Writers of Pro Football Prospectus 2008

Most Recent FO Features

SimBla15.jpg

» OFI: Don't Make Saban Angry

Notre Dame and Baylor entered the one-loss group in what is shaping up to be an extremely tight race for playoff consideration.

17 Sep 2010

VN: Quandaries and Rivalries

by Bill Connelly

My colleague Brian Fremeau's Wednesday column focused around a rather dicey topic: When making opponent adjustments, what the heck do you do with FCS opponents? Like Brian, I faced this question pretty early on, as I was committing to the task of entering a full season's worth of play-by-plays. Obviously, connectivity between FCS and FBS teams is nearly nonexistent, even with more FBS teams than ever playing FCS opponents. But I was concerned about the overall sample size of data -- this isn't baseball, where a team has more than 5,000 plate appearances in a given year. In college football, almost nobody has more than 1,000 plays in a season. We need to take into account every play we possibly can. So I wanted to use the data from FCS games. But how?

The FCS Question

My approach was different than the one that Brian chose: Tiers. I separate all FCS teams into six tiers based on two factors: overall record versus non-FBS teams and Sagarin rating. Every FCS opponent is then designated in the database as FCS Tier 1 through FCS Tier 6. By the end of each season, each tier of opponent has "played" between 10 and 20 games, and though they can vary drastically in terms of style (obviously one could run into a situation where a flexbone team and a pure, "55 passes per game" spread team are in the same tier), this gives us a pretty good idea of the quality of each tier of opponent. If we want every possible play to count in the ratings, this allows us a way.

(For in-season, week-to-week ratings, I set each team up in tiers based on last season's performance, then flip the switch to the current season as things begin to draw to a close.)

So how has each "tier" of FCS opponent performed through the years? Below is a table showing each tier's record and where the "team" would have ranked in overall S&P+. (In 2005-06, there would have been 125 FBS teams including the six FCS tiers; in 2007-09, there would have been 126.)

"Team" 2005 2006 2007 2008 2009 Avg.
FCS Tier 1 0-8
(100th)
4-9
(90th)
7-7
(95th)
1-14
(110th)
4-13
(106th)
3-10
(100th)
FCS Tier 2 0-7
(120th)
1-13
(97th)
0-16
(119th)
0-14
(118th)
1-13
(120th)
0-13
(115th)
FCS Tier 3 2-10
(112th)
2-11
(121st)
0-9
(124th)
1-12
(124th)
0-16
(118th)
1-12
(120th)
FCS Tier 4 0-10
(117th)
0-13
(123rd)
2-8
(117th)
0-15
(120th)
0-19
(123rd)
0-13
(120th)
FCS Tier 5 0-9
(124th)
0-12
(124th)
0-14
(125th)
0-14
(125th)
0-12
(124th)
0-12
(124th)
FCS Tier 6 0-8
(125th)
0-13
(122nd)
0-17
(126th)
0-16
(126th)
0-16
(126th)
0-14
(125th)

So using last year's final F/+ rankings as a guide, your typical FCS Tier 1 team is Florida Atlantic, FCS Tier 2 is Louisiana-Lafayette, FCS Tier 3 is Tulane, FCS Tier 4 is Eastern Michigan, and FCS Tiers 5 and 6 are a step below New Mexico State and Washington State. (And yes, it still blows my mind a bit that a team that was in the Rose Bowl less than a decade ago was quite possibly the single worst team, BCS or non-BCS, in the Football Bowl Subdivision last year. Come on, Paul Wulff. What are you doing?)

For what it is worth, Jacksonville State (conqueror of Ole Miss) is currently an FCS Tier 1 team, while James Madison (vanquisher of Virginia Tech) is FCS Tier 2. North Dakota State, the Ted Kevorkian for Kansas in Week 1, is FCS Tier 4. All of these ratings could change once 2010 results are taken into account.

In the end, I do not really think my method is any more or less correct than Brian's. We both have what I consider decent logic behind our choices, and I look at this as one more reason why the combined F/+ ranking is the best thing for Football Outsiders as a whole. It combines two reasonably sound approaches to create logical results.

