Best Conference in NCAAF--Big East?

[MyTurn2Raise]

Ok, based on some feedback from rwperu, I decided to use this new tool I've created and show off on my blog to determine which conference is the best in NCAAF. What I do is use the sagarin predictor to obtain 'spreads' between a base number and all the teams in a given conference. For this exercise, I used the #5, #15, and #50 teams in the country. I then convert that 'spread' into an expected win pct based on a smoothing of all college games of that spread from 1993 through 2006. I then go through all the possible permutations of independent events of those win pct (probabilities) to list the total outcome space of different number of wins and losses. In other words, I project the possible records a team would have playing every team in the conference and list out the resultant expected win/losses and win pct.

For a #5 team in the land playing every conference member:
Big East 6.042W 1.958L 0.755winp
SEC 9.129W 2.871L 0.761winp
PAC10 7.639W 2.361L 0.764winp
BIGXII 9.361W 2.639L 0.780winp
Big11Ten 9.374W 1.626L 0.852winp
ACC 10.602W 1.398L 0.884winp

For a #15 team
Big East 4.830W 3.170L 0.6038winp
SEC 7.249W 4.751L 0.6041winp
PAC10 6.052W 3.948L 0.605winp
BIGXII 7.607W 4.393L 0.634winp
Big11Ten 8.016W 2.984L 0.729winp
ACC 9.082W 2.918L 0.757winp

For a #50 team
SEC 4.543W 7.457L 0.379winp
PAC10 3.811W 6.190L 0.381winp
Big East 3.143W 4.858L 0.393winp
BigXII 5.086W 6.914L 0.424winp
Big11Ten 5.562W 5.438L 0.506winp
ACC 6.139W 5.861L 0.512winp

Current conference rating systems suffer from a linear weighting of teams. Win pcts don't vary linearly with a spread. This ranking actually looks in terms of win/losses.
What comes out is essentially a 3-way tie between the SEC, PAC10, and BigEast with the slightest nose going to the Big East but that is well within the margin of error.

It will be interesting to see how this all shakes out during the rest of the year.

[TruFloridaGator]

Awesome stuff, thanks. Do you think it will change that much by the end of the season?

[MyTurn2Raise]

yeah...I think it can change quite a bit
the SEC in particular has many out of conference games left that mean something
Georgia-G Tech
Florida-FSU
S Car-Clemson will say quite a bit

[MyTurn2Raise]

oh yeah...they'll all change mostly because sample sizes of 7-8 suck

we still have a ton of data points to come out in the future

[TruFloridaGator]

Yeah, that's why I asked, just want to make I understand this [img]/images/graemlins/smile.gif[/img]

[Billy Bibbit]

Can you change one game and run the numbers again, just to give an idea of how much variance there is? For example, say Auburn had beaten South Florida. Thanks for doing this.

[pvn]

[ QUOTE ]
Ok, based on some feedback from rwperu, I decided to use this new tool I've created

[/ QUOTE ]

very awesome imo

Will you keep updating this thread each week?

[TomCollins]

[ QUOTE ]
Can you change one game and run the numbers again, just to give an idea of how much variance there is? For example, say Auburn had beaten South Florida. Thanks for doing this.

[/ QUOTE ]

Since his tool uses the Sagrarin Predictor, chances are you can't do this. Since the Predictor uses margin of victory, a close win by SF isn't much better than a close loss, so it probably wouldn't change things too much.

[MyTurn2Raise]

yeah....tomcollins is right...and sagarin hasn't released his formula ever

right now, SEC, Big East, and Pac10 are all within the margin of error with each other and one or two games wouldn't change that

[rwperu34]

We might have to take a hindsight look at 2004 USC, Oklahoma, and Auburn.