drhoops

Fellow hoop junkies,

RPI has always bugged me because of its obvious flaws and how much the committee apparently relies on it for seeding and selection.

My biggest problems with RPI are:
1-You gain in ranking by playing a good team and losing to it
2-You lose in ranking by playing a weak team even if you blow them out
3-No allowance for home court advantage
4-There are other, more subtle flaws, such as the arbitrary weights that make sure that SOS is more important than win/loss and take records as whole (without checking who your wins and losses came against).

These end up favoring teams from big conferences, in my opinion, since they get a free pass to a strong SOS and can force smaller teams to play them at home.

I recoded RPI to be able to see how important these factors are. The first change is to ignore wins that end up reducing your RPI. If you win a game against a cupcake, then doesnt make you a weaker team. Also, ignore losses that end up raising your RPI. If you lose a game against a tough opponent that doesnt make you a stronger team. These games are discarded because we learn nothing new from them. This takes care of number 1 and 2 above.

This early in the season, for undefeated teams, this ends up ignoring all but their best win; for no-win teams this ignores all but their worst loss. I used the raw game result data from Ken Pomeroy (www.kenpom.com). Ratings to follow.

drhoops

Top 50
Rank Team (conf) W-L Adjusted RPI (RPI) SOS

1. St. Joseph's (A10 ) 7- 0 0.942634 (0.717695) 0.623594
2. Stanford (P10 ) 6- 0 0.929762 (0.690192) 0.586923
3. Creighton (MVC ) 6- 0 0.914286 (0.601713) 0.468951
4. Vanderbilt (SEC ) 7- 0 0.892857 (0.694558) 0.592744
5. Georgia Tech (ACC ) 8- 0 0.891518 (0.678990) 0.571986
6. Kentucky (SEC ) 5- 0 0.889583 (0.752817) 0.670423
7. Louisiana St. (SEC ) 4- 0 0.883333 (0.558681) 0.411574
8. Oklahoma (B12 ) 7- 0 0.882812 (0.605291) 0.473721
9. Auburn (SEC ) 7- 0 0.860000 (0.609541) 0.479388
10. Pittsburgh (BE ) 7- 0 0.828075 (0.576689) 0.435586
11. Florida St. (ACC ) 7- 0 0.825000 (0.683282) 0.577710
12. Mississippi St. (SEC ) 7- 0 0.819048 (0.589033) 0.452044
13. North Carolina (ACC ) 6- 0 0.818527 (0.708423) 0.611231
14. Gonzaga (WCC ) 7- 1 0.803013 (0.737730) 0.691974
15. Wake Forest (ACC ) 6- 0 0.797768 (0.709179) 0.612239
16. Cincinnati (CUSA ) 4- 0 0.787500 (0.604942) 0.473256
17. Virginia (ACC ) 5- 0 0.784014 (0.608111) 0.477482
18. Nebraska (B12 ) 6- 1 0.771597 (0.669290) 0.606672
19. Murray St. (OVC ) 5- 0 0.768750 (0.543796) 0.391728
20. Rhode Island (A10 ) 6- 1 0.768447 (0.663603) 0.599090
21. Dayton (A10 ) 8- 0 0.759226 (0.629981) 0.506642
22. Wisconsin (B10 ) 6- 1 0.759061 (0.647239) 0.577271
23. Purdue (B10 ) 7- 1 0.758968 (0.628696) 0.546594
24. Kansas (B12 ) 4- 1 0.757812 (0.697614) 0.663485
25. Boston College (BE ) 7- 1 0.755215 (0.674564) 0.607751
26. Rutgers (BE ) 5- 1 0.745441 (0.714770) 0.675249
27. Connecticut (BE ) 7- 1 0.743229 (0.625436) 0.542248
28. UCLA (P10 ) 3- 1 0.739915 (0.605808) 0.557744
29. Michigan (B10 ) 5- 1 0.739484 (0.630778) 0.563259
30. Providence (BE ) 4- 1 0.737756 (0.699945) 0.666594
31. Arizona (P10 ) 4- 1 0.734462 (0.673380) 0.631173
32. Southern Illinois (MVC ) 6- 0 0.731548 (0.646670) 0.528893
33. Duke (ACC ) 6- 1 0.722975 (0.658056) 0.591694
34. Texas (B12 ) 5- 1 0.722292 (0.614013) 0.540906
35. Detroit (Horz ) 4- 2 0.721165 (0.652401) 0.647645
36. Maryland (ACC ) 6- 2 0.714882 (0.672823) 0.647097
37. Illinois (B10 ) 6- 2 0.710556 (0.617634) 0.573512
38. Utah (MWC ) 6- 2 0.705122 (0.639948) 0.603264
39. Texas Tech (B12 ) 8- 2 0.704704 (0.647568) 0.596757
40. Brigham Young (MWC ) 4- 1 0.701971 (0.661683) 0.615577
41. Louisville (CUSA ) 4- 1 0.701825 (0.667266) 0.623021
42. Missouri (B12 ) 3- 1 0.690365 (0.628596) 0.588128
43. New Orleans (SB ) 3- 1 0.685069 (0.602713) 0.553618
44. Iowa St. (B12 ) 5- 0 0.683333 (0.594459) 0.459279
45. Syracuse (BE ) 3- 1 0.683036 (0.653888) 0.621850
46. Toledo (MAC ) 5- 1 0.678294 (0.584084) 0.501001
47. Lafayette (Pat ) 4- 1 0.675099 (0.606441) 0.541921
48. Wichita St. (MVC ) 5- 2 0.671396 (0.602953) 0.565843
49. Florida (SEC ) 5- 2 0.671220 (0.579072) 0.534001
50. Georgetown (BE ) 7- 0 0.666667 (0.528639) 0.371519


