Post Weekend Musings

Happy Monday!

Or not. Depends on how your team did on Saturday, I suppose. 

Anyway, I figured that this season I would split the Top 25 and ranking comparisons instead of lumping it all into a single Sunday word vomit. At this point in the season (very early still, but enough data points to start talking) we can start to point out the voter bias again. Also, I am adding the Massey CFB Ranking Composite to compare, as well as manually adding my rankings to see if we move the needle at all. I will also take a look at conference rankings.

Ray vs the AP

If you've followed any of my No Season Left Behind series, you'll know that the AP and I do not get along at all. Some seasons we are closer to agreement, like maybe 1 in 10, but even a broken clock is right twice a day. Right now, we're pretty far apart in general:

	AP		RR	Diff
1	Alabama	SEC	3	2
2	Georgia	SEC	2	0
3	Oregon	Pac-12	14	11
4	Penn State	Big Ten	16	12
5	Iowa	Big Ten	34	29
6	Oklahoma	Big 12	8	2
7	Cincinnati	American	28	21
8	Arkansas	SEC	4	-4
9	Notre Dame	Independent	6	-3
10	Florida	SEC	13	3
11	Ohio State	Big Ten	46	35
12	Mississippi	SEC	10	-2
13	Brigham Young	Independent	35	22
14	Michigan	Big Ten	1	-13
15	Texas A&M	SEC	51	36
16	Coastal Carolina	Sun Belt	53	37
17	Michigan State	Big Ten	12	-5
18	Fresno State	Mountain West	42	24
19	Oklahoma State	Big 12	7	-12
20	UCLA	Pac-12	5	-15
21	Baylor	Big 12	73	52
22	Auburn	SEC	49	27
23	North Carolina State	ACC	30	7
24	Wake Forest	ACC	37	13
25	Clemson	ACC	40	15

As you can see, the voters are close to my ranks 8 or 9 of 25 times. That's not too bad, really, but when they're off they're way off. Iowa, Cincy, OSU, BYU, A&M, Coastal Carolina, Fresno, Baylor, and Auburn are all voted too high by 20 places or more. OSU, A&M, and CC are 35+ places too high, and Baylor 52!

Mean	11.76
Variance	313.52
SD	17.71
SEC	8.86
Big Ten	11.60
Pac-12	-2.00
Big 12	14.00
ACC	11.67
American	21.00
Mountain West	24.00

The mean difference is 11.76 spots. Only the Pac 12 is close to where they should be, and only because of the offset between Oregon and UCLA. There is simply not enough data for voters to say, "Well, OSU is a good school with a great program so we know they'll be better than they are right now and we'll just vote where we think should be." Derp.

Ray and Massey

Ken Massey has put together a composite ranking system that I've talked about before. As I type this, there are 43 different ranking systems that he uses. It's pretty straightforward: each team adds up all of its rankings and he takes the mean, median, and standard deviation and ranks them according to the mean. He includes a lot of algorithm-based systems, and of course I find that appealing.

My system is not included. I'm a nobody, and my system isn't set in stone still. 

Anyway, I find that his composite ranks are usually closer to mine. Usually. Today, that is not the case. 

	Massey's		RR	Diff
1	Georgia	SEC	2	1
2	Alabama	SEC	3	1
3	Iowa	Big Ten	34	31
4	Michigan	Big Ten	1	-3
5	Notre Dame	Independent	6	1
6	Penn State	Big Ten	16	10
7	Oregon	Pac-12	14	7
8	Oklahoma	Big 12	8	0
9	Arkansas	SEC	4	-5
10	Cincinnati	American	28	18
11	Mississippi	SEC	10	-1
12	Florida	SEC	13	1
13	Texas	Big 12	27	14
14	Ohio State	Big Ten	46	32
15	Oklahoma State	Big 12	7	-8
16	Michigan State	Big Ten	12	-4
17	Texas A&M	SEC	51	34
18	UCLA	Pac-12	5	-13
19	Wake Forest	ACC	37	18
20	Brigham Young	Independent	35	15
21	Baylor	Big 12	73	52
22	Louisiana State	SEC	17	-5
23	Kentucky	SEC	15	-8
24	Auburn	SEC	49	25
25	Clemson	ACC	40	15

This really isn't very different than the AP. I'm not super surprised. Any algorithm-based model does better with more data, and like me most of these don't have enough.

