blog | rpi | pomeroy ratings | stats

Wednesday, March 08, 2006

New vs. Old RPI

While you’re waiting impatiently (You are impatient, right? Tell me you are. Please.) for my latest at espn.com to get posted, you can check out our boy Yoni’s piece on how the road rage RPI is possibly affecting the selection process. Let’s just say the CAA will owe a big thank you to the new formula if they get an at-large bid or two.

Also, reader Sean McGrath took the log5 formula and applied it to the Pythagorean Ratings to come up with some probabilities for the Big East and ACC tourneys. Enjoy. 

Seed Team         Quarters Semis  Finals Champion
  1  UConn         100.00% 81.66% 61.21%  39.71%
  2  Villanova     100.00% 80.60% 52.64%  28.00%
  6  Pitsburgh      66.54% 37.86% 17.53%   7.66%
  3  West Virginia 100.00% 48.79% 19.52%   7.33%
  4  Marquette     100.00% 47.14% 14.17%   5.34%
  5  Georgetown     55.89% 30.99% 10.67%   4.56%
 12  Notre Dame     44.11% 21.86%  6.50%   2.42%
 11  Louisville     33.46% 13.35%  4.19%   1.22%
  8  Cincinnati     57.95% 11.77%  5.10%   1.70%
 10  Rutgers        51.81% 10.33%  3.33%   0.80%
  9  Syracuse       42.05%  6.57%  2.35%   0.63%
  7  Seton Hall     48.19%  9.06%  2.79%   0.63%

Seed Team         Quarters Semis  Finals Champion
  1  Duke          100.00% 83.06% 65.29%  40.83%
  2  UNC           100.00% 84.94% 66.47%  38.08%
  3  Boston Col.   100.00% 61.49% 18.74%   6.11%
  4  NC State      100.00% 54.43% 15.37%   5.34%
  5  Florida St.    67.02% 34.33% 10.01%   3.58%
  6  Maryland       61.73% 26.40%  6.65%   1.80%
  9  Clemson        51.78%  9.02%  3.97%   1.15%
  8  Miami          48.22%  7.92%  3.35%   0.93%
 10  Va. Tech       53.86%  8.60%  3.54%   0.81%
  7  Virginia       46.14%  6.46%  2.42%   0.50%
 12  Wake Forest    32.98% 11.24%  1.99%   0.45%
 11  Georgia Tech   38.27% 12.12%  2.18%   0.42%

Line o’ the Night 

                           FG    3pt  FT   Reb
                      Min  M-A   M-A  M-A  O-T  A F  S TO BLK Pts
Karandick Ogunride     32  5-10  0-0  8-12 9-16 0 0  2  4  0   18
Result: Win. Norfolk St. 73, UMES 71.

Posted on 03/08 at 12:45 AM
RPI • (1) Trackbacks • Permalink • E-mail me

Friday, October 14, 2005

The new RPI and its effect on scheduling

I love it when data speaks for itself. It doesn’t happen often, but I believe I’ve found one of those cases.

When it was revealed last summer that the RPI would be altered to give an undisclosed bonus to road wins and then when the formula was announced last summer, it was hailed as a boon for teams in non-power conferences. Not only would they be rewarded with a boost in the RPI for all of the non-conference road games they have to play, but they would also reap the benefit of power conference teams that want to take advantage of the road win component of the RPI. Power conference teams would in turn schedule a few more road games because of the new formula. Life would be great.

I recall this claim being rather common. Unfortunately, the only example I can find at the moment is from Dick Vitale.

[The new RPI] will become a factor as schools realize the added importance of winning on the road. Also look for this to influence scheduling in the future, because a big road win will help even more on the road to securing a bid to the Big Dance. Dick Vitale, espn.com, 12/27/04.

Dick Vitale didn’t get to where he is because of his journalistic prowess, so I feel a little sheepish about singling him out. And to be precise, he doesn’t specifically claim that power-conference teams will schedule more road games. But I received e-mails last season from people wondering if the new formula would spur the kind of movement that Vitale alluded to. Now that schedules are set for this season, we know the answer. 