F/+ Top 25 and a Quandary

So since S&P+ records FCS games, surely that means Virginia Tech has suffered a precipitous drop in the S&P+ and F/+ rankings because of its loss to James Madison, right? Not so fast, my friend. The Hokies rank fourth in the S&P+ rankings, higher even than in FEI.

While Fremeau fought the good fight on Wednesday, the fact is that the inclusion (or lack thereof) of FCS games is not the problem with our early rankings. The problem, of course, is that only two weeks' worth of games have been played, and a vast majority of the current rankings are still derived from the preseason projections. Most of the time, this is perfectly fine. Alabama has certainly lived up to its ranking so far, as have Ohio State, Oregon, and a few other top ten teams. But when a team severely overachieves or underachieves compared to its projections, it is going to take a long time to filter out the projections. Eventually, those teams will end up where they belong, but they aren't there quite yet. (See: DAVE and the Houston Texans.)

Is this a bad thing? And if it is, what can we do about it?

At the end of the season, the F/+ rankings should look pretty good, no matter how they look in mid-September. All preseason projections will be phased out by early October, and we will be left with the ranking system we originally intended. After all, the goal (one of the goals, at least) is to determine the team which performed the best all season long. What we track week to week while we get there isn't as important. But if it unsatisfying in some way, should we change it?

There is an alternative. We could somewhat ignore the opponent adjustments early in the season and judge teams primarily by what their raw statistics say. By the time the preseason projections are phased out, so are the raw, unadjusted numbers. The final rankings aren't affected, but those in the first six weeks are.

To show you which teams would benefit from this approach, take a look at the top ten teams according to pure S&P margin (Raw Offensive S&P minus Raw Defensive S&P).

Top 10 Teams According to Raw S&P Margin
1. California (+0.975)
2. Alabama (+0.949)
3. Houston (+0.857)
4. Clemson (+0.855)
5. Nebraska (+0.824)
6. Arizona (+0.819)
7. Wisconsin (+0.745)
8. Iowa (+0.636)
9. Kentucky (+0.607)
10. Baylor (+0.592)

Those are the 10 teams who have dominated their opponents most thoroughly, no matter who those opponents may have been. (And, in basically every case but Alabama's, their opponents have been quite weak.) To show you the difference the raw numbers might make in our Week 2 ratings, you can see both the official F/+ rankings and the "What if" version below. The "What if" version supposes that we combined the current S&P+ with the raw S&P margin to form early-season S&P+.

F/+ Top 25 (After Two Weeks) With Raw S&P
Adjustment
Rk Team F/+ S&P+ Rk FEI Rk S&P+ Rk F/+ Rk
1 Alabama +27.1% 262.1 1 .241 2 292.5 1 +34.8% 1
2 Florida +26.1% 256.6 2 .248 1 251.0 8 +24.8% 2
3 Ohio State +21.1% 245.1 5 .205 4 247.9 10 +21.9% 5
4 Virginia Tech +20.9% 245.6 4 .198 5 223.2 27 +15.4% 17
5 Texas +20.1% 243.8 6 .191 6 246.8 12 +21.0% 6
6 TCU +19.2% 250.8 3 .138 20 249.4 9 +19.0% 10
7 Oregon +18.9% 235.6 8 .209 3 247.5 11 +22.0% 4
8 Oklahoma +18.1% 235.8 7 .191 7 237.9 17 +18.7% 11
9 Boise State +17.6% 234.3 9 .188 8 217.7 39 +13.6% 19
10 LSU +17.0% 233.0 12 .182 9 227.2 22 +15.6% 16
F/+ Top 25 (After Two Weeks) With Raw S&P
Adjustment
Rk Team F/+ S&P+ Rk FEI Rk S&P+ Rk F/+ Rk
11 South Carolina +15.0% 229.3 15 .161 13 232.5 21 +15.9% 15
12 Auburn +14.8% 233.6 10 .135 21 241.2 15 +16.8% 13
13 Clemson +14.5% 225.5 18 .170 11 262.3 2 +23.8% 3
14 Iowa +14.5% 229.0 16 .153 14 251.4 7 +20.2% 7
15 Miami +14.5% 230.5 14 .145 16 218.8 35 +11.7% 27
16 Penn State +14.4% 233.3 11 .130 23 200.7 63 +6.4% 41
17 Georgia Tech +13.9% 221.9 26 .177 10 218.7 36 +13.3% 20
18 USC +13.9% 223.4 22 .169 12 219.4 34 +13.0% 21
19 Georgia +13.3% 225.0 20 .148 15 218.2 38 +11.7% 26
20 Michigan +12.3% 222.9 23 .140 18 221.3 29 +12.1% 24
F/+ Top 25 (After Two Weeks) With Raw S&P
Adjustment
Rk Team F/+ S&P+ Rk FEI Rk S&P+ Rk F/+ Rk
21 Texas Tech +11.8% 223.9 21 .125 24 220.3 30 +11.1% 30
22 Arkansas +10.8% 225.0 19 .099 30 242.4 14 +15.3% 18
23 Wisconsin +10.7% 228.0 17 .081 37 257.4 4 +18.1% 12
24 West Virginia +10.5% 214.7 37 .144 17 217.2 41 +11.2% 28
25 Nebraska +10.3% 221.9 25 .103 29 258.1 3 +19.4% 9

