Tuesday, November 06, 2007

My Top 25 Has Given Up

RankTeamDelta
1Ohio State 25
2LSU 24
3Oregon 23
4Kansas 22
5Oklahoma 21
6Missouri 20
7West Virginia 19
8Arizona State 18
9Auburn 17
10Connecticut 16
11Southern Cal 15
12Florida 14
13Michigan 13
14Georgia 12
15Alabama 11
16Boston College 10
17Texas 9
18Virginia Tech 8
19Kentucky 7
20South Florida 6
21Boise State 5
22Arkansas 4
23Wake Forest 3
24Cincinnati 2
25Virginia 1

FYI, the Delta in my ratings is screwed up because I was swamped at work last week and didn't get a ballot in.
I wanted to be the guy who ranked LSU #1 for the entire season, but as I wrote on Saturday night, this LSU team is living right now more on reputation and the effects of its domination of Virginia Tech and South Carolina than it is on actual performance on the field right now. The defense seems to be less than the sum of its parts and Matt Flynn does not exactly inspire confidence. I bet they struggle at Oxford in two weeks. That said, I also bet that they look good against Arkansas and then really good in the SEC Championship Game. If I were an LSU fan, Florida would be the one team that would scare me in Atlanta because they could compel LSU to play a shootout. Kentucky doesn't have the depth to be excellent in December and I'm not sold on either Georgia or Tennessee standing much of a chance if LSU is really paying attention.
The bottom of my ballot was a difficult proposition. As I went through the list of teams alphabetically, I kept putting teams into the poll and saying "they'll be #25" so by the end, I had about eight candidates for that spot. I leaned on the Sagarin Predictor rankings pretty heavily, although I should have had the guts to put Texas Tech (Sagarin #15) ahead of Virginia (Sagarin #50) in the #25 spot. Consider Virginia's ranking this week to be the equivalent of Spurrier ranking Duke at #25 before each season.
If you want an example of the imperfection of human rankings, when I put Arkansas in the poll, all I was thinking about in my mind was Darren McFadden's Bo/Herschel impersonation on his 80-yard touchdown run on Saturday night. I forgot about Arkansas's suspect defense, their negligible passing game, and the fact that they haven't beaten a top 30 team all year. All my mind could see was Humanity Advanced and that thought went directly to my keyboard. Speaking of the flaws of mortal man, here are some teams that are rated significantly higher by humans than by Sagarin's Predictor and vice versa:
Test Group One
Boston College - 8 AP, 24 Sagarin
Georgia - 10 AP, 28 Sagarin
Virginia Tech - 11 AP, 29 Sagarin
Hawaii - 14 AP, 32 Sagarin
Virginia - 23 AP, 50 Sagarin
Test Group Two
Florida - 17 AP, 7 Sagarin
Auburn - 18 AP, 11 Sagarin
Cincinnati - 27 AP, 13 Sagarin
Arkansas - 31 AP, 12 Sagarin
South Florida - 33 AP, 14 Sagarin
Texas Tech - unranked AP, 15 Sagarin
Kansas State - unranked AP, 17 Sagarin
Pay attention to how these teams perform over the last month of the season. We'll have a nice, unscientific mini-test as to how humans perform as compared to one particular computer.

11 comments:

LD said...

6-3 record.

Lost at home to a 4-5 MAC team.

Best win is over a 6-3 team in the middle of a 3 game skid.

Second best win is over a 5-4 team.

Strength of schedule in the high 70s-low 80s.

That's your #10 team.

You're right about not putting Hawaii in your top 25 - what looks like a good record is merely a veneer over a terrible schedule. You're equally wrong in putting UConn in your top 10 on the basis of what looks like a good record. They have a bad schedule (look at the OOC - Duke, Maine, Temple, Akron and a loss). And they should have 2 more losses if not for atrocious officiating.

You can use your subjective opinions to prevent a team from receiving undeserved credit for wins (as you do with Hawaii). You should do the same for Connecticut.

Of course, if you prefer to rely solely on objective reasoning... well... shouldn't Georgia be ahead of the team it hung 42 on 10 days ago?

Michael said...

You really have a thing for UConn, don't you?

If Georgia should be ahead of Florida, then unranked Tennessee and unranked South Carolina should be ahead of Georgia. I have the Dawgs significantly ahead of where Sagarin does.

