|
Post by Rich Kern on Sept 4, 2006 18:39:47 GMT -5
I don't know if I'll post this on my site but here is a very early season ranking for the RKPI. Keep in mind that the RKPI doesn't care how good a team is supposed to be. It only cares about "what have you done for me lately". This early in the season (and even later for that matter) you can get some rather strange results but based on limited current results, it has a "reasonable" shape to it.
1 Nebraska 2 UCLA 3 Purdue 4 Texas 5 LSU 6 Illinois 7 Arizona State 8 Texas A&M 9 Florida 10 Penn State 11 Oregon 12 Tulane 13 Santa Clara 14 Wisconsin 15 USC 16 BYU 17 Tennessee 18 Long Beach State 19 Hawaii 20 Arizona 21 Florida State 22 UConn 23 San Francisco 24 San Diego 25 Washington 26 Ohio State 27 Stanford 28 Wichita State 29 Notre Dame 30 Missouri State
|
|
|
Post by Rus Socoli on Sept 4, 2006 19:22:53 GMT -5
Where is Cal? Aren't they supposed to be undefeated also?
Tulane?
|
|
|
Post by canyoudigit on Sept 4, 2006 19:29:49 GMT -5
Last time I checked, PSU beat LSU, but they are ranked behind them? What is that all about? And Florida State has played the sisters of the poor, what gets them so high? I know it's early, but some of these don't make any sense.
|
|
|
Post by Revbang on Sept 4, 2006 19:37:13 GMT -5
Gotta love the Illini at #6!
|
|
|
Post by The Bofa on the Sofa on Sept 4, 2006 19:39:12 GMT -5
Last time I checked, PSU beat LSU, but they are ranked behind them? What is that all about? And Florida State has played the sisters of the poor, what gets them so high? I know it's early, but some of these don't make any sense. You can find descriptions of RPI around the web. If you aren't familiar with it, I recommend you check it out. Head-to-head matters in RPI because it affects the winning percentage of the team in question, but that effect is tempered because it decreases the winning percentage of the opponent. But since opponent's winning pct is 50% of the RPI rating, this early in the season, it is not unreasonable that a head-to-head win will actually hurt your ranking in RPI. However, keep in mind that RPI is not intended to be meaningful in any way at all this early in the season. In fact, the NCAA does not consider the RPI to be all that meaningful anywhere during the season, and that it should only be consider when the season is complete. Personally, I am impressed that it came out as well as it did, considering it is only about 6 matches/team. Pablo probably won't be available until tomorrow evening.
|
|
|
Post by Rich Kern on Sept 4, 2006 21:52:54 GMT -5
Where is Cal? Aren't they supposed to be undefeated also? Tulane? Yes Cal is undefeated but the only opponent with a winning record is Florida International so their wins are not very impressive at this point. Tulane only has 3 matches so far so hardly anything to go by.
|
|
|
Post by Rich Kern on Sept 4, 2006 21:58:36 GMT -5
Last time I checked, PSU beat LSU, but they are ranked behind them? What is that all about? And Florida State has played the sisters of the poor, what gets them so high? I know it's early, but some of these don't make any sense. It doesn't look at head to head at all. It just sees records, your opponents records and their opponents records. LSU has actually played some pretty decent competition and beaten all but Penn State (that went 5). Penn State on the other hand has playe mostly relatively weak opponents. Some things will start to iron out better even by next week but as Pablo pointed out, the NCAA doesn't use their results during the season. Still, it is fun to see the place they start out and the direction they head throughout the season.
|
|
|
Post by Mr. Wednesday on Sept 4, 2006 23:57:10 GMT -5
A brief note about my qualifications prior to engaging in some technical arguments... I'm very familiar with the general form of the RPI, and the specific implementation used for hockey (which I presume to be the same across other sports, modulo secret ad hoc bonus factors like in basketball). I have a general interest in ranking systems that improve on RPI, which I consider to be outright ad hoccery. Head-to-head matters in RPI because it affects the winning percentage of the team in question, but that effect is tempered because it decreases the winning percentage of the opponent. Except that, in the NCAA's RPI formulation, head-to-head games are not counted when computing the owpct part. (edit: incidentally, this feature makes it deceptively difficult to calculate; although the formula appears to be simple, it basically requires a computer program to do it right, pencil and paper is completely impractical.) It's always better to win than to lose, but it's possible in extreme cases to lose more in the hit to owpct than you can make up in wpct by winning (that is, the team was guaranteed to lose ground just by playing the match). The NCAA is actually fixing that problem for hockey this season, although on an ad hoc basis (but then, what else would you expect with the RPI...). While I agree that it's a bad idea to draw too many conclusions early in the season, where the amount of data for analysis is extremely limited, I also think it's a bogus argument to try to blow off symptoms of real problems with the calculation by claiming (as the NCAA does) that it's only "meaningful" when the season is complete. Even if there are odd outcomes during the season (in the sense that they fly in the face of conventional wisdom), they should always be rational, which is not always the case with RPI.
|
|
|
Post by OverAndUnder on Sept 6, 2006 10:40:32 GMT -5
Quality post. Welcome to the forum, Mr. Wednesday.
|
|
|
Post by reiker21 on Sept 6, 2006 11:48:34 GMT -5
Gotta love it. I think this is a accurate indication what might be happening this season.
|
|
|
Post by BearClause on Sept 6, 2006 11:55:51 GMT -5
Where is Cal? Aren't they supposed to be undefeated also? Tulane? Yes Cal is undefeated but the only opponent with a winning record is Florida International so their wins are not very impressive at this point. Baylor is now 7-1. Still - it's early in the season and I realize that Cal's opponents haven't really been world beaters yet.
|
|
|
Post by Pirate VB Fan on Sept 6, 2006 12:03:43 GMT -5
The NCAA is actually fixing that problem for hockey this season, although on an ad hoc basis (but then, what else would you expect with the RPI...). Wouldn't that be "an ad hockey basis"?
|
|
|
Post by reiker21 on Sept 6, 2006 13:07:39 GMT -5
This should prove to be an interesting season. I love upsets
|
|
|
Post by BearCat323 on Sept 6, 2006 14:28:06 GMT -5
I think Rich Kern just forgot about Cal. If Tulane only have 3 matches, what did you go on to put them at 12. It can't be the quality of the teams they played.
|
|
|
Post by BearClause on Sept 6, 2006 14:45:19 GMT -5
I think Rich Kern just forgot about Cal. If Tulane only have 3 matches, what did you go on to put them at 12. It can't be the quality of the teams they played. This is the RKPI, which is Rich's attempted duplication of the NCAA's Ratings Percentage Index. It generally takes nothing more than the won-loss records of a team and that of its opponents and opponents' opponents. It's a really blunt tool and many wonder why the NCAA even uses the RPI. Cal has only played two opponents with winning records, so it's going to be skewed until Cal plays teams with better records. The RichKern.com poll of 15 human voters has Cal at #12 and Tulane nowhere to be found.
|
|