The RPI Explained:Last saved: 2001/01/15 21:45:12 GMT by ac200
The RPI is a calculated ranking, an estimate of the relative power of all the teams in a league, like CIAU Men's Basketball.
It was originally devised by the NCAA to help in selecting wild cards to fill out their tournament field of 64 teams, and to help with seeding for Regionals. Imagine trying to remember all the games of 300+ teams each playing 30 games, and trying to keep track of records of games with common opponents. The RPI came out remarkably similar to the final judgements of a room full of experts spending 3 or 4 days in ranking the teams. NCAA still relies on committee selection, but have the RPI to help resolve difficult choices. It is now widely used by others, and distributed early in the season, where it is somewhat less reliable, but still reasonably accurate, and popular with fans.
Each week, Mark Cheers' software scans our master scores pages to read in all known games between CIAU teams. Only games against other CIAU teams are considered. Since all games are treated equally, the pre-season takes on a new importance.
The software is quite complex, but the math is simple.
Simply stated, the RPI Value is a total of:
25% of a teams's winning percentage
50% of their opponents' winning percentage against other teams
25% of their opponents' opponents' winning percentage
To show the calculation, here is a small early season example of Acadia Mens RPI (at that time).ACADIA DATE OPPONENT W- L SCORE MARGIN Fri Nov 14 Dalhousie 0- 3 W 67- 54 13 Sun Nov 09 Manitoba 3- 5 L 75- 76 -1 Sat Nov 08 York 3- 6 W 82- 70 12 Fri Nov 07 Bishop's 7- 2 L 64- 70 -6 Sat Nov 01 St. Francis Xavier 7- 1 L 67- 83 -16 Fri Oct 31 Saint Mary's 4- 3 W 100- 75 25 Sat Oct 25 Saint Mary's 4- 3 W 73- 71 2Acadia's record is 4-3. A next step is to delete all Acadia games from the Opponents W/L record. 50% of the RPI total technically comes from Acadia's opponents when they play other teams. Without deleting these Acadia games, their RPI value might be much lower. Let's say Acadia goes to 20-0. Without deleting, the Opponents record would include another 20 losses. Here is how The OPPONENTS W-L column is adjusted.DATE OPPONENT W- L Fri Nov 14 Dalhousie 0- 2 Acadia win removed Sun Nov 09 Manitoba 2- 5 Acadia loss removed Sat Nov 08 York 3- 5 Acadia win removed Fri Nov 07 Bishop's 6- 2 Acadia loss removed Sat Nov 01 St. Francis Xavier 6- 1 Acadia loss removed Fri Oct 31 Saint Mary's 4- 1 Two Acadia wins removed Sat Oct 25 Saint Mary's 4- 1 Two Acadia wins removed TOTAL Opponents Record = 25-17 in non-Acadia gamesSimilarly, each Opponent's Opponents record is established. For Acadia's Opponent's Opponents it came out as 159-124. The table below shows the calculation steps of RPI.ACADIA OPPONENTS OPP/Opp's RECORD 4-3 25-17 159-124 divide 4/7 25/42 159/283 W/L% .571 .595 .561 RPI % 25% 50% 25% Add .143 + .298 + .141 = .581 [RPI] 6. Acadia 4- 3 0.571 0.595 0.581A big component of the RPI is schedule strength. Fully half the weight goes to the Opponents Strength. A team may have a 6-8 record, but if it has only played quality opponents, it will be ranked much higher. A team at 8-6 might be ranked much lower because it has mainly played the weakest teams. Over the season, this effect diminishes somewhat.
The RPI is constantly self-correcting. An 8-8 team that goes 2-2 over a 2-week period may rise quite dramatically in the rankings if all its 16 former opponents won 4 straight over that time. The calculations gradually reveal the team had been playing a much tougher schedule than shown previously.
BoilerJP claims the RPI only becomes statistically significant for the NCAA by about December 15. That would be about November 15 for the CIAU where teams begin much earlier, then take off most of December for assignments and exams. With their semester or trimester systems, the NCAA begins about mid November and plays a heavy schedule right through December.
Once into conference play, team movement in the RPI depends on the strength of other teams in the conference. A team that went 6-6 in the pre-season against some of the better teams will only slide down in the RPI if the conference teams are generally weaker. However, in a strong conference, that team could climb quite high in the RPI rankings. What determines conference strength? It is not decided in advance. It only emerges each season as inter-conference pre-season games are played.
Teams who think they are in for a big year will select a tough pre-season schedule, so they will be noticed in the polls and rankings. This is particularly noticeable in the NCAA. At the same time, very good teams have nothing to gain by playing pre- season games against weak opponents. In the NCAA, with easier travel, they also pointedly avoid playing teams from their own conference in the pre-season. The top teams will not even enter a tournament that already has one of its own teams. They play the entire pre-season out-of-conference. CIAU teams may be moving in that direction. Some Womens tourneys are trying to improve that. In this year's Dal Subway, each Atlantic team played each of the 3 Ontario visitors once in a modified round-robin. In one 8-team tourney last season, organizers switched around first- round losers if they were to meet another from the same conference. Acadia and Saint Mary's will later see enough of each other, so they can flip-flop to meet Lethbridge and Calgary in the next round.
If you glance at our master scores lists and see there is a score missing, please e-mail. One missing win against a strong opponent may mean a difference of 3 or 4 in your team's rank.
CHEERS RPI and Expert PollsNow that we better understand the RPI, it is easy to see how it differs from a conventional expert poll, like a coaches poll or a sportswriters poll. Some would try to make them both do the same thing, but they won't, and shouldn't.
The poll has higher validity, especially face validity. The experts know a lot about the teams that they mentally include in their rankings. They know that team A has its two starting guards returning this week, and will almost certainly play better than in their last 3 games. They also know that team B's dominant centre may have a bad ankle and may not play for 3 weeks. They know that two of team C's former stars will become eligible this weekend. They aren't yet sure if young team D is for real, even after 6 straight wins and 12-3, so they move them up, but slowly. Polls repeated every week also are a bit like Delphi Techniqe, with the summary of last week's voting tending to influence the decisions of individuals polled this week. Results move toward more consensus as the season advances.
The computer calculating the RPI cannot know these things, or assimilate them in any meaningful way. It does not concern itself with whether games are Home or Away, whether team E played 5 games in 5 nights, whether team F is the defending champ, whether team G has won 10 straight.
On the other hand the RPI is much more reliable, or consistent, simply because it is calculated. Doing what it does, it can easily give equal consideration to the record and schedule strength of every team. The experts being polled will know more about the strength of teams in their own area, but will have to rely on rumor and tips from other experts for their knowledge of teams in other areas.
So which is best? Neither.
So which do I prefer? As a serious fan, I want to see both. I want to see what the experts think of the teams, how they would rank them, based on those intangibles that only the experts as insiders can know or hear about. But I also want to see how the teams stack up in terms of their season record, and the strength of their opponents.
When I follow my favorite NCAA team, I can check their position in BoilerJP's RPI by Jerry Palm, in the similar Associated Press RPI, a more complex one in The Sagarin by Jeff Sagarin, the new NCAA Power Index and in each of two major polls of sportswriters like USAToday Top 25 and the AP Top 25 . It is fan heaven... -:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-:-