The RTT computer BlogPoll ballot: raw numbers, week eight

by Joel on Oct 19, 2008 9:27 PM EDT in College Football Comment 2 comments

This here's the RTT Ready, Fire, Aim BlogPoll computer ballot. For a complete refresher on what we're doing, go here, here, here, and here. You can also hit the "blogpoll" tag below the post for a list of all of the related stories.

Prior tweaks to the system: (1) added experience ratings; (2) adjusted the weighting with a "square root of the sum of the squares." Why? I'll you why: I don't know. (3) revised the way tied teams were ranked; (4) began sorting the win/loss category based on losing percentage to hopefully alleviate the penalty to teams with early byes; and (5) (a) adjusted the weighting of the win/loss factor and (b) adjusted the strength of schedule factor by subjecting the NCAA past opposition figures to a strength of conference rating. Conferences were divided into two categories: BCS and non-BCS.

Tweaks for this week:

More fooling with the strength of schedule component. The major computer polls (Billingsley, Colley, Massey, Anderson, Sagarin, and Rothman) were generally in agreement that the Big 12 and SEC were the best conferences and that the other BCS conferences were better than the non-BCS conferences, but on the question of which non-Big 12 and non-SEC BCS conferences were strongest, there was a lot of diversity of opinion. What I decided then was to create three categories. The first includes the Big 12 and the SEC, and teams in that category get full credit for their NCAA past opposition percentage. The second group includes the other BCS conferences, and teams in that category have their NCAA past opposition percentages reduced to 75% of whatever the raw number is. The third group consists of all of the other conferences, and teams in these categories have their past opposition numbers reduced to 50%.
Doing that left the top three as Alabama, Oklahoma State, and Texas. Both Texas and Oklahoma State beat Missouri over the past two weeks, and Texas did so much more convincingly, even accounting for the fact that they had the Tigers at home. The Longhorns' SoS rank based on the new formula was 8, and the Cowboys' was 33, so I increased the weight of the new SoS component from 15 to 20, which put Texas on top of Oklahoma State. It still leaves Alabama above Texas, though, and based on what I've seen from those two teams over the past three weeks, I'd have to say that Texas is the better team, but their pass efficiency defense (72) is keeping them below Alabama. The only way I could think of to make the computer spit Texas out on top of Alabama would be to increase the weighting of the SoS factor to a ridiculous level (which makes everything else exceedingly wacky) or to weight the Big 12 more than the SEC. In the end, I decided to leave it, as I believe it's a matter of the SEC being a more defensive league and the Big 12 being a more offensive league. According to SMQ's stats relevancy watch, defense is more important, so I left it. I'll probably overrule it this week, though, and put Texas over Alabama.
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 current weighting of the categories: 100, 20, 9, 8, 7, 6, 5, 4, 3, 2, 1. Then tweaked by the square root of the sum of the squares manuever.
Current spreadsheet here.

All right, then. The Top 30 (so you can see where some of the teams in the traditional polls ended up in the computer rankings), according to the computer:

Rank	Team	WL	SOS	PED	RD	3DO	TD	PEO	EXP	OPPG	TO	3DD	AVG R	W AVG
1	Alabama	1	23	17	4	40	16	48	95	14	60	3	29.18	67.53
2	Texas	1	8	72	2	4	39	2	116	25	10	43	29.27	77.92
3	Oklahoma St.	1	33	33	35	16	47	3	45	40	7	10	24.55	78.81
4	Texas Tech	1	37	53	12	3	58	10	68	43	2	10	27.00	91.32
5	Ohio St.	10	25	10	17	50	10	47	59	12	92	50	34.73	113.23
6	Oklahoma	12	3	24	32	15	34	4	59	42	4	7	21.45	117.85
7	Georgia	12	4	51	3	29	12	27	90	27	26	32	28.45	124.06
8	Penn St.	1	67	4	22	8	8	19	38	6	11	5	17.18	124.52
9	Boise St.	1	78	8	44	33	32	7	114	2	27	22	33.45	154.16
10	Georgia Tech	12	52	6	19	59	5	25	109	5	66	47	36.82	156.00
11	South Fla.	12	62	20	6	19	7	23	66	25	20	12	24.73	160.63
12	Northwestern	12	64	35	39	18	49	82	59	28	44	23	41.18	172.88
13	Minnesota	12	49	66	42	77	83	26	112	32	68	37	54.91	173.71
14	Utah	1	94	50	9	38	9	24	14	31	35	8	28.45	178.54
15	TCU	10	83	12	1	7	1	85	26	3	38	13	25.36	181.07
16	Florida	19	17	21	15	26	15	22	116	8	37	29	29.55	182.98
17	LSU	19	26	46	16	53	24	50	77	44	40	20	37.73	190.87
18	Southern California	19	43	3	8	13	2	6	72	1	12	4	16.64	191.71
19	Florida St.	19	51	18	7	6	3	72	53	17	29	1	25.09	200.46
20	Pittsburgh	19	50	44	53	22	20	75	77	44	52	51	46.09	208.32
21	Boston College	19	65	1	26	44	4	97	68	11	65	20	38.18	218.30
22	Ball St.	1	114	19	87	5	55	9	53	16	15	63	39.73	220.56
23	Cincinnati	19	59	73	13	86	43	21	59	33	39	58	45.73	221.93
24	BYU	12	103	40	56	2	37	15	49	10	24	102	40.91	224.92
25	Tulsa	1	113	106	59	1	99	1	49	80	1	89	54.45	235.44
26	Kansas	28	11	34	29	10	65	14	32	47	14	44	29.82	260.91
27	Michigan St.	26	39	27	69	84	63	69	68	39	70	84	58.00	264.33
28	Kentucky	28	18	2	51	99	18	99	38	7	90	14	42.18	271.66
29	North Carolina	28	40	32	47	35	59	43	72	34	87	67	49.45	273.70
30	Missouri	28	2	95	45	12	100	5	32	71	5	104	45.36	274.88

