Wednesday, September 21, 2016

Weekly RPI Report, Games Through September 18

For those interested, I have published a weekly RPI report, in Excel format, at the RPI for Division I Women's Soccer website's This Year's Reports page.  At the bottom of that page, you'll find a link to the report.  It's a downloadable Excel file.  It covers games through September 18.

Although I can't be absolutely sure yet due to the data issues mentioned below, it appears that there have not been any changes in the RPI formula since last year.

There are some slight variations between the numbers in the weekly report and those in the NCAA's rankings report and in the NCAA's more detailed reports published at the NCAA's RPI Archive.  These variations are due to a couple of errors in the NCAA's data base:

*The NCAA, as of yesterday, did not have the September 18 game of UNC Asheville (H) 4 v Alabama A&M 1 in its data base.  I suspect this is a game that the teams scheduled after they initially entered their schedules into the NCAA's system and that they forgot to enter it into that system once they scheduled it.  I've let the UNC Asheville coach know the game isn't in the NCAA's system, so we'll see if there's a correction when the NCAA issues next week's reports.

*The NCAA's data base, at the time it did its RPI calculations, appears to have had the August 23 game of USC Upstate (H) 0 v Winthrop 1 with the score backwards -- USC Upstate 1 v Winthrop 0.  This now appears correctly in the NCAA's statistics data base, so I'm guessing there was a correction after the NCAA published its RPI reports.  Again, we'll see next week if the NCAA makes a correction.

Errors like these have ripple effects, particularly on these teams' opponents' strength of schedule components of the RPI.  If there are not corrections by next week, then my next step is to notify the NCAA staff responsible for Division I Women's Soccer of the errors.  Typically, the staff then checks out the games and makes the appropriate corrections.

No comments:

Post a Comment