INTERESTING STAT in updated Sagarin ratings

[rvwsleep]

With NC's loss dropping them to #17, the top 15 ranked teams have only 3 losses among them.
A combined 96 - 3
All of the losses inflicted by , you guessed it, UCONN !!!

 

[Wally East]

So, no other Top 15 teams have played anyone else in the Top 15.

Slackers.

 

[Samanthabrown3]

So, no other Top 15 teams have played anyone else in the Top 15.

Good point. UConn has played a couple of the higher ranked teams and still done pretty well. . .

 

[Phil]

Nebraska goes down. #10 in AP, but not close to that in Sagarin. Sounds like Sagarin knows something.

 

[Phil]

AP #12 TAMU in trouble against Syracuse. Howeevr, only ranked #35 per Sagarin, maybe Sagarin knows something.

EDIT TAMU goes on 7-0 run while I was typing this, maybe I should wait until game over

 

[DobbsRover2]

Also interesting to see the teams with the #4 and #5 SOS ranked at 308 and 330 in Sags, them being Prairie View and Arkansas-Pine Bluff. Obviously the fodder for the top-ranked teams that are sporting those #300+ SOS ratings.

 

[Icebear]

PSU beats Wisconsin after losing to Nebraska last week in their 3rd overtime game of the year using a team of 65 scholarship players. They finish at 7-5 and a hair's breadth from 8-4. Maybe the NCAA should cut all football teams to 65 scholarship players.

 

[vtcwbuff]

I'm not sure how Sagarin computes his schedule rank but UConn's is 205 despite playing teams currently rated #3, #11 and #17.

If you want a clear indication of the disparity between the haves and have nots take a look at the spacing between the top teams and the rest of the division. UConn at #1 has a Sagarin predictor of 106.19. The #3 rated team, Stanford has a predictor of 96.68. Duke at #2 is roughly in the middle. The difference between UConn and Stanford is about 10 points. The predictor difference between the remaining 22 teams in the top 25 is also about 10 points!

 

[Phil]

I'm not sure how Sagarin computes his schedule rank but UConn's is 205 despite playing teams currently rated #3, #11 and #17.

I'm fairly sure it is simply an average of the rankings of the opponent.

 

[DobbsRover2]

I'm not sure how Sagarin computes his schedule rank but UConn's is 205 despite playing teams currently rated #3, #11 and #17.

If you want a clear indication of the disparity between the haves and have nots take a look at the spacing between the top teams and the rest of the division. UConn at #1 has a Sagarin predictor of 106.19. The #3 rated team, Stanford has a predictor of 96.68. Duke at #2 is roughly in the middle. The difference between UConn and Stanford is about 10 points. The predictor difference between the remaining 22 teams in the top 25 is also about 10 points!

The schedule computations are not quite as bad as for straight RPI, but it still amounts to a top 10 team getting more credit for playing say six #100-110 ranked teams than for playing three #2-20 teams and having three exhibition-style games against #200-320 teams like Hartford, Boston U and Monmouth. The schedule assessment is a statistical tool that doesn't reflect reality, but as someone who works on and develops data ranking systems, it's more onerous to create a system with a more balanced and nuanced view. Personally, if I were Jeff, I'd scrap the SOS analytics that he uses and go with a regression based tool that weights a team's performance more closely against teams around its range, so top 25 teams would get more credit for playing top 50 teams but #150+ teams would have minimal impact, while it would be the opposite for a #250 team, which would be evaluated most closely against teams around the #225-275 teams and get minimal credit for getting smacked by a top 10 team, which is just a ridiculous way to get your ranking updated.

But I guess we should be glad to have a flawed but overall very solid evaluator like Sagarin versus other systems that put a team like the 2003 NC squad UConn at #3.

 

[Phil]

Measuring schedule strength is harder than it sounds, because it can't in general be done looking only tat the teams played.

Consider the following extreme example:
Which is tougher,
Schedule A

• #1 UConn
• #301 Maryland Eastern Shore

OR
Schedule B

• #150 Montana State
• #152 Hofstra

Sagarin rates these two schedules, each with an average opponent rating of 151, as equivalent.

A team ranked in the upper half of the bracket, but especially those around #75 Wisconsin see the first schedule as an almost certain win and an almost certain loss, while the second is an almost certain 2 wins. There is no question, the first is harder.

In contrast, consider teams in the lower half, especially around #225 Elon. They see the first schedule as an almost certain win and an almost certain loss, but the second as two almost certain losses. They'd prefer the first schedule, thinking of the second as harder for them.

 

[DobbsRover2]

Statistics and analysis are never easy, and the more complex systems sometimes make fans uneasy because they don't understand all the gear work that runs them. Sagarin does say that the SOS is generated by a simple mean average for opponents with home\away factored in. But I'm not sure if a opponent's score is fixed when you play them or whether every opponent's score is reconfigured as the season progresses. If you play the #1 team in the nation and get a great kick for your SOS does that number go down if they subsequently lose their top two stars and go into a tailspin? Or what happens if it's the opposite and that top 20ish team you play in November ends up playing in the NC game? Do you get a later season boost? There are all sorts of other contingencies that a complex ranking system needs to adjust for that are hard to factor for in the algorithms.