Below are this week's Adjusted RPI ranks based on all games played through Sunday, September 30. These are different than the NCAA's ARPI ranks published today for three reasons:
1. The NCAA's data, as of the time they published their ranks, were missing the 9/27 Wake Forest 2 v Boston College 1 game. This most likely is due to the teams not having reported the game into the NCAA's data system before the NCAA published the rankings.
2. The NCAA's data have the 9/30 Alabama State 10 v Alcorn State 0 game score backwards. This most likely is due to the score having been reported to the NCAA incorrectly.
3. With 335 schools fielding Division I teams this year, the bottom two tiers of RPI adjustment penalties are for losses and ties to teams ranked 256 through 295 and 296 through 335 respectively. Last year, there were 333 teams, so the tiers were 254 through 293 and 294 through 333. The NCAA, however, forgot to change its computer programming for the penalty tiers and still is using last year's tiers.
The NCAA now is aware of these problems, so I expect they will make corrections some time this week. The missing Wake Forest v Boston College game, however, is particularly unfortunate as it has resulted in this week's NCAA published rankings having some pretty serious discrepancies from what they would be with the game included in the data base.
In the mean time, the ranks published below as well as those at the All White Kit 2018 Division I Women's Soccer Schedules website are exactly correct. You might wonder how I know they're exactly correct, so here's an explanation of my weekly process that allows me to say that:
Each Monday morning, I produce current ARPI rankings and other information based on the results of all games I believe have been played through the preceding Sunday. I get my game results from the All White Kit website. I then compare my ratings (and thus rankings) to those of All White Kit, to be sure they match. If they don't, I locate the reason for the difference, either an error I've made or one they've made. If I've made an error, I correct it; and if they've made an error, they correct it. This results in our two systems matching exactly on a weekly basis. Then, as soon as the NCAA has published its RPI rankings, I compare their rankings to mine to see if we have an exact match. I do this using documents they publish at their RPI Archive. If our rankings do not have an exact match, I go through a series of steps to see where we have data inconsistencies, either in game results or in game locations. I check game locations because they can affect adjustment bonus and penalty amounts. When I've found all the apparent data inconsistencies, I check to see whose data is correct: the NCAA's or mine/All White Kit's. If I have a data error I correct it and also notify All White Kit, where it also gets corrected. If the NCAA has a data error, I notify the NCAA and they make corrections. In the meantime, however, I temporarily modify the data in my system to insert the "errors" that the NCAA has in its system and produce my own modified ratings/ranks based on what appears to be a data base exactly identical to theirs. I then again compare my ranks to theirs. That's what I did today, and with those "errors" in the data base, my ratings/rankings exactly matched theirs except for the rankings and ratings of a few odd teams. When that happens, I consider possible reasons for the differences other than data discrepancies. Having one this for a number of years, it was relatively easy for me to figure out that the rankings that didn't match this week had to do with the NCAA's not having changed their penalty tier programming this year to match the increase in the number of Division I teams from 333 to 335. (This is not the first time the NCAA has forgotten to adjust the tiers.)
So, after going through the above process to assure that my numbers, with the NCAA "errors" introduced, exactly match the NCAA's, I then reinstall the correct data and am able to say that my ratings/ranks are exactly correct. Once the NCAA makes its corrections, their ratings/ranks will be exactly the same as mine.
So, with that little essay, here are this week's exactly correct RPI rankings:
No comments:
Post a Comment