Here are the new teams who would rank in the Top 25 using a factor of the Raw S&P: California (No. 8), Arizona (No. 14), Stanford (No. 22), Houston (No. 23), Michigan State (No. 25).

Factoring in the raw numbers for now would certainly produce a list of teams who have looked most impressive so far. Virginia Tech drops considerably, while teams like Oregon, Ohio State, Clemson, Iowa, and, of course, Alabama, get credit for remaining mostly unchallenged to date. The raw numbers would also introduce a couple of new factors:

1. Volatility. What if California gets run off the field by Nevada tonight? They could fall all the way from eighth to unranked.

2. "Yeah, but ..." "Yeah, but Clemson has only played North Texas and Presbyterian!" ... "Yeah, but California only played Cal-Davis and Colorado!" That sort of thing.

Honestly, both approaches -- the current, conservative, slow-moving approach of choice and the volatile, schedule-be-damned alternative -- have significant drawbacks, and the reason for that is simple: We're two weeks into the season. There isn't going to be a great way to rank teams no matter how much we may want to. But I'm open to opinions in the comments below.

Box Score of the Week

We are going to skip the Box Score feature this week. This column is already getting awfully long. Plus, we were going to look at James Madison-Virginia Tech, but it was too depressing. Let's just say that, taking turnover points into account, the Dukes should have won by closer to 15 than their five-point scoring margin. It was a disaster for the Hokies in every way possible, and I'm incredibly curious to see how they respond this weekend against an East Carolina squad that has a bit of fire to them.

Biggest F/+ Movers of the Year So Far

Despite the conservative nature of the current F/+ rankings, some teams have still seen their lot in life change quite a bit, for better or worse, with their early on-field performances. Below are the teams who have risen or fallen the most in comparison to their preseason projections.

Strongest Rises

Miami (Ohio): 22 spots (Proj: 106th, Current: 84th). This what happens when you hang tight with Florida into the fourth quarter after starting the season ranked awfully low. The Former Redskins will get a couple more shots at the big boys in coming weeks. They play Missouri next Saturday, then make a less-intimidating-by-the-minute trip to Cincinnati on October 9.

Michigan: 19 spots (Proj. 39th, Current: 20th). Their secondary is obviously as shaky as the girl who played Crystal on True Blood this season, but they have Denard Robinson, the man causing sprained ankles and grass stains throughout Midwest high schools. And that's all they've needed to get by so far.

South Carolina: 17 spots (Proj: 28th, Current: 11th). Aside from Alabama, South Carolina has possibly been the most physically impressive SEC team so far this season. What they lack in explosiveness (Marcus Lattimore's longest carry is only 24 yards so far), they make up for in strength and Lattimore's incredible between-the-tackles running ability.

