RTT preseason BlogPoll drafty draft draft ballot: now with 75% more mess!

Remember, this is ready, fire, aim. Emphasis on the post-fire aim. Especially this very first attempt, which is the sculpting equivalent of throwing a wad of clay on the table and poking it a few times before running off to another appointment. Ridicule if you must, but if you do, try to at least get in a few pokes of your own so the rest of us can return the favor.

So . . . the first stated goal from a few days ago was to finish the computer program that will inform the human opinions. Well, FAIL right out of the gate. I did manage some progress toward the goal before face-divoting, though, and it gives us more than enough to get started. So here it is.

The first step was to determine the factors that matter, and fortunately, SMQ has already done the sweaty work for us, divining the anwer to the question of which stat[s] correlate[] most closely to success, with the term "success" roughly equating to the term "record." The results over the past two seasons:

I also wanted to take a look at what Phil Steele considered to be the most important statistics. Steele isn't nearly as publicly methodical as SMQ, so it's much more difficult to determine what he deems important, but there are certain factors that he repeats more often than others in analyzing teams. From my reading, the stats of which he is most fond appear to be (1) offensive yards per game, (2) offensive points per game, (3) defensive points per game, and (4) returning starters.

In trying to figure out the best way to weave Steele's factors into SMQ's sweatervest, I decided first that offensive yards per game was likely the same thing as total offense and decided that it had already been factored in by SMQ, so I dropped it. Second, because Steele seemed to mention offensive points per game more often than yards per game, I decided to rank its importance higher than SMQ's total offense. How much higher was purely arbitrary. Third, because Steele appeared to mention defensive points per game less often than offensive points per game, I ranked its importance lower, although that would seem to cut against the fact that defensive stats seem to be more important overall. The degree of the drop was purely arbitrary. Fourth, experience seemed to be as important to Steele as offensive points per game, so it's ranked about the same in importance. I actually bumped it one higher just because it seemed to make sense to do so. Finally, to break ties, I favored consistency over wider disparity and then went with the most recent over the . . . um, less recentist. Ish. Okay.

The result:

The working theory is that as you work your way down that ranking of importance, the correlation of that particular stat to success decreases. At some point, it may even fail to correlate to success at all. Where is that line? I'll tell you where that line is.

I don't know.

Is the theory even correct? I don't know. Yes, perhaps, maybe, no. In case you've forgotten, we're trying to steer a bullet that's already been fired here.

Of course, all of that stuff above is an attempt to determine the ingredients and measurements of actual records, but what about the actual records themselves? I'm inclined to start with the premise that the winner of a head to head match up should always be ranked higher than the loser, provided they have identical records, and the rest of the factors just help flesh things out. There's a strength of schedule aspect of this idea to monitor, though. For instance, suppose that Westerrn Kentucky beats Georgia in the first game of the season and loses the following week to the No. 120 team in the nation as Georgia beats the No. 2 team in the nation. They'll both be 1-1, and WKU will have the head-to-head, but there is an argument to be made that Georgia is the better team despite having lost to WKU two weeks prior and an even better argument as Georgia continues to beat really good teams as WKU keeps pace against worse teams.

Anyway, at this point, I'm going to say that actual wins and losses trump most all stats and should be given the highest importance. So put wins and losses above pass efficiency defense in that table thingy up there.

Weighting. Consider weighting of the factors to be the most "work-in-progress" aspect of the system right now. Here's the starting point:

Note that I only went down to the 9th level of importance. Why? I'll tell you why.

I don't know.

Experience. Look, nobody's played a game yet this season. These are last year's numbers. What do last year's numbers have to do with this year's expectations? Some a little maybe perhaps notsomuch. The only way I can think of to even attempt to translate last year's numbers to this year's teams is to adjust them for experience, meaning returning starters. For this, I turned to Steele's experience ratings (despite the fact they're largely outdated by this time -- just ask Georgia, Florida, and Southern Cal) and, as this is the preseason and that seems to be the key point of differentiation from last year, I doubled the weighting. To my surprise, it really didn't make much difference. Huh.

