RR v Massey's v AP v CFI

I Like Big Games and I Cannot Lie

I truly believe that the CFI committee ranks teams below the top 10 in such a way that they can justify their final top 4. And I believe that they make their selections in such a way as to imply big games where none exist (more accurately, every game is big at this point for anyone in the top 15). Manufactured drama filtered through the ESPN machine. I mean, there are some very smart statisticians out there who do market research in such a way as to tell ESPN which match-ups will make them the most money while alienating the fewest possible eyeballs.

That said, big wins can move teams around a lot, as I think I've shown more than once this season.

RR vs the World!

Let's take a look at the AP poll. Be aware that UGA only got 62 of 63 1st place votes, and I don't see anyone with a single vote. It appears that voter 63 may have had an attack of conscience and withheld his vote?

	AP		RR	Diff
1	Georgia	SEC	1	0
2	Alabama	SEC	13	11
3	Cincinnati	American	20	17
4	Oregon	Pac-12	16	12
5	Ohio State	Big Ten	4	-1
6	Notre Dame	Independent	2	-4
7	Michigan State	Big Ten	6	-1
8	Michigan	Big Ten	3	-5
9	Oklahoma State	Big 12	5	-4
10	Mississippi	SEC	12	2
11	Baylor	Big 12	11	0
12	Oklahoma	Big 12	9	-3
13	Wake Forest	ACC	10	-3
14	Brigham Young	Independent	15	1
15	Texas-San Antonio	C-USA	51	36
16	Texas A&M	SEC	19	3
17	Houston	American	39	22
18	Iowa	Big Ten	8	-10
19	Wisconsin	Big Ten	7	-12
20	Pittsburgh	ACC	26	6
21	Arkansas	SEC	21	0
22	Louisiana	Sun Belt	77	55
23	San Diego State	Mountain West	43	20
24	Utah	Pac-12	28	4
25	North Carolina State	ACC	24	-1

Look at Bama, Oregon, and Cincy. That is the direct result of old thinking. Ancient, actually.

Mean	5.80
Variance	221.08
SD	14.87
SEC	3.20
Big Ten	-5.80
Pac-12	8.00
Big 12	-2.33
ACC	0.67
American	19.50
Mountain West	20.00

The mean difference is pretty small, but the SD is rather large. The mean is small due to the offsets (SEC and Pac12 ranked too high, B1G and Big12 ranked too low).

How about the College Football Invitational? I'm all but convinced that they don't actually watch the games like they claim. Only diehard Oregon fans would think the Ducks are better than OSU after watching what the Buckeyes did to Purdue. The committee has chosen to die on the hill of "head to head" this season, and now  they cannot back down without their already questionable integrity questioned even more.

	CFP		RR	Diff
1	Georgia	SEC	1	0
2	Alabama	SEC	13	11
3	Oregon	Pac-12	16	13
4	Ohio State	Big Ten	4	0
5	Cincinnati	American	20	15
6	Michigan	Big Ten	3	-3
7	Michigan State	Big Ten	6	-1
8	Notre Dame	Independent	2	-6
9	Oklahoma State	Big 12	5	-4
10	Wake Forest	ACC	10	0
11	Baylor	Big 12	11	0
12	Mississippi	SEC	12	0
13	Oklahoma	Big 12	9	-4
14	Brigham Young	Independent	15	1
15	Wisconsin	Big Ten	7	-8
16	Texas A&M	SEC	19	3
17	Iowa	Big Ten	8	-9
18	Pittsburgh	ACC	26	8
19	San Diego State	Mtn West	43	24
20	NC State	ACC	24	4
21	Arkansas	SEC	21	0
22	UTSA	C-USA	51	29
23	Utah	Pac-12	28	5
24	Houston	American	39	15
25	Mississippi State	SEC	23	-2

Mean	3.640
Variance	90.32
SD	9.504
SEC	2.00
Big Ten	-4.20
Pac-12	9.00
Big 12	-2.67
ACC	4.00
American	15.00
Mtn West	24.00

Alabama and Oregon. Again. And Cincinnati. Yeah, bottom half isn't great, but look at 6-18!

I will tell you why this excites me. It points out exactly where and why the committee is hedging its selection options. Look at the top 10. At least 2 of the 3 B1G teams will have 2 losses so it's safe to keep teams like Notre Dame and OkSt outside the top 5. They artificially keep Bama and Oregon up there because they are contractually obligated to love on the Tide and they can point at their BS criteria and say they're looking at head to head for the Ducks. But the rest of the rankings tells the real story - they do have a criteria, and they will set it aside to please their ESPN masters. The hypocrisy is nearly overwhelming. Look, it is impossible to mimic a strictly mathematical system and have it vary only in 6 of 25 opportunities all in the same direction (all of the big variances are the committee ranking teams too highly) without making the variances deliberately. They know the correct placement of these teams and they lie to us about it.

And the 75 systems in the Massey's Composite? I do prefer this to either of the above, but I still believe that too many of them try too hard to emulate human voters.

	Massey's		RR	Diff
1	Georgia	SEC	1	0
2	Ohio State	Big Ten	4	2
3	Michigan	Big Ten	3	0
4	Alabama	SEC	13	9
5	Notre Dame	Independent	2	-3
6	Oklahoma State	Big 12	5	-1
7	Cincinnati	American	20	13
8	Michigan State	Big Ten	6	-2
9	Wisconsin	Big Ten	7	-2
10	Mississippi	SEC	12	2
11	Oklahoma	Big 12	9	-2
12	Wake Forest	ACC	10	-2
13	Oregon	Pac-12	16	3
14	Baylor	Big 12	11	-3
15	Iowa	Big Ten	8	-7
16	Texas A&M	SEC	19	3
17	Brigham Young	Independent	15	-2
18	Penn State	Big Ten	17	-1
19	Pittsburgh	ACC	26	7
20	UTSA	C-USA	51	31
21	Arkansas	SEC	21	0
22	NC State	ACC	24	2
23	Utah	Pac-12	28	5
24	Purdue	Big Ten	14	-10
25	Kansas State	Big 12	18	-7

Mean	1.40
Variance	62.67
SD	7.92
SEC	2.80
Big Ten	-2.86
Pac-12	4.00
Big 12	-3.25
ACC	2.33
American	13.00

Look at this. Bama and Cincy are screwing this up, but just look at the rest. Mean difference is tiny! SD is reasonably small. UTSA is the only ridiculous one in the list. Sometimes you can trust the wisdom of the multi-mind.

Conclusions

I gotta say that the CFI rankings variances were a huge surprise. With the final tweak to my system to account for relative strength losses, I feel like I've cracked the code on the ESPN FPI and acquired actual evidence that the committee is full of it. But hey, if you have an argument for Bama, Oregon, and Cincy that isn't "SeC iS aWeSoMe" or "signature win" or "head to head" with some actual evidence that these things matter more than body of work, let's have it.