BlogPoll draft ballot: week three

by Joel on Sep 8, 2008 8:43 PM EDT in College Football Comment 9 comments

Third draft of the Ready, Fire, Aim RTT BlogPoll ballotty thing. Complete refresher. Objections about computer polls, subjectivity, reliability of numbers, and more.

So on to this week, which has its very own caveats:

Sticking with simply ordering every team in every category from first to last and then weighting the categories in order of importance rather than using things like total yards and trying to figure out how to turn oranges into apples.
The stats are still very unreliable at this point. Yes, it bears repeating.
Everybody is tied for either first, second, or third in the all-important Win/Loss category.
Re: strength of schedule -- I decided to use the NCAA's "Past Opposition" figures, as this is supposed to be a resume system -- how good have the teams been, respectively -- as opposed to a forecast of how the teams might be going forward.
Since every team has now played at least one game, I re-added Phil Steele's Experience numbers. Okay, so that probably doesn't make sense to you. It was a logistical thing in the spreadsheet. Trust me. Or not.
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. Why these? Because SMQ said so. The AVG R column is Average Ranking, and the last column is Weighted Average, with multipliers of 1-11 for each successive stat category. Why these weights? Because we're experimenting.
Expect Something Approaching Rampant Human Override Without Reproach this week.
Mistakes? Oh, yeah. Solve.

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

Rank	Team	WL	SOS	PED	RD	3DO	TD	PEO	EXP	OPPG	TO	3DD	AVG R	W AVG
1	Florida St.	1	3	14	10	23	6	3	53	1	5	14	12.09	81.64
2	Kansas St.	1	3	21	37	14	14	17	11	14	27	12	15.55	100.64
3	Oklahoma	1	3	11	8	37	8	7	59	28	11	3	16.00	111.00
4	Southern California	1	3	20	3	53	11	10	72	10	7	5	17.73	115.55
5	TCU	1	3	5	2	11	3	66	26	4	46	14	16.45	129.36
6	Kansas	1	2	1	51	11	15	25	32	4	30	33	18.64	135.18
7	Arizona	1	3	10	19	9	7	13	77	14	22	22	17.91	135.91
8	South Fla.	1	3	13	18	22	1	26	66	33	17	9	19.00	137.27
9	Air Force	1	1	23	26	48	5	64	7	4	54	4	21.55	142.73
10	Boise St.	1	3	42	57	16	16	4	114	10	2	19	25.82	163.27
11	Iowa	1	3	6	42	53	19	20	53	2	24	41	24.00	167.73
12	Oregon	1	3	52	36	14	33	32	38	41	1	37	26.18	171.73
13	Penn St.	1	3	49	23	2	43	28	38	22	15	52	25.09	175.82
14	LSU	1	3	35	9	82	22	30	77	26	31	2	28.91	189.27
15	East Carolina	1	3	27	74	21	24	23	6	25	61	45	28.18	197.45
16	Arkansas St.	1	1	22	55	115	29	5	32	22	10	52	31.27	199.91
17	Arizona St.	1	1	17	64	46	31	16	45	29	26	50	29.64	206.27
18	Florida	1	3	19	13	23	12	41	116	8	63	6	27.73	207.82
19	Georgia	1	3	46	28	3	45	12	90	48	9	43	29.82	212.82
20	Duke	2	1	30	68	57	37	78	21	33	40	7	34.00	214.64
21	Utah	1	1	48	30	60	23	33	14	58	55	34	32.45	226.73
22	South Carolina	2	1	7	53	72	9	75	26	22	76	19	32.91	231.91
23	Mississippi St.	2	3	8	15	60	17	72	38	39	64	24	31.09	232.45
24	Oklahoma St.	1	3	62	47	19	64	18	45	76	12	41	35.27	236.45
25	Kentucky	1	3	3	12	100	10	95	38	3	85	12	32.91	236.55
26	Ohio St.	1	3	2	17	55	2	63	59	10	59	60	30.09	238.27
27	California	1	2	15	19	80	39	51	90	41	23	17	34.36	238.45
28	Fresno St.	1	3	31	49	109	73	69	12	10	38	30	38.64	241.64
29	Auburn	1	2	37	13	90	28	94	84	8	53	1	37.36	248.09
30	Texas Tech	1	3	57	63	23	95	52	68	54	14	8	39.82	248.82
31	Indiana	1	3	58	6	92	18	55	100	14	29	34	37.27	249.82
32	Louisville	2	3	26	7	82	13	97	18	46	83	9	35.09	251.64
33	Wisconsin	1	1	43	24	5	50	14	77	33	25	89	32.91	253.91
34	Missouri	1	3	80	30	44	98	24	32	62	4	63	40.09	262.73
35	Miami (Fla.)	2	1	74	19	23	21	83	18	40	102	30	37.55	270.73

Human eyebrows cocked:

Florida State. A single 69-0 shutout will do that for you. Yeah, but Carolina, of the Western variety rather than the Eastern, so, consider them high fallutin' numbers skeptified.
Kansas State. Hmm. Yeah. Well. 69-10 over Montana State and 45-6 over North Texas. I am not impressed by your competition, Mr. Purple Feline.
Oklahoma. Half a Hundred wins over both Cincinnati and Chattanooga, with that Chattanooga win basically happening in the first half alone. Cincy is not bad (No. 63 in these here rankin's.)
Southern Cal. One win against a Virginia team currently ranked No. 93. Perhaps too high, but what else are you going to do with them, what with the showdown against the Buckeyes coming up, huh?
TCU. Stephen F. Austin? Really? Who is he? No.
Duke. Fluke? Jury out. A 24-20 loss to Northwestern isn't entirely shameful. Watching.
Utah. Actually, not bad. 25-23 over Michigan and 42-21 over UNLV. Possibly top 25, I dunno.
Other skeptizoids: Air Force, Boise, Iowa.