Well, doing that put Hawaii in the second spot, which really underscored the need for a strength of schedule correction, so I input the data from the final 2007 NCAA strength of schedule standings and then tweaked the weighting until it looked better, meaning until Hawaii fell to a more comfortable No. 18. Apologies to the Warriors, but you know. Doing that resulted in SOS ranges from 1259 to 748, which looks really strange when the other stat categories range from the single digits to the low 200s, but that's where we're at right now. Expect major tweaking here in the future. If you have ideas, by all means, speak.

So the system is extraordinarily messy and surely unreliable behind the curtain right now, but the results actually don't look all the strange for a preseason poll:

All right, one more time. This is about as far from a finished product as you can get. It's the equivalent of drunk scratching out an idea on a barroom napkin in the dark with his fingernail . It's not set in stone, m'kay? It's the beginning of . . . something . . . whatever we make of it.

The First Major Problem: Not so happy with the weighting of the Experience component. It's too small and insignificant compared to the overall total number, which is giantenormous primarily due to the tweaking (wrenching!) of the weight of the SOS component. The problem, though, is that when I trumped the Experience up to match the SOS, Hawaii ended up back on top again. So hmm and all that. I don't have an idea for a satisfactory fix at this time. In fact, it looks to me like the thing is in far a complete recalibration. Perhaps instead of using actual numbers (such as the number of yards of total offense, the number of points per game, etc.), we should just rank each team from 1 to 120 for every category and then determine the weighting. Thoughts?

Anyway, even if we were completely satisfied with the weighting and the formulating and the other 'ings, this is the preseason, and it's subject to Total Human Override, so what do you think? What looks fine? What looks wrong?

My thoughts on the vomit: There are some things at the top that look a little weird to me, but nothing really jumps out as being completely wrong until you hit Boston College at No. 12. I'm on record as believing that Georgia is overrated, so seeing them at No. 6 when the coaches and AP polls have them at No. 1 doesn't give me the jitters. Hawaii at No. 19 does give me heartburn, though, which is probably a hangover from the memory of their bowl game last year and the peer pressure at work in not wanting to "fall for that trick again," but when I have to twist the life out of the numbers to make it look right to me, I'm a bit wary of arbitrarily knocking them down much further just because "I've learned my lesson." Things sort of jump the track at that point, anyway.

So, my proposed adjustments:

• Boston College down to 25, just so I can keep an eye on them. Matt Ryan's gone, and the impact of that departure is largely unknown, but it just about has to be significant.
• Auburn should be much higher, I think. They're largley favored to win the West. I moved them up a whopping 15 spots. I'm thinking the computer had them so low primarily due to a low wins and losses number. New offense, installed in last year's bowl game. I just think they'll be better.
• I bumped West Virginia down one, just because.
• Southern Cal seemed quite low, and I bumped them up five spots.
• Cincinnati and Kentucky: out. Taking their spots are Texas Tech and Tennessee because I like Ts.

So, here's the drafty draft draft:

Rank	Team	Delta
1	LSU	25
2	Ohio State	24
3	Oklahoma	23
4	West Virginia	22
5	Virginia Tech	21
6	Georgia	20
7	Southern Cal	19
8	Missouri	18
9	Kansas	17
10	Florida	16
11	Auburn	15
12	Brigham Young	14
13	Clemson	13
14	Oregon	12
15	South Florida	11
16	Arizona State	10
17	Penn State	9
18	Texas	8
19	Texas Tech	7
20	Tennessee	6
21	Hawaii	5
22	Michigan	4
23	Illinois	3
24	Oregon State	2
25	Boston College	1

This is a community project. If you have thoughts on either the system or this week's ballot, leave them below.

By the way, I said I'd post the formula, so here's the current spreadsheet. Consider yourself warned: it's messy.