beknighted
Joined: 11 Nov 2004
Posts: 11050
Location: Lost in D.C.
Back to top |
 Posted: 04/04/07 10:00 pm ::: This week's Fun with RPI |
Reply  |
|
[cross-posted from the ESPN board]
I'm doing this one last time for the 2006-2007 season. This week, it's a comparison between how well the seeding and RPI predicted the tournament.
The approach here is pretty simple - look at every game in the tournament and see whether the seeding or the RPI did a better job predicting the results.
In the first round, it was even. Both the seeding and RPI predicted 27 of the 32 games. In fact, they got exactly the same games right and wrong.
The second round also was even, with each predicting 11 of the 16 games. This time, the RPI correctly predicted the 4/5 game between Texas A&M and GWU that the seeding got wrong and the seeding correctly predicted the 4/5 game between NC State and Baylor that the RPI got wrong. They both missed Marist over Middle Tennessee, Mississippi over Maryland, FSU over Stanford and BGSU over Stanford, but I doubt a whole lot of people got even two of those right.
The Sweet 16 was like the first round. Both got 6 of 8 games, and they both missed the same upsets - Mississippi over Oklahoma and Rutgers over Duke.
In the Elite 8, each got two of the four games wrong. They both missed LSU over UConn. The seeding would have predicted Arizona State over Rutgers, while the RPI had it right, and the RPI would have predicted Purdue over Carolina, while the seeding had that one right.
In the Final Four, the seeding was wrong and the RPI was right in the RU-LSU game, and both seeding and RPI were right in the Tennessee-Carolina game. (We know this because Tennessee was wearing the home unis.) And in the final, both the seeding and the RPI were right.
If you've been keeping track, you know that the RPI got one more game right than the seeding. The RPI correctly predicted 48 of the 63 games, while the seeding got 47 of 63 right. That's not what you'd call statistically significant. What's more interesting is that, of the games that were predicted wrong, they missed the same game 12 times. In fact, the RPI and the seeding disagreed about the result of a game only 5 times out of the entire 63 games, or a bit less than 8 percent of the time. That seems like a strikingly low percentage to me.
Fun with RPI will be back in November. Until then, go watch some baseball.
|
|