Mean	9.12
Variance	258.78
SD	16.09
SEC	4.78
Big Ten	13.20
Pac-12	-3.00
Big 12	14.50
ACC	16.50
American	18.00

Interestingly, neither the composite nor I have Fresno in the top 25. But, similar to the AP differences, we have huge numbers except for the Pac 12.

Ray included in Massey

I would like to reach a level of completeness someday that would allow me to be included in the Massey Composite. Until then, all I can do is add myself manually and see what happens. My voice is one of 44 (43 plus me) and so I don't really expect much movement, if any. 

Here's how I did it:

• Massey composite mean (UGA mean is 2.09 currently) multiplied by 43. Just a simple reversal of the mean. That gives us 89.87
  
• Add my rank (UGA is #2 for me) to that total for 91.87
• Divide the total by 44 for a new mean of 2.088
• Do the same for all 130 FBS schools and then rank them smallest to largest.
• Compare actual Massey to Ray's fantasy Massey

That's it. Basic high school stuff. Here's what we get:

	Massey's w/ RR		Massey	Diff
1	Georgia	SEC	1	0
2	Alabama	SEC	2	0
4	Iowa	Big Ten	3	-1
3	Michigan	Big Ten	4	1
5	Notre Dame	Independent	5	0
6	Penn State	Big Ten	6	0
7	Oregon	Pac-12	7	0
8	Oklahoma	Big 12	8	0
9	Arkansas	SEC	9	0
11	Cincinnati	American	10	-1
10	Mississippi	SEC	11	1
12	Florida	SEC	12	0
13	Texas	Big 12	13	0
14	Ohio State	Big Ten	14	0
15	Oklahoma State	Big 12	15	0
16	Michigan State	Big Ten	16	0
17	Texas A&M	SEC	17	0
18	UCLA	Pac-12	18	0
19	Wake Forest	ACC	19	0
20	Brigham Young	Independent	20	0
23	Baylor	Big 12	21	-2
21	Louisiana State	SEC	22	1
22	Kentucky	SEC	23	1
24	Auburn	SEC	24	0
25	Clemson	ACC	25	0

It takes a very large difference to swing a position change, and even then it's not guaranteed nor is it likely to be more than one spot. That said, we moved Baylor down 2 because the variance was so great. Anyway, a couple of position swaps is about all we get.

I do feel that my system can make a positive contribution to the sport, and that's all I'm really after. Equal treatment for all, and no bias.

Conference Ranks

I've been doing this from the start, but haven't really made much use of it. May as well start sharing, right?

We've all heard or read people from all walks of life prattling on about how awesome their conference is and how weak all the others are, especially from Power 5 conference fans. I suppose if you compare the SEC as a whole to the AAC or MAC, you're pretty safe in making broad generalizations. But is that true of the SEC and B1G? Is the Pac 12 really as weak as ESPN would like us to believe?

In Jan of 2020, I posted this: Conference Strength 2013-2019. That's about as deep as should need to dive into the details. The TL;DR is this: take the average 1st, 2nd, and 3rd level points for each team in a conference, add them up, and rank them. That's it.

Rank	Conf	Score	% of
1	SEC	0.2342
2	B1G	0.2133	91.1%
3	IND	0.1777	75.9%
4	BIG 12	0.1767	75.5%
5	ACC	0.1641	70.1%
6	PAC-12	0.0751	32.1%
7	MTNW	0.0674	28.8%
8	AAC	0.0651	27.8%
9	SUN BELT	0.0641	27.4%
10	C-USA	0.0586	25.0%
11	MAC	0.0397	17.0%

Rank is self explanatory, as is Conf, and Score was explained above. What is "% of?" That is the conference score in relation to the #1 score. What it means (and I explain this in the above linked post) is this: 

• If the average SEC team played the average B1G team 1000 times, and both conferences had equal scores, we would expect each team to win 500 games.
• The B1G is currently 91.1% "as good" as the SEC, which means that the average B1G team should win 91.1% of the 500 games, or 456 games
• The average SEC team should win 544 games of the 1000.

A key here is sample size 3 or 4 crossover games is not enough data to draw conclusions from.

Right now, the SEC is the best conference. Will that hold? Hard to say. We head into hardcore conference play for everyone now, so the whole cannibalism thing is going to rear its ugly head.

Anyway, I hope you enjoyed this. We'll do it again next Monday!

Ray