So what human overrides are called for this week? Candidates (those that are five spots or more off from the major polls):

Possibly too high

Georgia Tech.
Northwestern.
Minnesota.
Florida State.
Cincinnati.

Possibly too low

Penn State.
Florida.
Southern Cal.
Kansas.
Missouri.

Current draft ballot, without any interventions yet:

Rank	Team	Delta

1	Alabama	--
2	Texas	--
3	Oklahoma State	1
4	Texas Tech	1
5	Ohio State	4
6	Oklahoma	--
7	Georgia	8
8	Penn State	5
9	Boise State	3
10	Georgia Tech	16
11	South Florida	6
12	Northwestern	14
13	Minnesota	13
14	Utah	4
15	TCU	6
16	Florida	8
17	LSU	8
18	Southern Cal	11
19	Florida State	7
20	Pittsburgh	6
21	Boston College	5
22	Ball State	--
23	Cincinnati	3
24	Brigham Young	8
25	Tulsa	1

Dropped Out: Michigan State (#11), Missouri (#13), Kansas (#14), North Carolina (#18), Virginia Tech (#19), Wake Forest (#20), Vanderbilt (#23), California (#24).

Up to six vetoes of the computer rankings. Go.

0 recs | Comment 2 comments |

Story-email Email Printer Print

Comments

Wow!

That seemed like it fixed a lot. I think the latter could be fixed by a normalization of the stats generated. What I mean by that is this:

Adjust the stats to reflect how a team performed against another team, but a percentage of a team’s normal output. What I mean by that is this: if Team A allows 150 yards on the ground against Team B, that’s probably bad, right? Well, if Team B normally runs for 300 yards, that’s actually really good – but the stats (I think) only reflect that Team A allowed 150 yards on the ground. This would, in effect, convert the component stats into a rate stat, similar to OPS+ or something like that from baseball. The downside? This is a huge pain to compute; you need a matrix of each individual team’s rankings, and THEN you need to compare them, or some such. If that makes sense.

What this should do is knock down the ACC a bit (great defense doesn’t matter against terrible offense) and inflate the Big 12, but by a smaller amount. That works for me, since my overrides revolve around knocking GT and FSU down – which should be taken care of. Northwestern and Minnesota probably would get knocked down by the bump, too, and I suspect Cincy might have that happen too (but I wouldn’t touch them). On the other side, Florida should get a boost from that, same with USC (well, maybe. Before last week they would’ve).

To be honest, anyone ranked below 15 at this point is a crapshoot anyway. Can you just submit everyone below 15 at 25? That’’d solve even more problems. :)

by Graysnail on Oct 20, 2008 7:09 PM EDT reply actions 0 recs

I think you're right on both counts

It makes sense, and it would be a huge pain to compute. Will have to think about something like that over the summer. For now, I suppose we’ll just have to be content that some of that is accounted for to a degree with the strength of schedule stuff. Thanks.

Go Vols!

by Joel on Oct 21, 2008 9:46 PM EDT up reply actions 0 recs