RTT draft BlogPoll ballot

by Joel on Sep 15, 2008 10:16 PM EDT in College Football Comment 2 comments

Fourth draft of the Ready, Fire, Aim RTT BlogPoll ballotty thing. For a complete refresher on what we're doing, go here, here, here, and here.

Adjustments for this week:

I switched up the way the computer was ranking ties. Rather than have, say, three categories and all teams ranked 1, 2, and 3, I ranked them according to the point at which the new category starts. In other words, if there were only three categories for any particular stat and they were evenly distributed, the teams would be assigned rankings of 1, 41, and 81 instead of 1, 2, and 3. Math guys: any problem with that?
The categories again: Win/Loss, Strength of Schedule, Pass Efficiency Defense, Rush Defense, 3rd Down Offense, Total Defense, Pass Efficiency Offense, Experience, Opponents Points Per Game, Total Offense, and 3rd Down Defense. The AVG R column is Average Ranking, and the last column is Weighted Average.
The initial results included five teams with 2-1 records and two teams with 1-1 records in the top 35. I didn't much care for that result, and so I fiddled with the weighting. In the end, I settled on the following weights for the categories in the immediately preceding bullet: 30, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1. Doing so bumped Kansas -- the highest ranked team with a loss -- down to 27.

All right, then. The Top 35, according to the computer:

Rank	Team	WL	SOS	PED	RD	3DO	TD	PEO	EXP	OPPG	TO	3DD	AVG R	W AVG
1	Southern California	1	43	10	8	35	5	13	49	3	26	28	20.09	112.64
2	TCU	1	43	2	3	18	2	90	22	10	48	3	22.00	118.82
3	South Fla.	1	1	48	17	19	25	22	72	65	13	42	29.55	141.09
4	Kansas St.	1	99	6	32	6	6	11	11	10	18	5	18.64	143.45
5	Utah	1	43	29	15	32	8	29	26	44	43	17	26.09	145.18
6	Florida St.	1	99	38	2	4	1	7	53	2	9	6	20.18	153.55
7	Georgia	1	28	49	12	12	40	19	90	27	31	51	32.73	165.45
8	Oklahoma	1	99	23	14	15	17	2	45	24	5	21	24.18	167.18
9	Nebraska	1	1	47	23	55	65	21	32	25	28	62	32.73	168.00
10	Troy	1	43	1	24	100	7	51	38	14	11	8	27.09	171.09
11	Iowa	1	43	3	27	73	26	39	53	1	47	30	31.18	173.09
12	Penn St.	1	93	20	16	2	23	17	77	19	8	19	26.82	173.18
13	Alabama	1	74	26	1	9	13	60	95	9	55	17	32.73	180.64
14	East Carolina	1	43	31	56	35	29	20	6	33	71	60	35.00	184.64
15	Wisconsin	1	23	50	26	44	51	32	26	23	40	88	36.73	184.82
16	Boise St.	1	43	45	45	16	28	8	114	7	21	31	32.64	190.45
17	Florida	1	99	4	13	19	3	34	116	6	62	15	33.82	191.09
18	LSU	1	99	8	5	40	12	67	77	10	33	14	33.27	203.36
19	Kentucky	1	43	18	7	99	24	93	38	5	83	11	38.36	211.27
20	BYU	1	74	69	37	1	49	3	49	26	10	73	35.64	215.45
21	Connecticut	1	74	7	19	55	15	86	109	8	32	23	39.00	221.55
22	Air Force	1	43	54	43	22	38	115	7	21	53	32	39.00	225.09
23	Auburn	1	43	15	9	108	10	103	84	3	67	1	40.36	225.27
24	Missouri	1	43	61	53	27	90	5	32	56	1	64	39.36	231.73
25	Oklahoma St.	1	99	42	59	12	36	9	38	59	6	6	33.36	232.91
26	Wake Forest	1	1	64	40	74	46	31	113	54	69	22	46.82	241.73
27	Kansas	41	28	16	48	5	33	23	32	31	27	56	30.91	247.00
28	Texas	1	74	43	29	61	66	6	84	17	25	69	43.18	249.82
29	Ball St.	1	74	33	99	3	81	4	53	53	15	81	45.18	260.73
30	Oregon	1	99	25	60	40	52	71	12	50	4	46	41.82	265.82
31	Texas Tech	1	99	39	41	19	76	33	105	37	2	2	41.27	271.82
32	Vanderbilt	1	40	51	38	48	68	73	84	39	93	66	54.64	273.36
33	Northwestern	1	74	32	61	38	53	65	95	19	57	20	46.82	275.09
34	Indiana	1	99	57	4	98	11	57	100	10	20	33	44.55	282.27
35	Colorado	1	99	52	11	34	45	37	114	54	70	83	54.55	282.73

