2017 RPI Ranks 10.23.2017

Here are my Adjusted RPI ranks based on games played through Sunday, October 22.

If you compare my ranks to the NCAA RPI ranks published the morning of October 23, you may notice some differences.  These are due to three NCAA data errors.  One of the errors was the NCAA's data not including a game that had been cancelled due to weather and later re-scheduled.  This was the teams' fault, as they deleted the cancelled game from the NCAA's schedule but did not add it back in when they re-scheduled it.  Two of the errors have to do with what the NCAA's RPI software does when two opponents, when entering their game into the NCAA's schedule, have entered it differently.  The RPI software has a problem with dual game entries, and this week one game got entered twice into the RPI calculations and the other game, which was a tie, got entered as a win/loss.  I've notified the NCAA of these problems and, if they follow their usual course, they'll check to see whether I'm right and then will make corrections.

If you use All White Kit's "real time" website, you also may notice that sometimes they and I will have two teams inverted.  This is not due to data inconsistencies or programming errors, rather it's due to statistical rounding conventions and how we rank teams when their ratings are equal to a certain number of decimal places.  Our rankings, if the ratings are carried to enough decimal places (mine are to 12 decimal places, since that's Excel's default), are identical.  When the NCAA's data is error free, my rankings have matched theirs exactly this year, so I have a pretty high degree of confidence that my rankings will match theirs once they correct their current data errors.

As in preceding weeks, the rankings table shows which teams are potential #1, 2, 3, and 4 seed candidates as of this stage of the season and also which teams are in the at large "bubble" as of this stage.  As of this week, the "poorly ranked" edge of the bubble has narrowed considerably, to #62, and is getting close to the ultimate edge of the bubble, which is #57.  These seeding and at large ranges are based on the last 10 years' experience, so they're not absolute ranges, simply what's happened over the last 10 years.