Power conference non-conference games
              Home   Road   %Home
   2002-2005  1655    341    82.9
        2006   451     86    84.0

The numbers represent games played by power conference teams against teams from outside the power conferences.  By a power conference team, I mean a member of the ACC, Big East, Big 10, Big 12, Pac 10, or SEC. By 2006, I mean this season, and 2002-2005 represents the previous four seasons combined. The data ignores (nearly) all games played on psuedo-home courts and games against teams outside of Division I.

So with that said, nothing more needs to be said. I will add the caveat that this season’s data does not contain games that haven’t been scheduled yet, and I am referring to in-season tournament games that will played on home courts like the second round games of the Guardians Classic and Preseason NIT. I expect that this will push the percentage of home games incrementally higher for the power conferences.

We can only speculate on the reason for the lack of change in scheduling practices, but to me it’s clear. Teams with the big budgets are not willing to trade two or three spots in the RPI for the money that home dates bring in. This may come back to haunt one or two teams each season, thereby costing them revenue they would get from the NCAA Tournament, but most schools are willing to take that risk.

To give you an idea of how little things have changed, Syracuse will play their first game in a hostile arena the same night as Baylor - and due to NCAA sanctions, Baylor doesn’t start its season until the second week of January. Washington takes the cake by waiting until January 12th to hit the road.

For those that are curious, here is the breakdown of true home and road non-conference games for all conferences. Only games against Division I opponents are included. 

            Home  Road   %Home
Big XII       86    18    82.7
SEC           92    25    78.6
Big East     115    33    77.7
ACC           90    28    76.3
Pac 10        55    18    75.3
Big Ten       76    27    73.8
MWC           58    26    69.0
CUSA          73    48    60.3
Atlantic 10   82    57    59.0
MVC           40    29    58.0
WCC           44    40    52.4
WAC           43    40    51.8
CAA           40    43    48.2
Horizon       35    40    46.7
MAC           34    42    44.7
Ivy League    40    50    44.4
Patriot       35    47    42.7
Big Sky       28    39    41.8
Big West      32    45    41.6
Sun Belt      41    62    39.8
MAAC          27    41    39.7
SoCon         32    56    36.4
NEC           30    53    36.1
America East  30    54    35.7
Independents  49    98    33.3
OVC           14    34    29.2
MidCon        20    51    28.2
Southland     21    54    28.0
Big South     16    43    27.1
Atlantic Sun  17    48    26.2
MEAC          16    68    19.0
SWAC           6    60     9.1

Posted on 10/14 at 12:00 AM
RPI • (0) Trackbacks • Permalink • E-mail me

Monday, August 01, 2005

RPI Help

The Ratings Percentage Index (RPI) is the formula used by the NCAA to rate men’s and women’s college basketball teams. The formula is described below.

The RPI is calculated by adding three parts.

Part I (25% of the formula): Team winning percentage. For the 2005 season, the NCAA added a bonus/penalty system, where each home win or road loss get multiplied by 0.6 in the winning percentage calculation. A home loss or road win is multiplied by 1.4. Neutral games count as 1.0. More on the effect of these changes can be found here.

Part II (50%): Average opponents’ winning percentage. To calculate this, you must calculate each opponent’s winning percentage individually and average those figures. This is NOT calculated from the opponents’ combined record. Games involving the team for whom we are calculating the RPI are ignored.

Part III (25%): Average opponents’ opponents’ winning percentage: Basically taking all of the opponents’ Part II values and averaging them.

Only games against other teams playing a mostly D1 schedule count when computing the RPI.

Here’s a simple example of how it’s calculated:

Team A has played two games and beat Team B and Team C.
Team B has a 1-1 record. Team B beat a team with a 2-1 record and lost to 2-0 Team A.
Team C has a 1-2 record. Team C beat a team with a 0-2 record, lost to a team with a 2-0 record, and lost to Team A.

