Every conversation about college basketball involves at least one reference to a computer ranking. Usually it’s the RPI (Ratings Percentage Index), but occasionally Jeff Sagarin’s ratings are mentioned. The RPI however takes center stage because the algorithm is a known commodity. Or is it? In truth, most of what we already know about the RPI has been made public, but insiders keep telling us the NCAA Selection Committee uses an adjusted RPI at the end of the season that applies bonus points and penalty points based upon a predetermined set of criteria. Commonly referred to as the “nitty gritty” RPI, it’s held in secrecy by all but the committee members. That said, we know enough about the guts of the RPI to run our own computations throughout the season to follow along.

In an effort to counteract some of the inherent weak points of the RPI, we’ve developed our own in-house computation, the UDPride Power Rankings. To understand the Power Rankings however, one must fully understand the RPI and how it’s calculated. It’s not as straightforward as you think and requires a good deal of arithmetic to make it happen. For starters, 25% is based on winning percentage, 50% on opponent’s winning percentage, and 25% on opponents’ opponents winning percentage. But there are a few details along the way that few people realize – and if you don’t perform the calculations properly and in the correct order, you’ll miss the mark. Let’s take the RPI and UDPride Power Rankings one at a time.

THE RPI

Here’s the proper way to calculate a team’s RPI. In this example, we’ll use the Austin Peay Governors. To keep the numbers simple, we’ll use their 2-1 record through games up to and including 12/3/02:

Points are awarded on three levels:
First level points = Winning % (25%)
Second level points = Opponents winning % (50%)
Third level points = Opponents’ opponents winning % (25%)

THE FIRST LEVEL (Winning Percentage worth 25% against D-I teams)

Austin Peay is 2-1

• Throw out the D-2 win over Bluefield College
• Win over Memphis
• Loss to Missouri

• D-1 record is 1-1 for a .5000 winning percentage
• Multiply .5000 x .25(first level percentage) for .1250

— First level points = .1250

THE SECOND LEVEL (Opponents Winning Percentage 50% against DI teams)

Austin Peay has played Bluefield College, Memphis, and Missouri. Since only DI games count, Bluefield College is thrown out.

• (A) Memphis is 3-1 against DI teams
Throw out the game against Austin Peay and they are 3-0 or 1.0000
• (B) Missouri is 3-0 against DI teams
Throw out the game against Austin Peay and they are 2-0 or 1.0000

• Add (A)1.0000 + (B)1.0000 – divide by 2 for 1.0000 and x by .50(second level percentage) for .5000

— Second level points = .5000

THE THIRD LEVEL (Opponents’ opponents Winning Percentage against DI teams)

Memphis played –
• (A1)won over Syracuse who is 2-1 – throw out the Memphis
game and they are 2-0 or 1.0000
• (B1)lost to Austin Peay who is 1-1 – throw out the Memphis
game and they are 0-1 or .0000
• (C1) won over Arkansas Pine Bluff who is 0-4 – throw out the Memphis game and they are 0-3 or .0000
• (D1) won over Arkansas Little Rock who is 3-2 but only 2-2 against D-1 teams (game against Central Arkansas doesn’t count) – throw out the Memphis game and they are 2-1 or .6666

Missouri played
• (A2) won over American who is 1-3 – throw out the Missouri game and they are 1-2 or .3333
• (B2) won over Austin Peay who is 1-1 – throw out the Missouri game and they are 1-0 or 1.0000
• (C2) won over Cal St.-Sacramento who is 0-3 against D-1 teams (Dominican doesn’t count) throw out the Missouri game and they are 0-2 or .0000

1.6666(A1+B1+C1+D1)/4 = X = .4165
1.3333(A2+B2+C2)/3 = Y = .4444
Third level = .25(third level percentage) x (X+Y)/2 =
Third level = .25(third level percentage) x (.4165+.4444)/2 = .1076

Third level points = .1076

Level 1 + Level 2 + Level 3 = RPI
.1250 + .5000 + .1076 = .7326

Austin Peay = .7326 RPI

Notice at the third level, the division must be made at each individual opponents’ opponent, rather than adding up the entire opponents’ opponents wins and losses and dividing by the number of games played. This is purposely done so that no opponents’ opponent weighs more or less heavily on your third level point total. Were you not to do this, an opponents’ opponent with a 27-0 record would count far more than an opponents’ opponent with a record of 3-0 because of the totality of the games involved. That’s why the RPI is based on average winning percentage and not the average of wins and losses. You’ll also notice your matches against your opponents as well as matches your opponents play against their opponents are thrown out of the computation to determine winning percentage. This is frequently overlooked by the general public. These games are omitted from the RPI to ensure that winning against an opponent (or an opponent against their opponent) does not adversely affect your second and third level point totals.

