I'll Take the Machines

Here’s a top 21 poll that isn’t getting much coverage –

1 Oklahoma
2 LSU
3 Auburn
4 Michigan State
5 TCU
6 Missouri
7 Boise State
8 Oregon
9 Oklahoma State
10 Stanford
11 Utah
12 Alabama
13 Florida State
14 Ohio State
15 Arizona
16 Wisconsin
17 Iowa
18 Texas
19 Kansas State
20 Nebraska
21 Mississippi State

Oregon at 8th? Missouri at 6th? What’s going on here?

Well, this is a perfectly legitimate poll that actually has substantial power in the BCS era. What is it?

It’s the average rankings of the six “computer” polls, and counts 1/3rd towards determining BCS rankings. And it’s VASTLY different than what the humans think.

Why?

The obvious answer is the “computers” aren’t subject to the bias and interia of human voters.

Take Oregon. In the computer polls, the Ducks gets 1 second place, 1 seventh place, 1 eighth place, 2 ninth place and one 11th place ranking. Throw out the 2nd and the 11th and your average is 8th. The reason I suspect their rankings are so relatively low is their present weak SOS (111th per the NCAA).

Or take Missouri. Ranked in the mid-teens in the 3 human polls, the computers place them 6th.

When we average the 3 human polls (AP, Harris, USA Today/Coaches) and compare them to the computer averages we find a great deal of disagreement. The following shows how much difference exists, by averaging ranking, from the computers to the human rankers by team (from under ranked by humans to over ranked)-

		Preseason
Team	DIFF	Rank
Missouri	10.67	UR
Oklahoma State	6.67	UR
Kansas State	6.00	UR
LSU	4.00	21
Michigan State	4.00	UR
Texas	4.00	5
Florida State	3.67	20
Mississippi State	3.00	UR
Stanford	3.00	UR
Oklahoma	2.33	7
Arizona	2.00	UR
Auburn	2.00	22
TCU	-1.33	6
Utah	-2.00	UR
Ohio State	-3.67	2
Iowa	-4.67	9
Alabama	-5.00	1
Boise State	-5.00	3
Wisconsin	-5.33	12
Nebraska	-6.33	8
Oregon	-7.00	11

In looking at this list, it’s pretty obvious why the discrepancies exist. Of the 12 teams above under ranked by the humans relative to the computers, 7 were unranked to start the season. Of the teams that were ranked, 3 were 20 or higher (LSU, FSU and Auburn). Only two were top 10 teams (Texas and Oklahoma). Texas has lost twice, and Oklahoma, though undefeated, has only managed to move up to 3rd in the human polls (from 7th).

Of the teams over ranked by the humans to computers, only 1 of the 9 was not ranked to start the season (Utah). The other 8 were all ranked 12th or better.

What this shows is inertia – unreasonable inertia to me – on behalf of human voters. While the computers don’t care where they place a team, human voters clearly have a tendency to make minor adjustments based on preconceptions – the preconception in this case being the preseason rankings.

Thus the much maligned computers aren’t only more equitable here, they are indispensible. Had the season ended with 7 games, and were it up to strictly the human voters of the BCS, you’d be looking at a Boise State – Oregon BCS title game, with undefeated teams like Oklahoma, Auburn and LSU on the outside. This would be despite the fact that the Ducks and Broncos have played teams that have a cumulative 20-31 record. The computer’s influence on the BCS moves Oklahoma up to number 1 with Oregon 2nd, omitting (quite justifiably) Boise State.

There’s a lot of season left to be played, for sure, and it will be interesting to see how the teams over or under ranked by the computers do. But right now I know who’s rankings I prefer.

Who is Number 1?

As I consider my BlogPoll for this week, I’ve rarely found the question more difficult to answer.

The traditional pollsters have Ohio State as their number 1, but this seems (to me at least) more of a default pick borne of typical poll voter laziness. OSU has played a schedule currently ranked 86th (of 120) by the NCAA. Looking ahead at their schedule it seems reasonably likely they may run the table. We may very well be looking at a 2006/2007 scenario again when it comes to the Buckeyes.

The computer polls, not subject to the static inflexibility of the human efforts (not to mention the obvious groupthink), finds Ohio State less than appealing. The Buckeyes range from a high of 4th ranked in the Massey Poll to 15th in Sagarin (though he has them 8th in his “Predictor”.) They would average 8.4 in the computer polls now (Wolfe doesn’t start his until 10/17). Were the BCS released today I’m guessing they would be about 3rd.


So who do the computer polls like? Well, they seem mostly to like the Mad Hatter. LSU is first ranked in A&H, Billingsley, and the Colley Matrix (5th and 8th in the other 2). That’s an average of 3.2, or nearly 6 slots ahead of where the human polls have them at 9th.

Gaudily scoring Oregon, ranked 2nd in both human polls, is all over the map by the computers, ranging from 1st (Sagarin) to 10th (A&H and Colley). I too have trouble with Oregon who, despite the unbelievable offensive numbers (their 43 TD’s are 9 more than the second best team) have played a very soft schedule (110th per the NCAA). I too don’t know what to believe about them.

Like the computers, I also don’t believe in Oklahoma. They are ranked as high as 12th (Sagarin), who actually ranks them 26th in his “Predictor”. According to Sagarin, 25 teams would beat them on a neutral site, including NC State and Texas A&M. In Oklahoma I see a team with the 38th best offense and 56th best defense in the nation. They really aren’t even top 10.

