The 2012-2013 College Football Season in Review and a Look Towards the 2013-2014 Season

UNC plays a very similar style as Clemson. How do they get a #20 ranking and Clemson at #64?

Is this just subjective?

 
Last edited by a moderator:
No. Otherwise I wouldn't have ranked Clemson or South Carolina that low nor Michigan and Michigan State that high. In the opening post, I believe I described the process, but I'll describe it with slightly more detail now. Teams are broken into clusters based on similarities from all of the stats the computer is told are used as criteria for breaking teams into clusters. So for this example, completions, yards, yards per completions, completion to touchdown %, rush yards, yards per carry, and rushes to touchdown %, plays, and turnovers, as well as the same stats for the defensive side of the ball, and an adjusted strength of schedule were used as the criteria factors.

The clusters were then ranked based on the mean number of wins. Since all the teams are grouped together based on similarities, the mean number of wins should allow a way to see where teams should have finished. When teams were ranked, special consideration was taken to the strength of schedule. Teams in a cluster that had a negative strength of schedule score (meaning an easier schedule) were grouped with teams from the lower cluster who had a positive strength of schedule score (meaning a harder schedule).

So Clemson and South Carolina happened to be victims of being similar to teams who finished lower than where Clemson and South Carolina should have finished.

But it does help explain some of the bizarre results that happened this year. For example, why did UL-Lafayette play Florida close in Ben Hill Griffin Stadium? The Ragin' Cajuns should have been ranked 20th and Florida should have been ranked 14th; basically, Florida was overhyped and UL-Lafayette was under hyped. All this model does is states who was overhyped and who was under hyped last year.

 
Let's look at like opponents:

NC State: Clemson gave up 48 pts - UNC gave up 35 | UNC > 13 pts

Duke: Clemson gave up 20 - UNC gave up 33 | Back Even

Wake: Clemson gave up 13 - UNC gave up 27 | Clemson > 14 pts

VT: Clemson gave up 17 - UNC gave up 34 | Clemson > 31 pts

GT: Clemson gave up 31 - UNC gave up 50 | Clemson > 50 pts

Maryland: Clemson gave up 10 - UNC gave up 38 | Clemson > 78 pts

So UNC gave up 78 more points against like opponents, and only gave up less than Clemson in the NC State game.

http://espn.go.com/college-football/team/schedule/_/id/153/year/2012/north-carolina-tar-heels

http://espn.go.com/college-football/team/schedule/_/id/228/year/2012/clemson-tigers

 
Bye Bye Big XII said:
No. Otherwise I wouldn't have ranked Clemson or South Carolina that low nor Michigan and Michigan State that high. In the opening post, I believe I described the process, but I'll describe it with slightly more detail now. Teams are broken into clusters based on similarities from all of the stats the computer is told are used as criteria for breaking teams into clusters. So for this example, completions, yards, yards per completions, completion to touchdown %, rush yards, yards per carry, and rushes to touchdown %, plays, and turnovers, as well as the same stats for the defensive side of the ball, and an adjusted strength of schedule were used as the criteria factors.

The clusters were then ranked based on the mean number of wins. Since all the teams are grouped together based on similarities, the mean number of wins should allow a way to see where teams should have finished. When teams were ranked, special consideration was taken to the strength of schedule. Teams in a cluster that had a negative strength of schedule score (meaning an easier schedule) were grouped with teams from the lower cluster who had a positive strength of schedule score (meaning a harder schedule).

So Clemson and South Carolina happened to be victims of being similar to teams who finished lower than where Clemson and South Carolina should have finished.

But it does help explain some of the bizarre results that happened this year. For example, why did UL-Lafayette play Florida close in Ben Hill Griffin Stadium? The Ragin' Cajuns should have been ranked 20th and Florida should have been ranked 14th; basically, Florida was overhyped and UL-Lafayette was under hyped. All this model does is states who was overhyped and who was under hyped last year.
Click to expand...
Excellent job!