Kent State: 16 spots (Proj: 105th, Current: 89th). They failed to get the memo about the MAC being worse than the Big South Conference (Akron lost to Gardner-Webb this past weekend, while Ball State lost to the Liberty Flames). They whipped Murray State in their opener, then hung tough against an uninspiring Boston College last week. They're not a good team per say, but they have certainly acquitted themselves better than half the MAC.

Largest Falls

Colorado State: 24 spots (Proj: 87th, Current: 111th). The only team who can give New Mexico and New Mexico State a run for their money in the "Worst Early Team of 2010" race. They have lost by a combined 75-9 to Nevada and Colorado. Colorado followed up on their 24-3 triumph over the hapless Rams by getting steamrolled by California in Berkeley. In a little more than a year, coach Steve Fairchild has gone from the potential savior of the program to the man who makes Colorado State fans wonder if 114-year-old Sonny Lubick might still be a better option.

UCLA: 20 spots (Proj: 49th, Current: 69th). When Rick Neuheisel arrived in L.A. two years ago, the idea was rather simple. Hold steady on defense while resuscitating the moribund offense. Instead, UCLA has regressed slightly on defense ... and barely improved on the offense. The move to the Pistol offense this fall might work eventually, but it hasn't yet. Now the Bruins are tasked with keeping up with Case Keenum (if he plays) and the high-scoring Houston Cougars. This is an opportunity for growth, but it is also an opportunity for embarrassment. Houston will score 65 on you if you let them.

Tennessee: 19 spots (Proj: 20th, Current: 39th). We'll call this a product of their being ranked too high to begin with. The Vols take on a vulnerable Florida squad this weekend, and after holding steady against explosive Oregon for quite a while last Saturday night, they could be in position for an upset. But 20th was just too high for them to start the season.

Cincinnati: 18 spots (Proj: 24th, Current: 42nd). I'll admit it: I drank the Kool-Aid. I looked past the losses of Brian Kelly, Tony Pike, Mardy Gilyard, etc., and saw big things out of a team that returned Zach Collaros, Isaiah Pead, etc. But Butch Jones has yet to press the right buttons as Brian Kelly's replacement. As ESPN's Pat Forde said on Twitter during the Bearcats' loss to N.C. State last night, "Cincy cannot block or tackle. Fairly bad combination."

Random Golf Clap and Random Mini-Rant

To the same subject: rivalries. They are everything good and bad about college sports. One of my favorite commercials is the old ESPN bit with the couple on a date. She's driving, and he finds out she's a Michigan fan, so he immediately jumps out of the moving car, rolls to a stop, and shouts "GO BUCKEYES!" I love rivalries in every way. I even get to live through one of the more enjoyable and frightening ones in the country: Mizzou-Kansas, one that has its basis in a bitter war.

In a rivalry, fans often toe the line between witty/spirited and tacky/too far. To me, the "(352): Time to die" shirts are about three steps too far. When Florida receiver Chris Rainey was arrested for aggravated stalking this week, the most disturbing piece of the story had to be the "Time to die" text he allegedly sent to the victim before she called police. Some rivals' crimes may beg for mocking (I, for one, never tire of Chalupagate, for instance). But when this sort of violence is involved, threatened or otherwise, it stops being funny. Besides, Rainey's not even going to be in Knoxville to see the shirts anyway, as he is indefinitely suspended from the team. I'm all for edgy humor and pushing the limits of decency when it comes to rivalries (or television), but this gambit just made me uncomfortable.

Random Reasons to Love College Football

Watching Mississippi State-Auburn last Thursday night, I was again struck with the same wonder that nails me every college football season. Fans of NFL teams are just as obsessive as most (most) college fans, but aside from Green Bay, they are all based in major cities. College football shares the wealth. Starkville, Mississippi, was buzzing during the Bulldogs' tight loss to Auburn. Fans in Murfreesboro, Tennessee, were excited to find out that Dwight Dasher will be returning from suspension in a couple more games. Waco is a-twitter about Baylor potentially making its first bowl run since the Grant Teaff administration. Hell, Gardner-Webb's win over Akron probably had Boiling Springs, North Carolina, as crazy as Boiling Springs can get Saturday night. There is barely an inch of the country not occupied by the unique fan base of a local school, and they are passionate whether we even know they exist or not.

