The RTT computer BlogPoll ballot raw numbers, week ten

by Joel on Nov 2, 2008 8:05 AM EST in College Football Comment 29 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.

I'm posting this early this weekend because I'll be at the Titans game most of the day. I've highlighted some teams that immediately jumped out as needing review, but until the major polls come out I can't do a thorough review of where the computer is more than five off from those polls everyone trusts. Yeah, it's all relative. Anyway, I'll do that later.

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; (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; and (6) modified the Conference Ratings so that there were three categories instead of just BCS and non-BCS. The new categories are (a) the SEC and the Big12, (b) the other BCS conferences, and (c) the non-BCS conferences.

Tweaks for this week:

None. The initial results triggered the need for investigation of certain teams, but did not reveal, to me anyway, any systemic problems. Anyone else see anything bizarre that might necessitate a system tweak this week?
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, 15, 9, 8, 7, 6, 5, 4, 3, 2, 1. Then tweaked by the square root of the sum of the squares maneuver.
Current spreadsheet here.

All right, then. The Top 25, according to the computer:

Rank	Team	WL	SOS	PED	RD	3DO	TD	PEO	EXP	OPPG	TO	3DD	AVG R	W AVG
1	Texas Tech	1	11	53	10	3	54	9	68	46	2	23	25.45	63.85
2	Alabama	1	31	14	2	32	4	56	95	6	61	2	27.64	76.02
3	Oklahoma	8	5	34	35	22	59	4	59	51	3	10	26.36	93.19
4	Penn St.	1	56	4	11	9	6	18	38	4	11	12	15.45	104.08
5	Oklahoma St.	8	27	45	36	21	62	2	45	39	7	42	30.36	107.05
6	Texas	8	1	80	4	4	64	3	116	36	9	55	34.55	112.76
7	Florida	13	9	10	13	27	16	6	116	5	31	22	24.36	128.79
8	Boise St.	1	78	2	29	33	12	10	114	2	24	8	28.45	151.28
9	Ohio St.	16	23	8	14	38	8	52	59	8	97	53	34.18	158.16
10	TCU	1	86	5	1	6	2	55	26	3	29	7	20.09	159.09
11	Southern California	13	57	1	5	24	1	7	72	1	10	5	17.82	160.17
12	Georgia	16	2	76	7	36	30	23	90	53	23	39	35.91	165.15
13	Missouri	16	8	92	38	5	86	5	32	55	6	79	38.36	174.43
14	Utah	1	94	42	8	35	7	22	14	18	32	4	25.18	176.64
15	Georgia Tech	16	55	16	33	62	14	39	109	11	72	70	45.18	188.75
16	Northwestern	16	59	41	26	25	46	86	59	22	51	25	41.45	192.73
17	Michigan St.	15	54	24	75	80	60	71	68	37	66	69	56.27	192.86
18	Minnesota	16	49	51	53	66	65	36	112	20	68	21	50.64	193.53
19	BYU	8	107	50	52	1	48	8	49	26	17	109	43.18	217.75
20	Air Force	16	83	26	62	40	22	31	7	17	73	76	41.18	218.31
21	Ball St.	1	117	18	79	8	57	12	53	13	15	67	40.00	224.29
22	Tulsa	8	116	107	40	2	85	1	49	69	1	78	50.55	247.30
23	LSU	26	25	46	17	47	21	53	77	58	33	20	38.45	249.36
24	North Carolina	26	36	29	32	28	45	33	72	33	88	66	44.36	251.91
25	Florida St.	26	46	23	18	16	5	74	53	23	37	1	29.27	255.12

So what human overrides are called for this week? Candidates (those that are around five spots or more off from the major polls) (remember, this will be updated later):

Possibly too high

Boise State
Georgia
Minnesota
LSU

Possibly too low

Texas. Almost certainly. They've beaten Oklahoma and Oklahoma State, so they should at least be above them, dont' you think?
Florida
Southern Cal

Possibly out of order

Texas and their prior opponents.
Oklahoma and Penn State.

Current draft ballot, without any interventions yet:

Rank	Team	Delta

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

Dropped Out: South Florida (#18), Maryland (#21), Connecticut (#25).

We'll use up to four vetoes of the computer rankings. Now soliciting suggestions.

0 recs | Comment 29 comments | Share

Story-email Email Printer Print

Comments

First fix.

TCU’s W/L figure is wrong. Their record is 9-1, so they shouldn’t be tied with the undefeateds with a score of 1.

by Hooper on Nov 2, 2008 8:58 AM EST reply actions 0 recs

Thanks

Fixed, and it moved them down to 12th.

Go Vols!

by Joel on Nov 4, 2008 7:00 PM EST up reply actions 0 recs

Second fix (or note):

Southern Cal is still dragging anchor by having played only 8 games so far. Even though they’ve only lost 1, they’re in the category of only 3 teams at 7-1, which is giving them a W/L score of 13 rather than 8. If you ignore my first comment, a score of 12 would put them between Boise St. and Ohio St. Any score from 8 to 11 puts them between Florida and Boise St. at #8.