As of this writing I still don’t know who I consider number 1. But I’d bet this much – this season is going to look a lot like 2007 before it is over. Remember 2007? That’s the year 2 loss LSU won it all.

The Polls - Stuck on Stupid

How the @$&*! does this make any sense?

USA Today Poll
RK	TEAM	RECORD	PTS
1	Ohio State (49)	6-0	1455
2	Oregon (6)	6-0	1388
3	Boise State (1)	5-0	1335
4	Nebraska (2)	5-0	1272
5	TCU (1)	6-0	1213
6	Oklahoma	5-0	1193
7	Auburn	6-0	1060
8	Alabama	5-1	1029
9	LSU	6-0	1021
10	Utah	5-0	940
11	Michigan State	6-0	838
12	South Carolina	4-1	835
13	Arkansas	4-1	780
14	Iowa	4-1	710
15	Stanford	5-1	618
16	Wisconsin	5-1	542
17	Florida State	5-1	496
18	Oklahoma State	5-0	478
19	Missouri	5-0	434
20	Arizona	4-1	367
21	Nevada	6-0	329
22	Florida	4-2	242
23	Air Force	5-1	151
24	Michigan	5-1	137
25	West Virginia	4-1	88

The top 7, all undefeated, so who really cares where they land now.  But Alabama AHEAD of South Carolina, who just got done kicking their ever-loving arses? (Not to mention the Gamecocks's sole loss was to undefeated Auburn). And Florida, with all due respect to my Gators, ranked at all?

Here's hoping the BlogPoll shows much more thought.

Somebody Stop this Crazy (Polling) Train

Yesterday morning I was listening to ESPN analyst Robert Smith opine that, based on the current poll positioning, Boise State is in position to play for the BCS if they run the table.

This seems to be a pretty uniform opinion among the sports chattering classes. It’s also totally asinine.

The following is the current top 25 in the Coaches’ Poll with the remaining strength of schedule (games yet to be played). The SOS is broken down into a percentage of wins and losses of teams yet to be played on each team’s schedule. It is arranged by best to worst.

Boise’s remaining schedule is the 2nd worst among top 25 teams, behind only Nevada’s (they are, after all, in the same conference). The Bronco’s rank 87th in the nation by this metric, as their remaining opponents have a record of only 10-18. Notably, 4 of those 10 wins belong to Nevada.

This is the essence of the idiotic inflexibility of traditional poll voting. What the current poll is saying is that teams like Missouri and Oklahoma State – both currently unbeaten – have no chance to play for the BCS title if they would win all their remaining games, even though their opponents have winning records more than twice as good as the teams Boise plays (0.7916 and 0.7857 to 0.3571).

Not to mention the following –

Nevada – 5th best Scoring Offense, 38th best Scoring Defense, 4-0 record.

Boise State – 13th best Scoring Offense, 49th best Scoring Defense, 3-0 record

Yet Boise is ranked 3rd, and Nevada 25th, per the Coaches. Where is the logic in that?

Some adult needs to step up and stop this crazy train before it trashes this season.

Preseason Polls - Blame it on the Media

Gregg Doyel has a great piece today summing up much of my own thinking about preseason college football polls.

Doyel’s opinion – they hurt college football.

As Doyel aptly describes, preseason polls create a “conveyor belt” that simply moves teams along based on where they were ranked before the first game was played, regardless of evidence later gained from actual play on where teams should be. The starkest example of this in the current polls is Boise State ranked 3rd when we now know that their supposed "marquee" win was against a very bad Virginia Tech team. If the pollsters can’t place a team like Boise in the proper spot after only the 2nd week, it can only get worse as inertia sets in and the Bronco roll up wins against lousy opponents.

According to Doyel, the Coaches tried to get rid of the preseason poll, recognizing its weaknesses. Why was it kept? Because the media wanted to keep it, and convinced the Coaches otherwise!

Besides Boise, other examples of ridiculousness of preseason polling in the current USA Today poll include –

The Over Ranked –

- Texas at 4. They’ve played Rice and Wyoming, and rank 72nd in the NCAA in passing offense.

- Florida at 7. Our Gators have struggled to win over Miami (OH) and USF.

- Wisconsin at 11 . Wins over UNLV and San Jose State in which 21 and 14 were scored against the Badgers.

Penn State at 20. A win over FCS Youngstown State, and a loss to Alabama in which they were dominated.

The Under Ranked –

South Carolina at 16. Wins against Southern Miss and Georgia, and the nation’s 12th ranked defense.

Stanford at 19. Just shut out UCLA on the road, and have the nation’s 10th ranked defense.

Michigan at 22. Wins over U Conn and at Notre Dame. Probably the best two wins in college football thus far.

The damning aspect of this is all of these teams are ranked where they are solely because of where they were ranked in the preseason poll. That is the only reason. And that is just plain stupid.

Since the advent of the BCS era, the lowest ranked team ever to win the BCS title was Oklahoma in 2000 at 20th. Most have been considerably better ranked, and no unranked team preseason has ever played for the title. History tells us that, despite what they do, South Carolina, Stanford and Michigan have no shot at the BCS title this year.

So preseason rankings matter very much. Basing so much on so little actual knowledge – apparently to satisfy the media - is idiotic.