Random Playlist

In honor of a slate of games that seems like something of a letdown after last week's heavy hitters.

"Bottle Let Me Down," by Emmylou Harris
"Don't Let Me Down," by The Beatles
"How Do I Let a Good Man Down," by Sharon Jones and the Dap-Kings
"Lay Low for the Letdown," by Beulah
"Let Down," by Radiohead
"Let You Down," by Dave Matthews Band
"Never Let Me Down," by Kanye West

Then again, the heavyweight slate last week wasn't actually very entertaining, so maybe there's hope yet for these games.

"Hopefully," by My Morning Jacket
"I Hope," by the Dixie Chicks
"Love and Hope," by Ozomatli

In the history of the world, has there ever been a more attractive woman with a full head of white (or at least, salt and pepper) hair than Emmylou Harris?

Upset Watch

Kansas over Southern Miss. Spread: Kansas +6 | F/+ Projection: Kansas by 8.2. This is, of course, a total crapshoot. Did Kansas lose to North Dakota State because they were looking ahead to Georgia Tech? Did they beat Georgia Tech because Turner Gill knows how to slow down the option and/or the Yellow Jackets were looking ahead to North Carolina? We have no idea.

Notre Dame over Michigan State. Spread: Notre Dame +3.5 | F/+ Projection: Notre Dame by 0.4. It was alarming to hear about the concussion-induced blurred vision Notre Dame quarterback Dayne Crist fought through last Saturday. I'm still in favor of mandatory post-concussion downtime, but assuming Crist plays against Michigan State, this should be a pretty even battle. Michigan State has handled its business against lesser opponents so far, but the Irish are easily the most competent foe they've faced.

UCLA over Houston. Spread: Houston -3 | F/+ Projection: Houston by 1.9. After listing UCLA-Stanford as a possible Upset Watch candidate last week, I am using the tight projection to double down on the Bruins.

Clemson over Auburn. Spread: Auburn -7 | F/+ Projection: Auburn by 3.8. Here comes the toughest early test of the FO projections' love of Clemson. They may not win at Auburn, but hanging close and giving themselves a chance to pull this one out will say very good things about them in terms of the upcoming ACC title race. Needless to say, other possible ACC favorites have not looked great. In all, this and Texas-Texas Tech are the games that intrigue me the most on Saturday. Both could be quite enjoyable matchups.

Closing Thoughts

You are tasked with answering two questions this time around: 1) Which of the two early season ranking methods do you prefer (conservative and projections-based, or volatile and schedule-unadjusted)? 2) What other running features would you be interested in reading in Varsity Numbers? The goal for this format was to cover a MMQB-esque number of topics, but is there something getting missed? More numbers? More thoughts about last week's games, or this coming weekend's slate? This is a very enjoyable column to write, and I'd like for it to be a fun read as well.

And a third question: Is it her voice that makes Emmylou so attractive? It has to be the voice, right?

Posted by: Bill Connelly on 17 Sep 2010

26 comments, Last at 21 Sep 2010, 3:34pm by Bill Connelly

Comments

1
by Kevin from Philly :: Fri, 09/17/2010 - 3:54pm

"In the history of the world, has there ever been a more attractive woman with a full head of white (or at least, salt and pepper) hair than Emmylou Harris?"

Probably some French ladies in powdered wigs from the Marie Antoinette era.

10
by justanothersteve :: Fri, 09/17/2010 - 10:48pm

I'd nominate Marilyn Monroe when she was in one of her platinum blonde periods. Or Annie Lennox. If you're thinking of someone older, Lena Horne or Jamie Lee Curtis. But as one of the few gay guys here, I may not be the best judge of these things.