I wonder how it would look if you based the clusters solely on strength of schedule, ranked each SOL cluster by performance (e.g., wins), and then combined the clusters based on head-to-head matchups (best overall fit), keeping the intra-cluster rankings intact? otoh, no matter what sort of analysis you use, there will always be debate.

 
Bye Bye Big XII said:
NUance said:
Bye Bye Big XII said:
No. Otherwise I wouldn't have ranked Clemson or South Carolina that low nor Michigan and Michigan State that high. In the opening post, I believe I described the process, but I'll describe it with slightly more detail now. Teams are broken into clusters based on similarities from all of the stats the computer is told are used as criteria for breaking teams into clusters. So for this example, completions, yards, yards per completions, completion to touchdown %, rush yards, yards per carry, and rushes to touchdown %, plays, and turnovers, as well as the same stats for the defensive side of the ball, and an adjusted strength of schedule were used as the criteria factors.

The clusters were then ranked based on the mean number of wins. Since all the teams are grouped together based on similarities, the mean number of wins should allow a way to see where teams should have finished. When teams were ranked, special consideration was taken to the strength of schedule. Teams in a cluster that had a negative strength of schedule score (meaning an easier schedule) were grouped with teams from the lower cluster who had a positive strength of schedule score (meaning a harder schedule).

So Clemson and South Carolina happened to be victims of being similar to teams who finished lower than where Clemson and South Carolina should have finished.

But it does help explain some of the bizarre results that happened this year. For example, why did UL-Lafayette play Florida close in Ben Hill Griffin Stadium? The Ragin' Cajuns should have been ranked 20th and Florida should have been ranked 14th; basically, Florida was overhyped and UL-Lafayette was under hyped. All this model does is states who was overhyped and who was under hyped last year.
Click to expand...
Excellent job!

I wonder how it would look if you based the clusters solely on strength of schedule, ranked each SOL cluster by performance (e.g., wins), and then combined the clusters based on head-to-head matchups (best overall fit), keeping the intra-cluster rankings intact? otoh, no matter what sort of analysis you use, there will always be debate.
Click to expand...
Here's a quick follow up of the cluster to cluster match ups. I'm hoping that the closer the clusters are in terms of standardized wins, the closer the games will be. With anything, there are outliers--due to the sheer randomness of college football. The closest match ups should be from teams in the same clusters.

For the sake of brevity, each week I will attempt to provide two games which turned out closer than they should have and two games that were blowouts.

So here we go:

Week 1--Notable Results

#9 South Carolina 17 vs. (UNRANKED) Vanderbilt 13: I think it was safe to say that everyone was expecting the Gamecocks to not have any troubles with the Commodores, but yet they did. Looking back on the game, it really shouldn't have come as a surprise because Vanderbilt turned out to be a really good team.

Difference in standardized wins = .186

#21 Stanford 20 vs. (UNRANKED) San Jose State 17: Again, it should've been the case that the Cardinal would have dominated the Spartans. But the Spartans almost pulled the upset. But the end of the season rankings revealed that the Spartans finished 39th, whereas the Cardinal finished 10th.

Difference in standardized wins = .186

#18 Ohio State 56 vs. (UNRANKED) Miami (OH) 10: A blowout, as should've been expected. Miami couldn't handle the Buckeye's offense, and their offense couldn't get off the ground. Ohio State ended the season ranked 7th in my rankings whereas the Redhawks finished 95th.

Difference in standardized wins = 1.80

#1 USC 49 vs. (UNRANKED) Hawaii 10: Similar to the Buckeyes opening game, the Trojans dominated Hawaii from the beginning of the game and didn't let up until it was over. USC ended the season ranked 48th in my rankings. Hawaii finished 114th.

Difference in standardized wins = 1.56
Click to expand...
I would only argue that Vanderbilt was not that good of a team last year. They beat exactly ZERO teams that finished with a winning record and 2 of their wins came against UMass and Presbyterian.

 
You must log in or register to reply here.
Back
Top