So what does the RPI tell us? It tells us exactly what it’s designed to tell us. In other words, teams who beat other teams with good records that have beaten teams of their own with good records do well in the RPI. The RPI is literally nothing more than a two-tiered daisy chain of winning and losing. What the RPI doesn’t tell us however is whether you are beating good teams with good records (a 19-7 Missouri club) or not-so-good teams with good records (a 19-7 Radford club). Assuming Radford and Missouri have the same opponent’s record, there’s no advantage whatsoever in beating Missouri. And therein lies the fatal flaw, or at least one of them. To make matters worse, the RPI makes no distinction between winning at home and winning on the road. All things being equal, a team that goes 24-8 without playing a road game will generate the same RPI as a team that finishes 24-8 (assuming identical second and third level points) that doesn’t have the luxury of playing in front of the home crowd. In other words, there’s no incentive to win on the road.

THE UDPRIDE POWER RANKINGS

First and foremost, we call them power rankings because they measure a team’s power potential to play competitive basketball, with less emphasis on merely winning and losing. Power can be defined as the team’s inherent program strength and ability to be competitive against teams from around the country. Power also results from an institution’s ability to schedule teams from conferences that, on average, field high caliber players and programs. Beating teams helps your power ranking, but sometimes it’s less important to beat a 9-1 Cornell team than it is to lose to a 3-5 Oklahoma State program. On average, Oklahoma State remains the stronger foe, and we cater the power rankings accordingly to take into account Oklahoma State’s conference and name recognition that gives it a competitive advantage in college basketball. By the end of the season, the chances are good that Oklahoma State is playing better basketball than Cornell — at least they should be. Conference are weighted in four categories and given a multiplier:

• x1.1 Top Tier: BCS Conferences (Big East, ACC, Big10, Big12, SEC, PAC10)
• x1.0 High-Mids: Historical (A10, C-USA, Mountain West, WAC)
• x0.9 Mid-Majors: Next six conferences based on conference ranking (Varies throughout the year)
• x0.8 Low-Majors: Remaining conferences

Multipliers are also used for home, neutral, and away games:

• x1.1 Road game
• x1.0 Neutral game
• x0.9 Home game

Winning is worth 40% (.400) of a team’s total score.
Opponents winning percentage is worth 60% (.600) of a team’s score, with bonus/penalty multipliers factored in.

Let’s use Austin Peay again and calculate their UDPride Power Ranking under similar circumstances.

FIRST LEVEL POINTS (Winning Percentage 40%)

Austin Peay is 2-1 on the season with wins over Bluefield College, Memphis and a loss to Missouri.

• Throw out the game against Bluefield College (non DI)
• Record is 1-1 or Winning percentage = .5000

.4000(Winning percentage weight)x.5000(Winning percentage) = .2000

SECOND LEVEL POINTS (Opponents Winning Percentage 60%)

• (A) Memphis is 3-1 against DI teams
Throw out the game against Austin Peay and they are 3-0 or 1.0000
• (B) Missouri is 3-0 against DI teams
Throw out the game against Austin Peay and they are 2-0 or 1.0000

Memphis was played on the road from a High-Mid Conference:

(A) 1.000×1.1(road match)x1.0(High-Mid) = 1.1000

Missouri was played on the road from a Top-Tier Conference:

(B) 1.000×1.1(road match)x1.1(Top-Tier) = 1.2100

(A+B/2) x 0.6000(Opponents Winning Percentage) =
[(1.1000+1.2100)/2]x.6000 = .6930

UDPRIDE POWER RANKING = FIRST LEVEL POINTS + SECOND LEVEL POINTS
.2000 + .6930 = .8930

Austin Peay = .8930 UDPRIDE POWER RANKING

So what do we have? The RPI yields a score of .7326 while the UDPride Power Rankings report a score of .8930. While that’s a large discrepancy, it means next to nothing. First, we sampled a team with just two DI games under their belt. Over the course of the season, each individual game will count less. After two games, each game is ultimately responsible for 50% of a team’s score. By year end, it’s just 1/27th, making large discrepancies less prevalent. And most importantly, pure point totals are irrelevant when comparing the two systems because they operate on entirely different formulas of scale. It’s the ranking discrepancy that we’re after. Small discrepancies are what prompted us to create the UDPride Power Rankings to begin with and that’s what we hope you take note of throughout the college basketball season. For Pride Plus fans, our RPI and UDPride Power Ranking computations are much more in-depth and populate the ranking discrepancy for every DI team in the country so you can go down the list and see exactly who’s cheating the system (RPI) with inflated ratings against mediocre teams with good records from lower-tier conferences.

For those who’d like to know, it takes us approximately 10 minutes a night to enter scores. The software program calculates both the RPI and UDPride Power Ranking in approximately three minutes for all 327 teams from the beginning of the season up through the latest final scores entered into the database. At this point in the season, second and third level points number in the thousands and tens of thousands of mathematical equations.