Tuesday, December 7, 2010

Regional Rankings Explained

I've already posted my regional RPI rankings for two weeks now, but I thought it would be good to take a step back and let you all know what I'm doing, and why I'm doing it.

When we boil it down, I'm trying to guess who's going to be selected to the NCAA tournament at the end of the year. In a way, it's like what Joe Lunardi does for ESPN with "Bracketology".

All of the information regarding the selection criteria and process can be found in the 2011 NCAA Division III Men's Basketball Championship Handbook. This is where I get my info.

The NCAA will compile and release regional rankings five times this year (my chart ranks three more teams per region than the NCAA will each week):
  • Wednesday, February 2
  • Wednesday, February 9
  • Wednesday, February 16
  • Wednesday, February 23
  • Sunday, February 27
That final regional ranking will be the one that the selection committee uses to select and seed teams for the tournament. My goal is to predict what the rankings will look like each week (with as little weekly effort as possible).

Here's the NCAA's ranking criteria:
The primary criteria emphasize regional competition (all contests leading up to NCAA championships); all criteria listed will be evaluated (not listed in priority order).Win-loss percentage against regional opponents.
  • Win-loss percentage against regional opponents.
  • Strength-of-schedule (only contests versus regional competition).
    • Opponents’ Average Winning Percentage (OWP).
    • Opponents’ Opponents’ Average Winning Percentage (OOWP).
  • In-region head-to-head competition.
  • In-region results versus common regional opponents.
  • In-region results versus regionally ranked teams.
They give no indication of how they compare or weigh all of the criteria.

What I decided to do was compare all the teams based on their in-region RPI (that's ratings percentage index, but it's not important). RPI includes the first two criteria winning percentage (WP) and strength of schedule (SOS); it's defined as 0.25 x WP + 0.75 x SOS.

That's all my spreadsheet is doing right now; last year this method correctly picked 16 of 19 Pool C (at large) tournament teams.

This year, I'm hoping to add in some sort of adjustment for results versus regionally ranked teams (vRRO on my chart), but that will have to wait until the NCAA actually starts ranking teams. It'll be a sort of guess and check method to see if there is any discernible pattern.

So now I have teams ranked (according to region) based on their in-region RPI. From there, I attempt to project the tournament field.

42 conferences have the ability to award an automatic NCAA qualifier (this is called Pool A). The MIAA (like every conference except the UAA) awards this to the conference tournament champion. Since I have no way of knowing who will win the tournaments, I "award" Pool A bids to the team that currently has the highest RPI in the conference. I figure they should have as good a chance as anyone to win the tournament, and I'm not going to spend the time to figure out who's in first place each week for each league.

This season, one bid will be reserved for teams that are not members of an automatic qualifying conference. This is called Pool B. I think there are 13 (or so) teams that are classified this way this season. I will award the top RPI team (among eligible Pool B candidates) with this spot. You can think of this group as one big conference that awards their bid via an at-large process.

Now, the NCAA will select a total of 61 teams this year, so we have 18 bids to had out. These 18 are the true "at large" (Pool C) selections. I take all of the teams that are not currently in line for an A or B bid, and rank them according to RPI. The top 18 are "in", and the rest are "out". It's that simple. I usually list each team's Pool C ranking under the "status" column.

So, as you can see, there are still relevant adjustments (such as head-to-head or common opponent) that can (and should!) be made to these rankings. I'm not trying to be definitive here, just trying to get as much (relevant) info out to the public as I can.

I'll try to refine the process as the year goes on. As I said before, I'm going to try to incorporate results versus regionally ranked opponents, and I'm even going to try to incorporate some secondary criteria. I'll be checking to see if any of these adjustments line up better with the rankings behind the scenes before I release the newer method.

Remember, these rankings only consider games that are defined by the NCAA as "in-region"!

Much special thanks to the folks at d3hoops.com. I get all of my schedule, W/L, and in-region determination data from them. Can't do anything I do without them.

No comments:

Post a Comment