There's still a good deal of weirdness in there, including TCU, Troy, a slew of SEC teams clustered at the bottom rather than the top (!), and other eyebrow cocking results. You can begin to see why, though. Look at TCU's defensive stats. Quite impressive, really, until you consider they came against New Mexico, Stephen F. Austin, and Stanford. We'll know just how inflated those stats are in two weeks when they play Oklahoma. Perhaps that strength of schedule should be weighted more heavily. Of course, doing that would likely send South Florida over Southern Cal, and that doesn't seem right, either. Ideas, anyone?

Huh Bullets:

Texas appears to be ranked way too high by the MSM polls. Not a terribly strong schedule so far and not particularly stellar numbers, either. 28 is a long way from 7.
Ditto Texas Tech. Blinded by offense, are we. The computer is not overly impressed with offense when it comes at the expense of defense.
South Florida. The computer's got these guys much higher than the MSM polls currently do. But their numbers are strong, and a peek at the schedule and results to date don't really make me all that skeptical of them. Kansas is the best one loss team, and so the Bulls get credit where its due.
Ohio State. Hmm. Where are they? 47th, according to the computer:

47

Ohio St.

41

74

19

39

59

14

75

59

33

89

80

52.91

356.18

The Buckeyes are the 8th best one-loss team, per the 'puter. I'll rank 'em, but not very high.
Oregon. Overranked right now, in my opinion. Will get them into the Top 25, but mostly just to watch them.
What about Wake? I'm liking them, but I'm not quite sure why. It's early. Watching.
Kansas.See above. The highest-ranked one-loss team. In both numbers and perception, IMO.
Utah is underranked, I think. We'll find out more when they play Air Force this week.

I'm beginning to believe that the computer should be subject to complete human override until about halfway through the season and then backed off gradually from there. Shoot, perhaps it will need to be chucked all together. Anyway, here's the RTT draft BlogPoll ballot, based only in part on the above table:

Rank	Team	Delta
1	Southern Cal	1
2	Oklahoma	1
3	Georgia	2
4	Florida	--
5	LSU	2
6	South Florida	11
7	Missouri	2
8	Wisconsin	2
9	East Carolina	9
10	Penn State	1
11	Utah	8
12	Alabama	4
13	Brigham Young	13
14	Texas	2
15	Iowa	11
16	Texas Tech	1
17	Nebraska	9
18	Auburn	10
19	Oregon	6
20	Kansas	13
21	Ohio State	15
22	Wake Forest	--
23	Kentucky	3
24	Air Force	2
25	Florida State	1

Dropped Out: Arizona State (#14), Fresno State (#20), California (#21), Illinois (#23), UCLA (#25).

Thoughts?

0 recs | Comment 2 comments | Share

Story-email Email Printer Print

Comments

On the ranking of categorical variables (the 1, 41, 81 thingy): that’s perfectly fine. You’re essentially changing the weighting, and probably making it more what you expected it to be anyhow.

On the weirdness: keep in mind that the computer doesn’t know anything about these teams other than the numbers you feed into it. It doesn’t know that USC is supposed to be a top team, or that TCU isn’t supposed to be in the top 10. With 2 or 3 games in the books, it’s not very realistic to expect the computer to figure out which teams are the elite teams among 120 different teams. In terms of a resume poll, you can see that the results are starting to converge along the lines of our mental expectations. (That is, this week is closer to our reality than last week’s was.) That’s what you want to see. Hopefully you’ll see it reach some reasonable clarity somewhere after week 6, but most importantly, you want it to be zeroed in after the last week. After all, the BCS MNC game doesn’t care where teams are ranked before the last games are played; why should we?

One thing I’d suggest, though; keep track of the way the numbers run for all of your previous versions, too. After the season is over, you might want to take a look at the different schemes you’ll have run through and see how they rank things at the end. You might find out that one of your early editions of this ranking thingy works better than your final version, but that you corrected yourself out of it due to early-return jitters.

Resume polls are the Jedi of all ranking systems. Sure, the trainup to be a storm trooper power ranking is much faster and you get some good action a lot sooner, but in the end, the discipline and training of a good Jedi Master resume ranking allows it to win out with an unbelievable amount of support behind its argument.

by Hooper on Sep 15, 2008 10:39 PM EDT reply actions 0 recs