(promoted to front by PJS. A well done diary/fanpost).

I am writing this about college basketball, but let me first elaborate on my opinion of such tools by talking about college football and how it relates.

Humans vs Computers

If you will recall, there are two primary components to the BCS formula in college football--the polls and the computer rankings.  A few years ago, the BCS created an unfavorable match up in the national championship game that caused an uproar and therefore a tweak to the BCS formula regarding the weighting of the human and computer components.  This was probably going to happen in any event--computers or humans--in the absence of a playoff.  Critics argued that computers shouldn't hold as much weight as the human polls, that somehow humans were smarter.  I was outraged at that argument back then (2003, I believe) and still contend that computers are much better at ranking teams than humans.

First, humans are biased.  They go off of name or brand equity and exposure.  This is why Utah didn't get a chance to play for the title despite being undefeated in football.  I will take this time to toot my own horn where I predicted a double digit win for Utah over Alabama.  My point is that computer rankings (such as RPI) take an entire season into consideration and not just a snapshot of who is better this week.  For the record, Utah was ranked ahead of Alabama in the computer rankings.  However, this is not a predictive tool; rather it is a ranking tool for games played to date.  It really is the best way to rank teams in college football.

Does One Game Determine Who is Better?

All that human polls do is rank teams according to an iterative logic (who won this past week, who lost and shuffle accordingly--despite who that win or loss was against).  Also, a win early in the season against a supposed favorite can either be supported or refuted by a computer.  For example, the Gophers hoops team had a big win over Louisville earlier this year.  How big of a win is that really going to turn out to be?  Well, if Louisville wins the Big East, that is going to be a huge win.  But if Louisville keeps falling to teams like UNLV, as they did a few days ago (at home, no less) then it isn't as great as it once seemed.  And it shouldn't be--Louisville was ranked #9 based on perceptions and not reality at the time; and what their RPI ranking was at the time is completely irrelevant--it is only relevant at the end of the season and before the tournament.  That ranking at the end of the year tells us how good our opponents really were.

Furthermore, just because team X beats team Y does not necessarily mean that the winning team is "better".  They could have been better that day, but if they played 10 times, the outcome could be different.  A perfect example of this is last year's Super Bowl.  The Giants were definitely better that day, but do you think the Giants would have won more games over the course of a series?  I don't think one game makes an entire season.  But it does in sports with a one and done sort of playoff (NFL, college football's BCS championship game, and college basketball where one loss sends a team home) where the "best" team does not usually win the championship.  One can certainly make the case that a 7 game series (MLB, NHL, NBA) does not guarantee such a result either, but I would argue it does so more often.  Likewise, it is possible that team X can beat team Y, yet be ranked below team Y in RPI because of the two schools' "body of work"; that is the context of the entire season not just one head to head game.

Nuts and Bolts of RPI

So how is the RPI calculated?  The NCAA does not actually release an official RPI during the season.  A close approximation is 25% of your team's winning percentage, 50% of your opponents' winning percentage (strength of schedule), and 25% of your opponents' opponents' winning percentage (opponents strength of schedule).  The precise calculation involves adjustments for home/road/neutral results.

So while it took me a while to get here, my point is that the Gophers are currently in a great position RPI wise.  Here is a great site that updates RPI standings in a timely manner and is organized in an easy to use fashion.  As of this writing, the Gophers are sitting at #37.  This includes the win over Ohio State.  We were #53 or #54 before the game.  Ohio State was a good win for us since the Buckeyes have an RPI ranking of #19 even after the loss.  Louisville has a #47 ranking--not quite what it was a few weeks ago.  Our loss to Michigan State was to the #11 team--so if you are going to lose, that is the sort of team to lose to--one that won't hurt too badly.

Importance of Conference Play

Now that we are in conference play, you will see a convergence of Big 10 teams' RPI (as you will with other conferences).  This is because before conference play teams were playing unrelated teams, often teams that won't play anyone else in the conference.  Now, however, all eleven teams are playing against each other.  Think for a moment if the Big 10 was the only conference.  There would not be a need for an RPI or other computer rankings.  The records would speak for themselves since everyone plays one another.  So conference play will tighten the standard deviation of the rankings--think of the distribution of RPI in terms of a graph.  After conference play, the shape of the graph will be closer to a bell curve than it is now.  The current range is #11 (Michigan State) to #192 (Indiana).  The Gophers are currently the median at #37; that is they are the middle team in the conference.

You'll notice that the average rank is just over 54.  That means 8 of the 11 teams are better than the conference average.  Expect the average RPI to fall since any win for the conference is also a loss for another conference team from here on out (with exceptions for any teams playing out of conference games in the middle of conference play, but the effect will be negligible).  But this also means that the average will normalize and there will only be five or six teams above the conference average (closer to one half).  The range will probably increase because some teams' winning percentage will be a mirage since it was compiled against weak competition.  Indiana, for example, could fall into the 250 range as wins will get even tougher for them.

Conference vs Conference RPI

The Big 10 is in a good position this year relative to other conferences thanks to a strong performance against other conferences.  The Big 10 has the second best conference RPI right now (ACC).  This means that each win against a Big 10 team is going to be worth that much more (although wins against Indiana may prove not helpful).

Final Thoughts

Remember that RPI is not a predictor.  It is a way to measure a team's performance to date based on varying schools' schedules.  Also, try not to think of RPI in an iterative manner.  That is, don't look at an individual game's impact on the RPI.  You will often hear pundits talk about a supposed absurdity how a team's win could lower it's RPI.  While in isolation that may be true, it isn't just that one game.  It is the team's overall performance and the context of that performance--schedule and opponents' schedule.  If a team has a gripe about its RPI, it should have won more games or scheduled tougher opponents.