2
by Joseph :: Fri, 09/17/2010 - 4:20pm

I would say the same thing to both Bill and Brian regarding their systems:
Maybe a more rapid phasing out of the projections could be in order. I don't know how quickly they phase out, but did VT drop completely out of the AP for their loss? (I honestly don't know--I don't pay much attention to the AP poll down here in Mexico.) So--maybe 75% projection, 25% reality after 1 game, 50/50 after two, 25/75 after 3, and 100% reality after 4 games. Or, am I missing the boat that your systems can't do that for each team, and you do it for the WHOLE FBS by the date on the calendar, regardless of how many games they've played? This would probably be more volatile, but would possibly make a team like VT drop down somewhat (like to 17th in your "what if" world). Sure, for those guys that have the top 10 raw margin, some teams would be overranked by pounding patsies. At the same time, maybe you could "adjust" their "reality" stats BASED ON last year's F+ finish until having enough data to properly rank the teams this year. For example, you don't get a HUGE bump by blowing out a FCS team, or a "cupcake" in the bottom 20. But Alabama gets a good bump by dominating Penn St.
Applying this to VT--after game 1, their F+ rating is based 75% on your preseason projections, and 25% based on their close loss to BSU. This probably doesn't drop or raise their rating significantly. But after game 2, they are getting 50% F+ projection, and 50% reality (1/2 being very good in week 1, 1/2 being HORRIBLE for losing to JMU.) This probably mimics most polls, no matter who does them. However, would phasing out the projections SOONER actually change the overall system and final rating? (I kind of doubt it.)
In my mind, the advantages are: if an overrated team (you mention Tennessee, for example) begins the season playing poorly, they start falling rather quickly--but they shouldn't drop 15/20 spots PER GAME; if an underrated team (you cite Michigan) begins beating decent/good teams handily, they rise--and if they keep it up, they will rise no matter what system you use.
I perceive that there may be increased movement in the NUMBER of spots the teams ~50-100 move because their ratings are so similar, but I doubt there would be a large movement in their actual RATING.

So, I guess I'm saying that a little more volatility wouldn't hurt either system, and might keep the critics quieter.

4
by Brian Fremeau :: Fri, 09/17/2010 - 4:32pm

Virginia Tech has, in fact, dropped entirely out of the AP top-25. They did not receive a single vote. James Madison received 11 points. East Carolina received 4 points.

I think it is possible to discount projected data more rapidly than we do, but I don't know that it would necessarily make for better ratings overall. Nor do I think that the AP's treatment of Virginia Tech is "better" in terms of reflecting their relative strength than the method we are using. Is Virginia Tech better than East Carolina? The Vegas line is VaTech -20. Per the current FEI ratings, the forecast picks VaTech by 24.

3
by Ben Johnson (not verified) :: Fri, 09/17/2010 - 4:28pm

Bea Arthur

5
by Will :: Fri, 09/17/2010 - 4:50pm

1) Which of the two early season ranking methods do you prefer (conservative and projections-based, or volatile and schedule-unadjusted)?

- I like the volatile approach. The early rankings are fairly useless either way, but it would be kind of cool for "FCS Tier 1" to show up as a top 25 after week 1 some year :)

2) What other running features would you be interested in reading in Varsity Numbers?

-Bring back the Box Score of the Week - it's the one feature I really liked. If you have to save space on the virtual paper, ditch your iPod playlist.

Will

6
by DFJinPgh (not verified) :: Fri, 09/17/2010 - 5:47pm

What's the purpose of early season rankings? As you say, it's really only the end ones that count, and those have all projections phased out (and I-AA separated correctly into their tiers).

I think they're useless - except as fodder for discussion. You've run them one way (slow phasing out of preseason projections) and gotten lots of bitching. Try 'em the other way, see if people bitch less. Pick the one they bitch more about. *sly wink*

Really, it's only when top teams lose to AA opponents that anybody ever complains about the slow-moving projections, and that's happened twice ... ever. Don't change anything.

7
by cfn_ms :: Fri, 09/17/2010 - 5:58pm

What's the point of F/+ to begin with? Isn't it to make accurate predictions of future games? If it is, then what the authors should do is re-examine their weighting procedure, and choose the process that is optimal within the context of making predictions, whatever that turns out to be.

If the point is instead just to provide discussion fodder, then they should do whatever they feel like.

8
by Jeff Fogle :: Fri, 09/17/2010 - 6:47pm

Agree with cfn's question here. Start with first principles. What's the goal of F/+? Are you trying to capture "reality" as best as possible at that moment? What is it you're measuring when you're measuring...and what is it you're ranking when you're ranking in your view BC?