LD said...

I know I'm shouting in the dark about the Huskies, but if SMQ can convince people how much Hawaii sucks...

The point is that when it comes to UConn, you seem to be just like a lot of voters (which is why I don't like opinion polls generally). You see a shiny 8-1 and think that's as far as you need to go. We should be able to count on you to think a little more in depth about these teams. And you do - you don't have Hawaii.

Have you seen UConn play this year? Seriously. They are not good.

I've been waiting all year for someone to make a reasonable argument in favor of UConn.

"But they're 8-1!" Terrible schedule, two bogus victories.

"But their only blemish is to UVA, who is good!" Nope, Temple and Louisville also should be considered blemishes. And is UVA really that good, or are they 7 points from 3-7?

"But they've beaten Rutgers and South Florida and both those teams were in the top 10!" Not anymore. USF is a shadow of the team they were the first month. Rutgers has a loss to Maryland on and hasn't beaten anyone out of conference or on the road.

"Forget everything else. UConn simply finds a way to win!" So does Hawaii. Where are they?

Do you have any defense?


I think you're over-reliant on Sagarin, but that's just me. Texas Tech, Florida, and South Florida are overrated by Sagarin because of (a) massive offensive output against weak teams (not enough qualification for schedule strength), and (b) early season successes that were against teams significantly different from those today.

For example, South Florida's win over Auburn looks great today because the Tigers have gotten their act together. But the Auburn team they played was without their best RB and with several young players on the field for the first time. Injuries in the WVU game also played a massive role (Pat White didn't play for most of it, and WVU still outgained USF significantly). From a computer's perspective, those two wins look huge for USF, but when put in context, I'm not sure USF wins over either of them more than 20% of the time.

I don't get Sagarin's liking of Texas Tech at all. Their schedule is atrocious. They have one win against a team even slightly mediocre (A&M - the rest of their wins are over simply terrible opponents). The fact that you trust Sagarin's numbers enough to have wanted to rank them 25 is weird to me. 0 wins against good opponents, 0 secondary wins against good opponents (wins by teams they've beaten). And probably 2 of their 3 toughest opponents remain on the schedule. They'll hurt your test group I feel. Colley has Texas Tech at #47, behind 3 MWC teams, 2 C-USA teams and the .500 Colorado team that beat the Raiders on their home field. Sagarin's algorithm is simply off when it comes to Texas Tech - it doesn't account for how terrible their schedule is.

That said, I do think you have a point about looking to a computer instead of the AP poll. Human voters don't do enough legwork to have complete information, sure. But computer rankings also limit and prioritize information. And it ends up with output that may not tell the whole story.

Now that I'm taking a closer look at your rankings, I'm seeing something else... you really seem to favor teams without a win over a good team. Ohio State, Kansas, Missouri, WVU, Arizona State, USC, Michigan, Texas, Virginia Tech, Boise State, Arkansas, and Wake Forest. Not a one of them has a win over another team in your top 25. Half of your best teams can't tally a win over one of your best teams. Any thoughts? LSU has 3 wins over your top 25. Auburn, Georgia, Oklahoma and Oregon have 2. Does beating quality opponents matter?

LD said...

err... Oregon also has 3 wins over teams in your Top 25.

Michael said...

1. Sagarin likes UConn. They're unblemished in a pretty good conference. They did benefit from a bad call against Louisville, but the bad call happened early in the third quarter. There are bad calls in every game. When the game was on the line, they made plays and Louisville did not. I don't know about the Temple call. Yes, there is a case to be made against them, but there are cases to be made against the teams behind them like Florida (no defense), Georgia (two losses to unranked teams), and Michigan (Appalachian State).

2. I thought you were an advocate of straight resume ranking. If that's the case, then how do you justify the "they're a different team now than they were then" line of thinking? Personally, I think that's usually a crutch for recency. There are cases when it can be made. For instance, I can point to specific personnel changes that Michigan has made that make them much better than the team that shat the bed in the first two games. The problem is that I know the Michigan resume a lot better than any others, so it's hard for me to make that assessment for other teams. For instance, why did Auburn suck at the start of the season and then get better? The defense has been consistent. Is it the offensive line maturing? A different Brandon Cox? The natural progression of a Tuberville team? How do I know that the AU team that lost to Miss. State won't make a sudden reappearance? (BTW, I'm somewhat confident that AU will win in Athens this weekend.)

