What a difference a day makes!
I realize I should have pulled Sagarin both before and after the ACC-B1G challenge, and failed to do so. I did tabulate the numbers yesterday before the games, and then again today, and we had quite a few pretty significant finishes. Huge wins from Purdue and OSU, and solid performances from PSU and Maryland, balanced against lackluster finishes from Nebraska and Wisconsin.
So, this early in the season, let's look at the impact of these results...
As of pregame, Dec 4 2019:
Then compare that to this morning, Dec 5 2019:
- Ohio State's bump was minimal, at 1 game added in both scheduled and RR wins. Their rating jumped by over 2 points, though. So there just wasn't much room to improve the projected record.
- Maryland remained with the same projected results, and only jumped rating by 0.05 points with a 21-point drubbing of Notre Dame.
- Purdue was the big gainer, with 5 additional projected schedule wins and 6 additional projected RR wins. The rating jumped around 3 1/2 points, which is pretty sizable for a single game. I also note that yesterday, Virginia's rating was 89.58, and today it was 85.81. They dropped from somewhere in the top 10 per Sagarin's ratings to 26th, while Purdue went from 40th in Sagarin yesterday to 17th. Huge swing for both teams.
- Wisconsin and Penn State both showed minimal movement based on their results yesterday.
- The big loser was Nebraska, with a massive loss to a mediocre P5 program in Georgia Tech.
What this means:Over on the Hammer and Rails site, there is a lot of question about the value of KenPom, who somewhat aggressively "smooths" out results, particularly early in the year. Part of this is continuing to weight last season's results until you get even into January. Purdue and Virginia, for example, swapped within a few positions after yesterday's result. Purdue rose to 5th from I believe 7th, and Virginia dropped from somewhere in the middle of the top 10 to only 10th.
At the same time, Sagarin is showing wild swings. Purdue went from 4-3 and 40th to 5-3 and 17th. Virginia went from 7-0 and mid top 10 to 26th. Whether either of those ratings were correct/incorrect before yesterday's game, or are correct/incorrect after yesterday's game is immaterial. What is material is that any rating system with swings that wild is useless as a predictive tool based upon the limited data set that we have. If the swings are that great right now, the signal:noise ratio is just too low to be useful.
KenPom tries to smooth those swings. Which doesn't make him any more correct, mind you, but he will be at least
consistently correct or incorrect. He also uses previous season's results as part of that smoothing function, which can be hard in cases where, like Purdue and Virginia, a huge portion of minutes, scoring, and team chemistry graduated or went to the NBA.
Personally, I'm more inclined to agree with the KenPom approach, at least early in the season. While I think Sagarin will round into form later, it makes his system basically useless as a predictor until there is enough data. Team strength of schedule can vary extremely widely in the non-con, and the teams are breaking in new players, new starting lineups, and new rotations. While there are areas where KenPom will be proven wrong over time, Sagarin is too volatile to be proven anything this early in the season anyway.
