The RPI rating is one of those computer ratings that gets a lot of press leading up to "March Madness".  Here's how you decipher what's really going on.

RPI is a method of ranking teams based on the opponents that a team plays. It is one of many types of computer ranking systems that are used in the sports world. Unlike most computer ranking systems, however, RPI has a standard definition. This definition is the one that is used by the NCAA basketball selection committee as an aid to determine who will qualify for the tournament and what their seed should be. I will note that the NCAA selection committee "adjusts" this ranking, however, the exact formula is not known.

One basic assumption that must be made, is that this can only be applied to games against opponents within their own class. This is how the NCAA does it (only games against Division I opponents count). I do not want to try and make adjustments to a standard formula to account for the rating distortion that would occur from a 3A school that plays a 4A school and loses. A weaker 4A school may be at the same level as a stronger 3A school, yet the 3A school would be punished more by losing to the 4A school since a weaker 4A school will have a poorer overall record than a stronger 3A school. Try the Power Index for a ranking that tries to account for these games.

There was a lot of confusion in 2005 with regard to the NCAA's definition of RPI. Apparently, the NCAA decided to add a weighting for home versus away games. A lot of the college basketball RPIs varied for the same team, based on whether the source was using the old or new definition. I have updated to the new formula. Background information on this subject can be found here.

The RPI calculates 3 percentages and assigns each of them a weight, which is published by the NCAA. Let's say we're calculating the RPI for Portland High School (I know it doesn't exist know--I don't want to sound prejudiced towards any school here).

• Team's winning percentage. It used to be pretty simple to calculate. The new formula is a little harder. For wins, you take home wins * 0.6 and add road wins * 1.4 and add neutral wins (no adjustment). For losses, you take road losses * 0.6 and add home losses * 1.4 and add neutral losses (no adjustment). Then calculate your winning percentage using wins / (wins + losses). This % is weighted 25%.
  If Portland is 2-2 with a 2-1 home record and a 0-1 road record, their adjusted wins = 2(0.6) + 0(1.4) + 0 = 1.2 and their adjusted losses = 1(0.6) + 1(1.4) + 0 = 2. Their adjusted winning percentage is 1.2/(1.2+2) = 1.2/3.2 = 0.375. They get .375 * 25% = .09375 RPI points.

• Opponents' winning percentage. This looks at the the teams Portland has played and calculates their adjusted winning percentage using the method above, excluding the games played against Portland. Let's say Portland beat Seattle in Portland. Seattle is 1-4 right now (1-1 at home, 0-3 on the road). Seattle's record minus the Portland game is 1-3 (1-1 at home, 0-2 on the road). Calculate the adjusted wins and losses by weighting home/road games as above for all of Portland's opponents. This % is then weighted 50%. Let's say the adjusted records add up to 6-4, or .600. Multiply by 50% and you get .300 RPI points.

• Opponents' opponents' winning percentage. This is a little trickier. Here you look at all of the teams Portland has played (Seattle, Boise, Sacramento, and Las Vegas) and see who each of these teams have played. For example, Seattle is 1-4. You look at each of the 4 teams (besides Portland) they have played and calculate those teams' adjusted winning %. Repeat for Boise, Sacramento, and Las Vegas. Then average the adjusted winning % for these 4 teams and multiply by the 25% weighting factor. Let's say Seattle's opponents are .333, Boise's are .500, Sacramento's are .500 and Las Vegas' are .667. The average of those is .500. Multiply by 25% and you get .125 RPI points.

• By adding up the RPI points, you get the total RPI ranking. In this case, Portland's RPI is .09375 + .300 + .125 = .51875.

The RPI will change after each game played--it does not look at the records when the games were played, but rather what the current records are of the teams they played.