I will post these periodically if there is interest.

xubrew

i think people misunderstand the importance of the rpi and how it comes into play in the selection process. teams from stronger conferences can manipulate the system and get their rpis up because most of them win their ooc games, and that causes the rpi to increase with ever conference game they play.

the rpi is just one thing out of a list of several things that the committee looks at. a high rpi does not ensure anything if a team does not meet any of the other requirements.

btw, the strength of schedule does not in any way directly affect the rpi. it is the number of wins of your opponents that is important. at this point in the year, the rpi is not accurate because there just has not been enough data to make it take shape.

here is something that may interest you about the selection process and how the rpi comes into play along with everything else:

http://www.basketballboards.net/foru...63&forumid=147

__________________
IT'S MILLER TIME!!!

HKF

Next time post it on the Bracketology forum.

__________________
My favorite NBA players: Kobe Bryant * Carmelo Anthony * Ben Gordon * Deron Williams * Joe Johnson * Kevin Martin * Tim Duncan * Kyle Lowry * Thad Young * Amare Stoudemire * Dwight Howard * Josh Smith * Brandon Roy * LaMarcus Aldridge * Rudy Gay * Al Horford * Brandan Wright * Craig Smith * Renaldo Balkman

2008-09 Favorite Rookies: Jerryd Bayless (POR), Roy Hibbert (IND), Greg Oden (POR), Jamont Gordon (undrafted), Chris Douglas-Roberts (NJN), Joe Alexander (MIL), Patrick Ewing, Jr. (SAC)

College Basketball: "Where 7-Foot Bigs Foul Out Happen"

drhoops

Thanks for the links Xubrew. I am familiar with the selection process; I didnt mean that RPI is used exclusively, but it seems to play an important role for such a flawed rating system.

I was also referring for the RPI SOS which makes up 75% of the RPI score (RPI is 0.25 * win percentage + 0.5 * opp win % + 0.25 opp opp win %). The last two terms together are the RPI SOS. They have nothing to do with the quality of your team and everything to do with who you have played. So, going 28-0 or 0-28 against a schedule makes a difference of no more than 0.25 in your RPI score, while who you play makes up 0.75 of your score regardless of how you do against them.

Can someone please move this thread to bracketology? I wasnt aware that is the right place for this kind of stuff

HKF

No problem dude. I am trying to write up all the teams that will be in the running for at-large bids, but I will wait till the right before the end of December.

__________________
My favorite NBA players: Kobe Bryant * Carmelo Anthony * Ben Gordon * Deron Williams * Joe Johnson * Kevin Martin * Tim Duncan * Kyle Lowry * Thad Young * Amare Stoudemire * Dwight Howard * Josh Smith * Brandon Roy * LaMarcus Aldridge * Rudy Gay * Al Horford * Brandan Wright * Craig Smith * Renaldo Balkman

2008-09 Favorite Rookies: Jerryd Bayless (POR), Roy Hibbert (IND), Greg Oden (POR), Jamont Gordon (undrafted), Chris Douglas-Roberts (NJN), Joe Alexander (MIL), Patrick Ewing, Jr. (SAC)

College Basketball: "Where 7-Foot Bigs Foul Out Happen"

BlueBaron

Moved from College Hoops Central.

__________________