Massey comparison/corrections/outliers

posted Oct 7, 2010, 2:08 PM by Ted P   [ updated Oct 8, 2010, 2:13 PM ]
The abbreviation for these ratings on the Massey comparison page is now KNT.  I believe it was KR last year.

When you look at the comparison page, there are red numbers and blue numbers for a team's highest rank and lowest rank.  My ratings are not that bizarre, so if you see those, they USUALLY indicate mistakes and will be corrected if necessary later in the week.

Mistakes are more often toward a lower rank, because most teams are toward the middle, so I notice a team that should be toward the middle if they're mistakenly toward the top, but I won't notice as easily if that team is toward the bottom.

The only team for which I still can't find anything wrong is Tulane.  I've checked everything I can imagine for them, and they are #63.  They are in the 60s in a few of the ratings, and #70 in GBE.  The fact that I added another step seems to have helped them further, and made my ratings rank Tulane the highest.

I think it's a matter of there not being bad losses and the fact that I segregate out I-AA/FCS wins so they do not unduly weaken an overall schedule.  They provide very little credit, but in the second half of the list, the rating is much more about how much you're penalized for losses than how much you're credited for wins.  Also this early in the season, the I-AA/FCS games do carry more weight since the credit given is divided by the number of overall games.  I could start out by dividing by 13, but I think this is best for the context of a week-to-week rating.  It would be too much of an advantage to those who have played an I-A/FBS team every week at this point.  Also, I like it better that the I-AA/FCS wins fade into the background as the season progresses.  Tulane opponent Ole Miss has rebounded nicely from an embarrassing start; and the Green Wave's other loss is to Houston, which despite not having a great schedule, is 3-1 with the only loss coming to a currently top-25 team (UCLA, ironically Tulane head coach Bob Toledo's former team).  Houston, Rutgers, and Ole Miss are all more highly rated as opponents than they are as teams overall.

I didn't like the old formula (where the three Tulane opponents mentioned would be higher {Houston would even be ahead of UCLA} and Tulane would be slightly lower), because I don't like games treated as better as the other games on the schedule change.  Under formulas like that, Arkansas gets more points for beating ULM now that Arkansas has played Alabama.  Losing should not strengthen the wins, no matter who the opponent is.  GBE is the closest to the old formula (that I know of) if you want to compare.  Kansas St. would be #1, however.  That sounds alarming, but the same number of ratings have Kansas St. #1 as have LSU #1 even though LSU's average rank is 9th (which is the 8th-best average) and Kansas St.'s average rank is 20.5 (Which is the 22nd-best average).

I'm not trying to impress anyone, I'm just explaining the things that I look at in order to correct either mistakes or situations that I believe are not properly addressed by my formula.  Either way, I make an effort to notify anyone who might be interested here.