Composite pre-season ranking

[Vowelguy]

Put this together the other day based on the 3 pre-season rankings. The last column shows the % of total voting points. ND got nearly all the 1st place votes so it's 99.9% (rounded to 100%). Allows you to more easily see the gaps in the rankings. Lou & Bay are very close, but a much larger gap between Bay & Ms St, and similarly between #12 Tn & #14 Iowa.

 

[Vowelguy]

And comparing these ranks with the BY poll from last spring which had 14 voters.

 

[LoTrader]

Please keep this around....would like to see how this plays out during the season. Thanks for creating.

 

[oldude]

Interesting stuff. Out of curiosity, I took a look at the Top 10 teams to determine how many games they scheduled vs other Top 10 teams. Obviously some teams have "in conference" games vs Top 10 teams, but only 3 teams have scheduled more than 1xOOC game vs a Top 10 team (UConn, Baylor & SC) as follows:

• UConn - 4 OOC/0 conference
• SC - 3/2
• Baylor - 3/0
• Oregon - 1/3
• MS St - 1/2
• Stanford - 1/2
• OR St - 1/2
• ND - 1/1
• Lou - 1/1
• MD - 1/0

Beyond the ranked teams, a number of these schools are going to get sugar highs from the many cupcakes they've also scheduled on their OOC schedules. Those highs may only last until NCAA selection time. The teams that win the ACC, SEC & Pac12 should be just fine. But winning the BIG in particular, and possibly the Big12 as well are no guarantee of a #1 seed if you end up with a loss or two along the way.

 

[Vowelguy]

revisiting this 4 months later

 

[Plebe]

Turned out worse than expected: Texas, Tennessee, Georgia, DePaul, Missouri, South Florida, Duke.
Turned out better than expected: NC State, Marquette, Gonzaga, Utah, Rutgers, Michigan State.

 

[triaddukefan]

Lawd hammercy

 

[IWearShoes]

revisiting this 4 months later

Completely forgot Georgia was that well thought of early on. They showed up in Starkville and a few other times, but what a disappointing season.

 

[HopJim]

And comparing these ranks with the BY poll from last spring which had 14 voters.

View attachment 36071

Nicely done. Thank you for the work.

 

[Vowelguy]

Comparing to current AP top 10, each poll has rankings that are off by an average 3.2 spots.

Using a sum of squared errors approach, the BY poll does slightly better (4.1 v 4.5 err).

 

[LETTERL]

Comparing to current AP top 10, each poll has rankings that are off by an average 3.2 spots.

Using a sum of squared errors approach, the BY poll does slightly better (4.1 v 4.5 err).

Oooooh...as someone who teaches Probability and Statistics at a community college, you made my heart beat a little bit faster.

 

[Plebe]

Oooooh...as someone who teaches Probability and Statistics at a community college, you made my heart beat a little bit faster.

I need a refresher; I used to know this stuff well but I don't use it anymore

 

[Vowelguy]

I need a refresher; I used to know this stuff well but I don't use it anymore

Instead of just looking at the simple difference, square those differences.
Statistically if you've got a bunch of datapoints and you're trying to fit the "best" line through the data, it's been shown that using squared errors [differences] does a better job.

 

[Vowelguy]

bump