So the computer is basically penalizing USC 3 spots due to the lack of a 9th game. The reason that USC lags a game is because they don’t have a conference title game to worry about. They shift the UCLA game over to that spot and add in an additional bye week into the schedule. Very sneaky, I might add. Yet one more reason for the PAC-10 and Big 11 to get their conference numbers up to 12 and get those conference title games in.

by Hooper on Nov 2, 2008 9:02 AM EST reply actions 0 recs

You know as well I do

Jim Delaney won’t do anything that helps the system. :)

by Graysnail on Nov 2, 2008 9:23 AM EST up reply actions 0 recs

Heh.

The ’snail is wise.

by Hooper on Nov 2, 2008 9:26 AM EST up reply actions 0 recs

I've read through this entire thread before . . .

. . . so I know that there are many other comments along this theme. I’m not sure what the answer is, but let me add this to the discussion:

Isn’t possible that they should be penalized for playing fewer games? Those conferences with championship games have, rightly I think, criticized those conferences that don’t. The “dos” risk defeat in another game and usually against one of the better or best teams in their conference. With all of the upsets in college football, risking defeat and winning should perhaps be rewarded, and avoiding the risk should perhaps be penalized. And if this is truly a resume system, might early byes be appropriately factored in? You can’t really just assume that if USC played another game, they would win. Ask Oregon State.

So what do you think? Might this be something positive rather than a problem to correct?

Go Vols!

by Joel on Nov 4, 2008 7:07 PM EST up reply actions 0 recs

It's a matter of preference, really.

Mostly, I was just trying to figure out why the system was hitting those teams. It all depends on how you perceive the effect of playing one fewer game than other teams. On the one hand, they haven’t lost that game yet. On the other hand, they haven’t won that game yet. Do you reward the risk that’s already been taken (as is currently), or do you give benefit of the doubt to those who haven’t played?

If it were me, I’d leave it. First, why should USC get credit for taking extra bye weeks because they don’t have a champ game? Second, why would the PAC-10 or Big 10 ever consider a champ game if there’s no incentive? (Not that they care about our feelings, but the mindset still holds. If all conferences used a standard seasonal system of 12 games and a champ. game, at least you could be consistent about it.)

Lastly, consider this: A 10-2 team who loses their champ. game would be 10-3. As the system currently stands, they would have a better W/L mark than a 9-3 team who never played a champ. game. That inherently rewards the team who earned a champ. game rather than exposing them to the risk of dropping to match a 9-3 team simply because they won their division. Conversely, the 9-3 team really hasn’t earned the right to be equated with the 10-2 (now 10-3) team in terms of the W/L points because they never faced the risk of playing the top team from the other side of the conference.

by Hooper on Nov 4, 2008 8:02 PM EST up reply actions 0 recs

3rd observation (not really a fix):

Texas is getting killed on their PED. They have played the #2, 4 and 9 teams in PEO in Ok. State, Ok, and Texas Tech respectively, which is partly the blame.

Fortunately for Texas, the rest of the schedule is much, much easier than their recent gauntlet. Assuming their numbers improve as their opposition declines, that will fix itself. (Right now, increasing the PED from 80 to 59 puts Texas above Penn State. Considering that the PED is the 3rd most important variable, that’s not really much difference at the top.)

There’s nothing wrong with leaving Texas alone for right now. They can make up that ground pretty easily.

by Hooper on Nov 2, 2008 9:06 AM EST reply actions 0 recs

I'd actually recommend a fix here

That would be adjusting the PED and other rate stats to adjust for quality of opposition. I think I’ve explained how to do this elsewhere, but the problem is that this isn’t easily implemented, and I would do that beginning next year (it’s much easier to set it up off a blank spreadsheet, when you have time0.

Also, I’d switch Texas and Oklahoma State.

by Graysnail on Nov 2, 2008 9:28 AM EST up reply actions 0 recs

True

I agree that Texas should intuitively have a better ranking. But as you said, it’d be a major hassle to change the system right now. And actually, I’m more interested in seeing how the chart holds up at the end of the season as it is.

If the rate stats were adjusted, would you then want to eliminate the SoS column, since is would be represented in the other stats?

by Hooper on Nov 2, 2008 9:31 AM EST up reply actions 0 recs

I haven't decided yet

As a general rule I hate the way most SoS calculations are done, although what we’re doing here seems to be one of the better fixes (really, I still think you need a permanently sliding scale and recalculate SoS on a weekly basis based on prior results).

What I’m not sure of – and I’d have to do the math and/or theory behind it – is if adjusting the rate stats WILL adjust and take into account SoS. I don’t think it would require eliminating it.

by Graysnail on Nov 2, 2008 9:38 AM EST up reply actions 0 recs

4th:

Florida suffers from the same condition as USC. If their W/L number was bumped to 8, they’d be a solid #3 ahead of Oklahoma. #9 puts them just behind OU; #10 virtually ties them with Penn State (but just behind, again); #11 virtually ties them with Texas (but just ahead).