So I made some adjustments. Basic rule (with many exceptions) was to mostly ignore the teams about which I and the rest of the pollsters are skeptical and to keep most of the rest in order of the stats. Again, with exceptions. So here's the drafty draft draft:

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

Dropped Out: West Virginia (#8), Brigham Young (#16), South Carolina (#21), Miami (Florida) (#25).

So. Thoughts? Need help for additional adjustments.

0 recs | Comment 9 comments

Story-email Email Printer Print

Comments

So we're basically ignoring gratuitous blowouts for the moment.

Fine with me.

By the way, I like it. Any ranking system that puts Ohio State well outside the MNC is fine with me. It’s nothing personal against them, but I’d like to think both teams have a chance at winning when they play.

by Hooper on Sep 8, 2008 8:49 PM EDT reply actions 0 recs

Yeah

You tell me — is the blowout factor already accounted for in the other stat categories? I’m guessing it is, but yes, that’s something I essentially ignored with the overrides b/c it just didn’t see right. But I’m rambling. Is margin of victory accounted for to a degree with the other factors?

Go Vols!

by Joel on Sep 8, 2008 9:04 PM EDT reply actions 0 recs

Let me play a bit, I'll get back to you on that.

by Hooper on Sep 8, 2008 9:44 PM EDT up reply actions 0 recs

It's a little odd; I'll try not to get too geeky about it. No promises.

It’s hard to see how well blowouts are accounted for because there are so few games. I tried simply keeping track of how many blowouts were achieved versus cupcakes; I figured that a blowout against a “legit” team deserves to be left in the mix. When doing so, the two stats with the best correlation were PED and EXP. However, EXP was inverted; teams with more experience had more cupcake blowouts. So I think that should probably be disregarded as a correlation. After all, that’s as much a function of scheduling as anything else.

In English (maybe): if you look at teams with low EXP, you’ll find that they also had fewer blowouts. If you look at teams with 2 cupcake-y blowouts already, they tend to have a lot of experience.

PED made a little more sense, and PED tracks pretty well with your W AVG score. So the blowouts are probably showing up most strongly through PED. (Side note: the blowouts do show up in most of the stats; these are just the two most obvious.) [English, againish: Cupcake blowouts go along with high PEDs.]

Oh, there were only 3 teams on the list that didn’t have cupcakey-ish blowouts (in my opinion): East Carolina, Utah, and Fresno State. Not exactly traditional names, so kudos to them.

Did that make sense?

by Hooper on Sep 8, 2008 10:31 PM EDT up reply actions 0 recs

Yeah

But I figured that margin of victory would be inherent more in stats like total offense and total defense. Why isn’t that the case? Or is it too early to really tell anything?

Go Vols!

by Joel on Sep 9, 2008 6:07 AM EDT up reply actions 0 recs

Mostly too early.

But it also depends on how you blow a team out. In most cupcake blowouts, all the damage happens in a short amount of time. Then the winning team sends in the band and the losing team runs a lot to kill the clock. You can see the stats equalize somewhat late in those games, so it can mask the blowout a little bit.

I’m guessing that the physical and training advantages are most effective with PED, where you have a well-trained secondary playing 500 with a lesser QB. Between that and an assumedly vicious pass rush, the completion rate is probably somewhere near an AL pitcher’s batting average.

Just guesses.

by Hooper on Sep 9, 2008 8:44 AM EDT up reply actions 0 recs

Ah

Makes sense. Thanks.

Go Vols!

by Joel on Sep 9, 2008 8:57 AM EDT up reply actions 0 recs

Oh, also

Desperate need of some tweak because South Carolina ended up ahead of Vandy. Any suggestions? Tweak the weighting? Or a head-to-head rule? Or is it okay for a team with better stats to be ahead of a team it just lost to?

Go Vols!

by Joel on Sep 8, 2008 9:06 PM EDT reply actions 0 recs

Not necessarily.

Remember that defense is weighted more heavily than offense in this, and South Carolina’s defense rang up some sweet defensive stats, thanks to the NCState no-show offense. Conceptually, you wouldn’t expect that to happen, but the mind has a tendency to autoweight things in favor of the outcome it wants, while your stats don’t even know who won the game.

If you wanted some kind of win/loss record modifier, you could always use Colley’s numbers. They would give you a pure win/loss component that’s uncluttered by other stats. If you wanted that, of course.

by Hooper on Sep 8, 2008 9:47 PM EDT up reply actions 0 recs