OT: Can somebody explain how RPI rankings are done?

Submitted by mGrowOld on

http://realtimerpi.com/rpi_Men.html

I know its early in the season but I also know that the RPI rankings play a big role in selection and seeding once the tournament arrives and I don't get it at all.  Take a look at this week's RPI and see if you can explain how some teams are ranked where they are.  And i looked at each of their respective wins and losses for "quality wins" and "bad losses" and still makes zero sense to me.

Somehow Long Beach State, with their glittery 5-6 record is considered the 12th best team in the county while Ohio State is 21st and Michigan is 54th?  How can any ranking system so obviously screwed up be utilized in determining something as important as the teams selected to the tournament?

AAB

December 26th, 2011 at 5:46 PM ^

1/4 winning percentage, 1/2 opponents' winning percentage, 1/4 opponents' opponents' winning percentage, with a multiplier for home versus road games. 

Edit: to answer your question, Michigan has played a lot of terrible teams, and has played all of them at home.  That kills their RPI at the moment.  

mGrowOld

December 26th, 2011 at 6:08 PM ^

So 75% of the ranking is determined not by your record but rather the record of  the team you played and the record of the teams they play?  I appreciate your prompt response but does that make sense to you?  Not that your inaccurate but rather that the methodology they use is flawed.

MaizeAndBlueWahoo

December 26th, 2011 at 10:50 PM ^

Look at it this way.  Suppose you're ranking teams that all played a 10-game schedule:

-- One individual game you play counts for 2.5%.  (One-tenth of 25%.)

-- One individual game played by one of your 10 opponents counts for 0.5%.  (One-tenth of one-tenth - or 1/100th - of 50%.)

-- One individual game played by your opponents' opponents counts for 0.025%.  (One-thousandth of 25%.)

So the truth is that each game you play is five times more important than each game played by your opponents, and a hundred times more important than each game played by your opponents' opponents.  So it's a misnomer to say that "win-loss record is half as important as strength of schedule."

It makes good sense to me.  If you weight the ranking too heavily towards win-loss record alone, 1) you might as well just rank teams by # of wins, and 2) we all know, to use a football example, that Michigan's 10-2 record is better than Arkansas State's 10-2 record, and why do we know that?  Because we care about strength of schedule.  The RPI is a way to distinguish teams with similar records, and if you look at it in that light, it's sensible.

His Dudeness

December 26th, 2011 at 6:35 PM ^

LBS has the toughest schedule in the country so far by a large margin. It will come back down to earth when conference schedules start up and they play the nobody teams in their conference.

mGrowOld

December 26th, 2011 at 6:48 PM ^

True but that also reinforces my point that the logic used in determining the RPI rating is terribly flawed because while they've played the toughest schedule so far....they've LOST every game they played against good teams.  The best RPI ranking of a team they've beaten is 64th.

 

Date Opponent Opp. RPI Rk Results
  11-12     Idaho  5-6 (0-0) 230  69-61  W
$  11-16   at  Pittsburgh  11-2 (0-0) 70  86-76  W
  11-19   at  San Diego St.  11-2 (0-0) 57  73-77  L
  11-22     Boise St.  8-4 (0-0) 133  72-62  W
?  11-26   at  Montana  5-5 (0-0) 168  71-73  L
  11-28   at  Louisville  12-0 (0-0) 6  66-79  L
  12-06   at  Kansas  8-3 (0-0) 30  80-88  L
  12-10   at  North Carolina  11-2 (0-0) 11  78-84  L
$  12-22   vs  Xavier  9-3 (0-0) 64  68-58  W
  12-23   vs  Auburn  8-3 (0-0) 185  64-43  W
  12-25   vs  Kansas St.  10-1 (0-0) 24  60-77  L

 

mfan_in_ohio

December 26th, 2011 at 6:49 PM ^

After 18 games against the now-gaudy records of B1G teams, Michigan and Ohio, along with most of the rest of the conference, will go up quite a bit, while Long Beach has to deal with about 20 games against mostly bad Big West teams (the rest of the conference has an overall record of 23-51), so they'll go down even if they win all those games.  To be fair to LBSU, though, they're pretty good, and I think they'll be in a position at the end of the year to make the dance even if they lose in their conference tourney.

UMfan21

December 26th, 2011 at 7:27 PM ^

Keep in mind RPI is one of many things the committee might consider. Overall record, marquee wins, last few games, strength of schedule, etc all play a factor. RPI attempts to quantify these things, but we all know it's not perfect (like pair wise rankings or the BCS formula)

aiglick

December 27th, 2011 at 12:14 PM ^

The committee has said that it doesn't officially look at last 10 games anymore though people are still free to take that into consideration. Not sure this link talks about that change but it is an interesting insight into how the Selection Committee goes about their process.

http://sportsillustrated.cnn.com/2011/writers/andy_staples/02/21/mock.t…

85Lee

December 26th, 2011 at 9:50 PM ^

confusion but you can have a high rpi and not be a good team for the exact reason you pointed out above.  75% is not based on your record that being said it will all even out when michigan gets into the meat of the schedule.  Do what I do and pretend it does not exist until mid to late february.  

AMazinBlue

December 26th, 2011 at 10:23 PM ^

college football is looking at to "improve" the way they determine a national champion.

And you guys are forgetting the dumbest part of the whole thing.  The chairman of the committee is none other than moron from ohio, Gene Smith!  Enough said.