So Florida is getting penalized as many as 4 spots (potential 3 down to 7) for not having played a 9th game yet. I’ll blame this on that late-term major nonconference game against Florida State – a rarity in the NCAA.

by Hooper on Nov 2, 2008 9:10 AM EST reply actions 0 recs

I'm starting to think this is a chronic problem

Given as how you’ve mentioned it with a number of teams, maybe there should be a tweak to the W/L compendium?

I think it’s either that or there are about 9 teams that could be in the top 5 (and about 15 that could be in the top 25).

by Graysnail on Nov 2, 2008 9:31 AM EST up reply actions 0 recs

Heh.

Well said.

Fixing it is a matter of taste. By the end of the season, everybody will have 12 games. The conference champ players will have a 13th game, in which the winner gets a bump and the loser takes a hit. But the spread of records will be better then, so an extra win or loss won’t mean as much as it does right now.

That said, it is really playing games with the system right now. I personally don’t care for midseason polls, but that’s purely an opinion and not something to base a system on.

by Hooper on Nov 2, 2008 9:34 AM EST up reply actions 0 recs

Penn State:

Getting killed because their conference is weak.

by Hooper on Nov 2, 2008 9:11 AM EST reply actions 0 recs

Boise State:

Moving them would be a meh to me. Their SoS is weak partly due to the conference. But I don’t know of too many significant teams who would be willing to schedule them, particularly after their Fiesta Bowl. There’s too much downside and not enough upside for teams like Georgia, LSU, Ohio State, etc. to schedule them. Kudos to Oregon for taking them on, but they paid the price for it.

I don’t mind midmajors getting seemingly high results because of the scheduling trap unless you can demonstrate that they’re actively avoiding significant teams. With Oregon on the slate, it’s a hard argument to make against BSU, so I’m fine with them where they’re at.

by Hooper on Nov 2, 2008 9:14 AM EST reply actions 0 recs

LSU:

8 games again.

If they had played and won a 9th (and the rest of the numbers stood pat), they’d be in a virtual tie with Georgia (but about half a point behind). They are receiving an effective 10-game penalty for not playing 9. The reason this is so much greater than Florida or USC is that there are far more teams bunched up with relatively close scores in that area of the poll. A few points makes a huge difference.

by Hooper on Nov 2, 2008 9:18 AM EST reply actions 0 recs

Georgia:

buoyed by the #2 SoS. They finish with Kentucky, Auburn and Ga. Tech. Of those, only Ga. Tech has promise of helping their SoS. That number should drop from 2 to somewhere around 15-20 by the end of the season. That’s not enough to move them, but it would give a chance for other teams to elevate.

On that note, Georgia’s ranking is actually very robust to decreases in numbers. You have to drop their SoS to 31 to get them into a tie with Missouri, for example. I’m beginning to think they deserve their spot, even with the fiasco, because their early numbers are still very good. It’s the late numbers that are their worst (with the exception of PED).

by Hooper on Nov 2, 2008 9:22 AM EST reply actions 0 recs

Minnesota:

2 losses, 9 games. The 2 losses to OSU (34-21) and N’western (24-17) respectively. Lost of decent wins in there, but no other major names yet. They get Michigan, Wisconsin and Iowa up next, which should lower their SoS a bit. It’s a matter of a very middle-of-the-road kinda schedule that they’ve been largely successful with.

You might check to see how the Montana State game is figured in, but it’s hard to knock their schedule.

by Hooper on Nov 2, 2008 9:25 AM EST reply actions 0 recs

We must define "decent" differently

They avoid Penn State and Michigan State in their schedule, and aside from Ohio State, I give no intuitive weight to any of their other games. Although they lost completely bizarrely.

I think they’re an override to drop out – really, they’re last year’s Mississippi State team playing in the Big 10. Win by turnover margin, lose by turnover margin. Of course, I’d only drop them out if we can find enough teams with a pulse ahead of them…. ;)

by Graysnail on Nov 2, 2008 9:40 AM EST up reply actions 0 recs

By decent

I just meant that it’s not blatantly filled with suck. Only OSU bumps their SoS up, but nobody really drops it down too hard either.

And you nailed the other problem. Nobody else really deserves a raise. You can make the argument for the 3 8-game teams at 23-25, but they’ll naturally rise up once they play as many games as the other teams.

And we’re back to the problem of ranking teams with different numbers of games played.

by Hooper on Nov 2, 2008 9:51 AM EST up reply actions 0 recs

Yep, that's it.

Just checked: by pretending that LSU, UNC and FSU played and won a 9th game (with no other changes), they all fall between Georgia and Utah, and mixed around Missouri.

That would drop Minnesota 3 spots to 21, which is probably much more palatable.

by Hooper on Nov 2, 2008 9:53 AM EST up reply actions 0 recs

Also,

Cincinnati, West Virginia, California, Pittsburgh and Maryland would all jump into the top 20 if their W/L was 16 rather than 26. So you have 8 teams out of the top 20 because they haven’t played a 9th game yet. Even if you assume that 3 of them lose that 9th game, that still would slide Minnesota back 3 or 4 spots.

by Hooper on Nov 2, 2008 10:03 AM EST up reply actions 0 recs