3. Should I really dismiss South Florida's results from earlier in the year? What's different about that team now? And I disagree with your statement about the WVU game, as USF was kicking their ass before White got hurt. The yardage doesn't affect my judgment because USF was in "sit on the lead" mode for the second half.

4. Texas Tech has destroyed most of the teams they have played. They have conclusively shown that they are much better than bad teams. There is some value in that. Sagarin has tested his formula and found that the one that includes MOV has the best predictive value by a significant margin. I'm willing to put some value in that.

5. On the teams that "haven't beaten anyone": you're placing an arbitrary value on beating a top 25 team. What about convincingly beating some teams that are between 25 and 50 (as Ohio State and WVU have done)? What about Kansas going unbeaten and beating the one team that beat Oklahoma (at the same venue, no less)? You're also ignoring the question of "whom did you lose to?" Georgia has more quality wins than Kansas, but they also lost to two unranked teams, whereas Kansas has not lost a game. Auburn has more quality wins than Missouri, but Mizzou's one loss was a close loss to a top five team, whereas Auburn lost at home to USF (a team you don't rate highly) and Miss. State. As mentioned above, Oklahoma has better wins than Kansas, but they also have a loss against a team that KU beat.

Michael said...

One other note: I like Sagarin's rankings because they're easily accessible, the database goes back to 1998, and he juxtaposes rankings with and without MOV. If there are other good sites with reputable computer rankings, I'd be happy to take a look.

Incidentally, finding the Sagarin archives has given me an idea for a column (possibly for the offseason): I want to compare final AP polls and Sagarin rankings to determine which teams are over or underrated (at least as measured against objective criteria). Naturally, I have an agenda here. I've already done the work for Michigan and found that their human rankings match up perfectly with their final AP rankings, so I can make the argument that they are not a media darling (or at least that their perceived media darling status does not bolster their rankings).

LD said...

1. Sagarin is wrong - mainly because his rankings don't consider (as they should) the Temple and Louisville games to be losses. If both were considered losses, it's likely that Sagarin wouldn't look as fondly on them. Yes, the atrocious punt return non-call happened in the third quarter, but it added 7 points to UConn's tally. They only obtained 14 others. If not for the 7 illegally obtained points, UConn loses, whether it happened in the first minute or the 59th. Look up the Temple game. Google it. A go-ahead TD by temple with 40 seconds to go was ruled out of bounds on the field and to the shock of everyone who has seen the play, the replay team (Big East) refused to overrule the ruling on the field. The Big East later admitted they got it wrong. In my opinion, there is no reasonable argument FOR UConn, and there are a lot stronger arguments against them than there are against about 12 teams you have ranked behind them.

2. I'm an advocate for straight objective ranking - but if subjectivity is going to be used, it ought to be used all the way. The "they're a different team now than they were then" I think is a valid subjective method - and it works lots of ways (both for and against recency). For example, Oregon doesn't deserve as much credit for beating the Michigan team that was a certifiable mess early in the year as Ohio State would if it beats Michigan in 2 weeks. You're familiar enough with the Michigan program to know that. There's a flip side to it - like how Tennessee's loss to Cal with a largely healthy Jackson and Longshore has gotten somewhat devauled by Cal's subsequent losses when those guys were injured. The point is that there are subjective judgments that can be made regarding teams in a particular point in time, and which might appear different as weeks pass. Auburn's win over Kansas State looked great the week after they beat Texas, but the shine comes off a little after the Wildcats got worked by terrible Iowa State.

"it's hard for me to make that assessment forother teams": this is why subjective polling is a problem. Because you are able to make certain judgments about teams you follow, but you can't about others. Every team has injuries, or matures, or goes through certain personnel or strategic changes over the course of the season. Playing team X in week 1 can be very different from playing Team X in week 9. The vast amount of knowledge out there makes the job of a pollster to make informed and completed judgments on teams virtually impossible.

