Even the most die-hard Eagles fan would probably admit that Eastern Michigan isn't as good as UConn. Though not as dominant as in past years, UConn is still a legitimate top-20 team and a contender to reach Indianapolis. Eastern Michigan doesn't even have a shot at an at-large bid. UConn has an RPI around 12 or 13. Eastern Michigan has an RPI somewhere in the 80s.

Suppose you're coaching another team, and you're looking to schedule some tough games to build your resume and and impress the Selection Committee by improving your RPI and Strength of Schedule (SOS) numbers. You know that UConn is a better team by any measure, including the RPI.

You would think that playing UConn would help your RPI and SOS much more than playing Eastern Michigan.

You'd be wrong.

As I described last month, the RPI uses a 25/50/25 weighting for winning percentage (WP), opponents' winning percentage (OWP), and opponents' opponents' winning percentage (OOWP).

The unequal weighting causes a quirky reversal. For UConn's own RPI, its WP counts for half as much as its OWP. But for UConn's opponents' RPI, UConn's WP counts for twice as much as UConn's OWP.

As a result of this quirk, it helps your SOS to play teams with a good record against a weak schedule (even though their own RPI is low). Conversely, it can hurt your SOS to play teams with a bad record against a tough schedule (even though their own RPI is high).

Some examples:

(1) If you play UConn this year, it adds a game worth about .689 to your SOS. If you play Eastern Michigan, it adds a game worth about .681 to your SOS. Thus, playing and beating UConn helps your RPI about as much as playing and beating Eastern Michigan.

(2) If Duke plays Virginia Tech (RPI 19) on Saturday, its SOS will go down. If Duke were to play Chattanooga (RPI 52) instead, its SOS would go up. In fact, beating the Mocs would help anyone's RPI more than beating UConn, Penn State, or Texas.

(3) Auburn, with an RPI of 37, is a bubble team. On January 9, Auburn played and lost to Florida (RPI 39), a team with a low WP against a very tough schedule. If instead Auburn had played and lost to Chattanooga, a team with a high winning percentage against an easy schedule, the Tigers' RPI would be about 32 instead of 37.

The Tigers could even have improved their SOS by playing Tennessee Tech (RPI 110) instead of Florida. Even if it had lost to TT, Auburn's RPI would be better than it is now. And if it had beaten TT, Auburn's RPI would be around 30. In bubble territory, the difference between 30 and 37 is significant.

What does this all mean?

It should just reinforce the truth that the RPI is only an approximation and a starting point for other analysis. No sensible system gives you equal credit for beating UConn and Eastern Michigan. And when the Committee really gets down to cases and starts comparing quality wins, it of course counts a win over UConn more than a win over E-Mich.

Also, if you're a coach who wants to game the system and manipulate your SOS and RPI, don't schedule games against mediocre teams in tough conferences. Instead, schedule games against mediocre teams in very easy conferences -- look for teams Alcorn State, Coppin State, St. Francis, and Liberty. That way, you can get easy wins and still move your SOS and RPI up.