So now let’s see how to compute Team A’s RPI.

At 2-0, their winning pct. is 1.000.

Their average opponents’ winning percentage (OWP):
Team B is 1-0 (1.000) without the loss to Team A. Team C is 1-1 (.500) without the loss to Team A. So the OWP is (1.000 + .500)/2 = .750.

Their average opponents’ opponents’ winning percentage (OOWP):
To do this, we need Team B’s and Team C’s OWP. When removing Team B’s results, Team B’s opponents are 2-0 and 1-0 for an average of 1.000. Team C’s opponents are 0-1, 1-0, and 1-0 for an average of .667. The average of these two numbers is .833.

So Team A’s RPI is 1/4 X (1) + 1/2 X (.75) + 1/4 X (.833) = .8333

Here’s how other columns in the RPI grid are computed…

SOS (Strength of Schedule): This is the last two components of the RPI formula:
(2/3) X Opponents Winning Pct. + (1/3) X Opponents Opponents Winning Pct.

Non Conference RPI: This is computed applying the basic RPI formula only to a team’s non-conference games. A common misconception is that a team’s non conference RPI will remain the same after they have played their last non-conference opponent. This is not true because a team’s non-conference opponents’ records and their opponents’ records change as they play their conference games. This can have a substantial effect on the NCRPI. For example, in 2003 Alabama’s NCRPI had a ranking of #23 when they entered conference play. By the end of the regular season it had risen to #3.

L10 (Last 10 Games): A team’s record in their last 10 games against Division I opponents.

RD/NT (Road/Neutral Record): A team’s record away from its home court.

How important is the RPI?
From the NCAA web site:

The Rating Percentage Index (RPI) was created in 1981 to provide supplemental data for the Division I Men’s Basketball Committee in its evaluation of teams for at-large selection and seeding of the championship bracket.

The RPI is intended to be used as one of many resources used by the committee in the selection, seeding and bracketing process. It never should be considered anything but an additional evaluation tool. No computer program that is based on pure numbers can take into account subjective concepts, e.g., how well a team is playing down the stretch, what the loss or return of a top player means to a team, or how emotional a specific conference game may be.

Several independent elements are combined to produce the RPI. These elements are a part of the statistical information that can be utilized by each member in an objective manner.

Each committee member independently evaluates a vast pool of information available during the process to develop individual preferences. It is these subjective opinions--developed after many hours of personal observations, discussion with coaches, directors of athletics and commissioners, and review and comparison of objective data--that dictate how each individual ultimately will vote on all issues related to the selection, seeding and bracketing process.

While the various elements of the RPI are important in the evaluation process, the tournament bracket each year is based on the subjectivity of each individual committee member to select the best at-large teams available and to create a nationally balanced championship.

To read more about the details of the tournament selection process go to this page. Prepared to be bored though.

Why are the RPI ratings different on other pages?
I don’t know and I don’t care. For most of the 2005 season, every site (including this one) that posted the RPI differed significantly from what the NCAA was using. That saga is documented on this page. As of August 2005, ESPN continues to carry bogus RPI data for last season. I am confident what I post is correct, subject to game data errors which are rare but do occur. Now that home/road wins matter, small errors can occur if the NCAA considers certain games neutral that I do not.

How does the RPI factor in margin of victory?
It doesn’t.

How does the RPI handle home site advantage?
As of the 2005 season, the RPI gives more weight to road wins and home losses (see above description of Part I). Prior to 2005, site of game was not a consideration in the formula.

Any more complaints about the RPI?
I’ve written a little on the RPI in an effort to defend it, here, here, and here. It doesn’t mean I think it’s perfect, it’s far from it, but it’s better than most folks think. Especially considering the selection committee is there to clean up any messes, although they do fail occasionally.

Any more questions? Write to ratings@kenpom.com

Posted on 08/01 at 05:49 PM
RPI • (0) Trackbacks • Permalink • E-mail me

Powered by ExpressionEngine