3. I think USF deserves credit for the win over WVU, but it was most certainly not "kicking their ass". WVU gained 170 yards more than USF that night. The reason for the loss was turnovers - 5 of them following drives into USF territory (4 inside the 30). Some of that can probably be attributed to the fact that Pat White wasn't running the offense (and he had a pretty decent night before the injury - 100 yards passing and 36 on the ground in a quarter and a half), as one fumble was off a bad snap (communication problems) and another was off a bad handoff. Injuries are part of the game, but a healthy Pat White probably means USF doesn't have their signature win. They had one good drive all night (one TD was a bomb pass on a busted play, another TD was a pick-6), and gave up a half dozen extended drives. That game was a swing point special. that doesn't mean WVU deserved a win, but it also strongly hints that those sort of breaks that USF capitalized on aren't something that could be expected.

4. Yes, there is some value in Texas Tech destroying bad teams. But when that fact matches up against the fact that they haven't beaten any good teams, it leaves a lot to wonder about. Hawaii is pounding a lot of terrible teams, but they're not in your list. Texas Tech's schedule (and especially their wins) are not much better than Hawaii's. I agree that MOV is of value for subjective analysis. But a terrible schedule can skew raw number analysis greatly. And MOV alone doesn't tell the whole story - i.e., garbage time scores by the losing team.

5. I used your Top 25, so I focused on that. If you have a top 50 list, I'll expand the analysis. The question I suppose is "whether 'tis better to have loved and lost than never to have loved at all". Yes, Georgia has 2 wins against your top 25, but also has 2 losses to unranked teams (though, if you're going to give credit for WVU or OSU beating teams 25-50 who are also "unranked", it's probably only fair to not grade too harshly those losses Georgia has had to teams frequently in your top 25 and most likely in that same category - if the 25-50 group isn't too shabby to give credit to some teams, they can't be too terrible to mark against others). What's better? Going 2-2 against top 30-35 teams (as Georgia has done) or going 0-0 against top 30-35 teams (as Kansas has done)? Who is to say that Kansas wouldn't have gone 0-4 had they played 4 teams as good as the 4 Georgia has split against? (of course, it's equally possible that they'd go 4-0) Playing strong opponents should matter in a subjective analysis - because if you operate under the assumption that a team is good already, the fact that they beat a bad team has little meaning. But that also brings me back to a central question. If Hawaii's gaudy unbeaten schedule is summarily dismissed because of it's lack of difficulty (and I believe it's correctly dismissed), why shouldn't Texas Tech and Kansas (who both have SOS above 100, like Hawaii) also be disregarded? Why does the rationale work for some teams but not for others?

Michael said...

Here's the difference between Hawaii and Kansas or Texas Tech: Hawaii has played a number of close games against bad teams. They nearly lost to San Jose State and Louisiana Tech. They simply aren't a very good team away from home. If they were blowing doors off of a bad schedule, then they would be worthy of a spot in the top 25. Kansas and Texas Tech have done to bad teams what I would expect a good team to do: they've destroyed them.

Wes said...

Florida has lost three games. Georgia has lost two games. Georgia beat Florida. And yet you rank Florida ahead of the Dawgs.

What?

Michael said...

Florida has lost to Auburn, LSU, and Georgia, all of which are in my top 20. Georgia has lost to Tennessee and South Carolina, neither of whom are ranked. Look at the common opponents.

Also, Sagarin has Florida significantly ahead of Georgia.

Anonymous said...

Georgia 42, Florida 30. Take away some of the worst officiating of the year in the SEC, and I say Georgia hangs half-a-hundred on the Gators. Yes, Tebow was bruised...but sowwy, that's by design via Urban's playcalling.

Florida's best win: Vandy at home? Meh. Tennessee at home? They're leading the East. That's OK. @ Kentucky don't look so hot no more. The Troy dominance looks pretty good, though the Trojans, like Oregon State recently and Texas A&M of the early '00s, are Just One Of Those Teams (can either be super up or super down depending week-to-week). That's pretty much it. They also struggled with Ole Miss, the SEC's by-far worst team.

Georgia won at Alabama, who destroyed UT. Smoked that aformentioned Ole Miss team. Oh, and beat Florida. (Did I mention they beat Florida? Pretty soundly, actually. Hey, I think Moreno just scored again. Did I mention they beat Florida?)

I think it's pretty much a wash, really. And if the two were to face again, that Georgia would shred Florida's horrible D just as bad as they did before. Which is not to say this Dawg team can't or won't tank down the stretch, just that to this point, they've been pretty much on par (or a little better) than Florida, who they beat. Did I mention they beat Florida?