Monday, March 14, 2005

The RPI Project

Here's what I did: I took the NCAA bracket, ranked the teams via snake based on the order of the #1 seeds. Then I compared that list to both the "new" formula RPI ratings and the "classic" RPI ratings that I found here. The reason why I did this: the selection committee uses these rankings to select at large teams and to seed the ones in the tournament. However, I've always had an idea that the selection committee uses the ratings only when it suits their purpose, or offers them an argument/defense. I think if you're going to use something like that, you should use it fairly, across the board, and objectively. I had a feeling that they didn't, and I think I was right.

Now first, there's been some controversy over the "new" formula of RPI this year. I think the NCAA actually used something closer to the "classic" formula. The order of the actual bracket was significantly closer to the classic formula list than the new formula list.

So what's the point in all this? My hypothesis is to find out whether the RPI is an accurate way of determining what the best team is. Now, obviously there are some problems with trying to figure this out. I can't set up a bracket based solely on the RPI rankings. I'm stuck within the parameters of the current tournament.

Here's how I did it: by comparing the snake rankings of the actual bracket to both RPI rankings, I determined whether the selection committee overrated or underrated teams. Then we can watch whether the underrated teams are successful and whether overrated teams flounder. Of course, we're subject to the actual matchups (and unfortunately, there aren't many overrated vs. underrated games in the first round).

In each, I list the number of "seeds" and "ranks" a particular team is over or underrated. e.g., if Team X is the top 4 seed in the snake, but would've been the last 6 seed on RPI, the committee overrated it by 2 seeds and 11 "ranks".


cf New RPI
1. Vermont - 7 seeds, 30 ranks
2. Old Dominion - 5 seeds, 19 ranks
2. Ohio - 5 seeds, 19 ranks
4. UL- Lafayette - 3 seeds, 14 ranks
5. Southern Illinois - 3 seeds, 12 ranks
5. St. Mary's - 3 seeds, 12 ranks
7. UTEP - 2 seeds, 10 ranks
7. Utah State - 2 seeds, 10 ranks
9. Pacific - 3 seeds, 9 ranks
9. Pennsylvania - 3 seeds, 9 ranks
11. Wisconsin - 2 seeds, 9 ranks
12. Kansas - 2 seeds, 8 ranks
12. UCLA - 2 seeds, 8 ranks
13. Bucknell - 2 seeds, 7 ranks

cf Classic RPI
1. Vermont - 4 seeds, 17 ranks
2. UTEP - 3 seeds, 11 ranks
3. Kansas - 2 seeds, 9 ranks
3. UCLA - 2 seeds, 9 ranks
5. Boston College - 2 seeds, 8 ranks
5. Pittsburgh - 2 seeds, 8 ranks
5. Utah State - 2 seeds, 8 ranks
8. Cincinnati - 2 seeds, 7 ranks

Underrated by at least 2 seeds as compared to both RPIs
Kansas, Pacific, UTEP, UCLA, Vermont, Utah State


cf new RPI
1. Iowa State - 5 seeds, 21 ranks
2. West Virginia - 4 seeds, 16 ranks
3. NC State - 4 seeds, 15 ranks
4. Texas - 4 seeds, 13 ranks
5. Charlotte - 3 seeds, 13 ranks
6. Minnesota - 3 seeds, 12 ranks
7. Iowa - 3 seeds, 11 ranks
8. Georgia Tech - 3 seeds, 10 ranks
9. UAB - 2 seeds, 10 ranks
10. Connecticut - 2 seeds, 9 ranks

cf old RPI
1. Iowa State - 3 seeds, 12 ranks
2. Charlotte - 3 seeds, 11 ranks
2. Stanford - 3 seeds, 11 ranks
4. Texas - 2 seeds, 7 ranks
5. Georgia Tech - 2 seeds, 6 ranks
5. NC State - 2 seeds, 6 ranks

Overrated by at least 2 seeds as compared to both RPIs
Georgia Tech, Charlotte, Texas, Iowa State, NC State

Now, I am not advising anyone that these are prime "upset picks". My idea is solely to show whether the RPI serves a purpose or not. If all the underrated teams advance far and all the overrated teams fail, then we'll know whether the RPI serves a purpose (and the selection committee should use it). Of course, once I take a look at the actual bracket, I realize that this exercise may be for nothing.

The only matchups between an overrated team (by at least 2 seeds in at least one of the RPIs) and an underrated team are Texas Tech/UCLA, Syracuse/Vermont, Cincinnati/Iowa and in a way, Pittsburgh/Pacific (Pitt is overrated by 2 in the new RPI, underrated by 2 in the classic RPI). There are plenty of games between two overrated teams and two underrated teams. I'll update this for round 2.