPI top 50 losses weren't at-large selections). Combined with their 6-1 record against teams ranked 51-100, they still have some room for error in the closing weeks.