9
by mm (not verified) :: Fri, 09/17/2010 - 7:59pm

To build on what others have said, the goal of these early season ratings should not be to avoid volatility. The goal should be to get as close to the 'true rating' of each team. For 2010, we don't won't know that number until after the bowls are played.

I agree with Joseph that you probably should reduce the preseason projection's importance as you go along. One way might be to use a weighted average of your current method with the 'raw S&P'. Another way might be to regress the preseason projection towards the mean each week by a certain value. Or you could introduce a factor in the math you are currently using.

When you have the time and desire (maybe not until the offseason), I would say:
1) Decide which form of adjustment you want to look at. Make the adjustment a variable.
2) Using every year you have S&P, go week by week and see which value for that variable makes that week's numbers best correspond to the end of the season values. It's possible that the first week or 2 you need to keep the adjustment strong, and then reduce it quickly. The reverse may also be true. Heck, you might find the adjustment needs to be even stronger, in which case we just have to accept the math.

Then, in future years, you use your new weighing factor to determine the S&P for the early weeks. Every offseason, you'll probably need to recalculate it as you add a year of data and further refine S&P.

If you're reducing the value of the preseason rankings, then some teams might bounce around a lot early (like Kansas), but as long the overall results are moving towards the final rankings, that's fine.

12
by Scott de B (not verified) :: Sat, 09/18/2010 - 9:56am

This method works, but only if you assume that there is an ideal relationship between preseason projections, F+, and true talent.

But what do you do if your model shows that rapid phaseout of preseason projections produces the best results in 2001, 2005, 2006, 2008, and 2009, while slow phaseout produces the best results in 2002, 2003, 2004, 2007, and 2010?

Not everything is easily modeled by a simple mathematical formula, alas.

11
by Jeff Fogle :: Sat, 09/18/2010 - 1:08am

Format Comments:

*Don't worry about length. If you're interested in what you're writing about, people will follow. Just go with it. MMQB is a marathon but people read it start to finish. Simmons has some very long columns. Few hate long pieces if it's clear the author is into what he's talking about.

*Take the word "random" out of the subheads. Did it mentally and it read better. Regulars know you do a golf clap and a playlist or whatever. My nieces stopped saying random all the time two years ago. I'm not sure what the new word is actually. I'll ask at the next family get-together.

*Talk about anything in the results/boxscores/summaries that surprised you. If the mainstream media wasn't talking about it, then you're the guy teaching everyone about something important that they probably missed.

In terms of the rankings...don't get me started on rankings. Too late.

*First, the "preseason rankings" are more like "program rankings" based on recent historical data rather than actual 2010 team rankings. The pro guys do the equivalent of "franchise rankings" in the same way...emphasizing recent history rather than long term history of course. Those aren't "2010 team rankings" either.

It's like everyone's trying to throw together different kinds of mud and pretend it helps you see more clearly. They're separate things...and nobody's going to see much clearly in the first few weeks anyway. Nature of the beast.

I could see using the "program rankings" for strength of schedule purposes in the first month of the year. I can't really see them being used to tell everyone how everyone ranks in early September, or mid September. That's selling something that isn't actually on the menu. It's a "here's where teams would rank if nothing changed from prior norms" chart. Stuff changes. In August, call them preseason estimates, then only use them for strength of schedule measures until you have enough from this season. They're not actual rankings of actual 2010 reality.

I keep seeing comparisons to Bill James. Can anyone imagine Bill James in the 1980's putting together "rankings" of Major League teams at the 13-game mark, or the 26 game mark? There's nothing in the toolbox that's going to build a reality replica this early. Why not think of alternatives instead of asking with unsatisfactory approach people prefer?

*Secondly, reality isn't a hierarchy. Stuff clusters. The stars aren't up in the sky in a hierarchy. They're in constellations. If the stathead industry is ever going to capture reality on paper...it's going to be in a form that recognizes the constellations, then does what it can do differentiate within the clusters. There's probably a way to express the data in a way that shows the clustering rather than going 1-2-3-4-5-6-30-31-32-45-46-47, etc... Maybe start the season with an effort to figure out who's in which clusters with early season S+P and FEI measurements. Then, after a month start differentiating more within the clusters, and use those as defacto rankings.

