Differences from NCAA Rankings. If you compare these ranks to those the NCAA has published, you will see that there are some differences. Nevertheless, these ratings and ranks are correct. There are a couple of data-related reasons for the differences:
First, the NCAA's data at the time the NCAA produced its ratings and ranks are missing two games:
9/23 Duquesne v Dayton @ Duquesne
9/23 Towson v William & Mary @ Towson
Since all of the teams' ratings are interrelated, not having the data from these games affects virtually all teams' ratings and ranks. My data, and the data at All White Kit's 2018 College Women's Soccer Schedule website, include these games. This accounts for most of the differences.
The NCAA's not having these games in its data as of the time it published this week's ratings is due to the schools not having entered the results into the NCAA's data base by the deadline for having them included. Hopefully the schools are getting the message that they need to get their game results into the system in a timely fashion.
Second, the NCAA's game location data for these three games are incorrect:
8/17 Loyola Marymount v Bowling Green @ Hawaii -- the NCAA incorrectly has this as a home game for Loyola Marymount
8/19 Loyola Marymount v Texas A&M @ Hawaii -- the NCAA again incorrectly has this as a home game for Loyola Marymount
8.24 Lipscomn v Ohio @ Vanderbilt -- the NCAA incorrectly has this as a home game for Lipscomb
The game location errors are due to the schools having incorrectly reported the game locations to the NCAA. Typically this happens when there is a tournament. For game where both teams are at a neutral site, for box score purposes one team is designated as the home team and the other the away team. When the school's responsible person enters the game data into the NCAA's system, the person includes these home/away designations and the game shows up in the NCAA's data base as a home/away game when it really was a neutral site game. This is a problem because it can affect bonus and penalty adjustments. I've advised the NCAA staff of these errors and expect to see correct game locations next week.
Third, the NCAA has what I consider to be a slight problem in its programming related to the award of bonus adjustments. Among the adjustments, there are bonuses for wins and ties against teams ranked 80 or better by the Unadjusted RPI. The NCAA's system currently shows three teams with rankings exactly at #80 -- Dayton, Villanova, and DePaul -- and gives bonuses for wins and ties against all three of them. When looking at the Unadjusted RPI ratings for the three teams, they are identical when their ratings are rounded off to four decimal places. Computations of the Unadjusted RPI to more decimal places, however, show that these three teams actually have different ratings and the true #80, for the games in the NCAA's data base, is Dayton. The NCAA's system thus is awarding bonuses for wins and ties against Villanova and DePaul when it shouldn't be. Ultimately, I anticipate these problems will disappear, as by the end of the season it is very unlikely that this problem will occur right at a bonus (or penalty) adjustment limit.
The table below does not have these problems and thus should be exactly correct for all games played through Sunday, September 23.
Historic Ranges for Seeds and At Large Selections. This is the first week in which I'm including color coded columns to show teams that are in the historic ranges of possible seeds and at large selections for the NCAA Tournament. These are based on data from the last eleven years, 2007 through 2017. For at large selections, those data indicate that all teams currently ranked #10 or better will get at large selections. These are, however, simply historic ranges so there is no guarantee they will hold true this year.
No comments:
Post a Comment