So...offseason provides the program rankings as a general guideline, but NOT any sort of ranking for the teams in 2010. The first month is used to figure out who's in which clusters (powers, near-powers, bowl caliber but not powers, not bowl caliber...basically like your tiers now that I'm typing it out). Once there's four weeks of action, start differentiating within the tiers/clusters. Through the course of the season, emphasizing the clustering rather than laying out a 1-120 totem pole. Totem poles aren't reality. They're an instinctive human construct that may work great in organizing social structures, but they don't capture reality very well.

There's no rule that says everything has to be ranked. Get rid of as much mud as possible.

13
by FireOmarTomlin :: Sat, 09/18/2010 - 1:37pm

Michigan struggling with UMass at halftime. LMAO.

---------------------------------
Men are more ready to repay an injury than a benefit because gratitude is a burden and revenge a pleasure.

14
by FireOmarTomlin :: Sat, 09/18/2010 - 3:06pm

damn, pulling away late against Umass.

Virginia Tech losing @ Half to ECU.
0-3 start?

-------------------
Men are more ready to repay an injury than a benefit because gratitude is a burden and revenge a pleasure.

15
by lionsbob :: Sat, 09/18/2010 - 3:57pm

and UMASS almost made it a huge upset...they lose 42-37...I smell late season swoon for Michigan.

20
by cfn_ms :: Sat, 09/18/2010 - 7:28pm

Sigh.. ECU gives up a REALLY cheap cover, just falling apart late. Tenn gives up a pick six to push the line. USC gives up a cheap last second TD to blow the cover. Probably ND goes ahead and loses by 4 just to screw things up. Sigh...

16
by lionsbob :: Sat, 09/18/2010 - 4:23pm

Alabama's offense is on another level right now.

17
by FireOmarTomlin :: Sat, 09/18/2010 - 5:00pm

I'm not sure Duke could go 500 in the D3, let alone FCS. LOL.

-------------------
Men are more ready to repay an injury than a benefit because gratitude is a burden and revenge a pleasure.

19
by lionsbob :: Sat, 09/18/2010 - 7:24pm

Well they are in the ACC, that is in the FCS right?

18
by FireOmarTomlin :: Sat, 09/18/2010 - 5:11pm

Baylor getting rolled 21-3 and outgained about 220-40 with 1:00 to go in the 1st.
LMAO.

------------------
Men are more ready to repay an injury than a benefit because gratitude is a burden and revenge a pleasure.

21
by johnmartinolive... :: Sat, 09/18/2010 - 11:38pm

142 yards total offense for Tech against the Longhorns. Their defense seems better although I'm not sure how much that has to do with Texas insisting on running the ball up the middle for two yards a pop. They are going with a new 'pro style' offense. Based on that its hard to look past Oklahoma for the Big Twelve South. Texas' defense does look mean though. Loved the commentators' lust for Tubberville during the game. Apparently he is just the opposite of everything Leach stood for, including apparently putting up yards against quality opponents.

22
by cfn_ms :: Sun, 09/19/2010 - 12:47am

the Pac-10 this week? Cal and Wash face-planted, but AZ, Stanford and UCLA are winning BIG. Wazzu at least showed up at SMU, UO kept cruising, and ASU gave highly-ranked Wisconsin a TOUGH, TOUGH game.

23
by Flounder :: Sun, 09/19/2010 - 12:51am

Just got home from Spartan Stadium. That was incredible. Everyone in the stands was so shocked, it looked like a botched play at first because the holder stood up, and ND had so little reaction to it..... craziness.

24
by Joseph :: Sun, 09/19/2010 - 12:55am

Agree w/the highlight I just saw on ESPN.com--call of the year. MSU's coach has a pair of big brass ones. Reminds me of Boise going for 2 against Oklahoma (iirc) in their bowl game a couple of years ago.

25
by witless chum :: Mon, 09/20/2010 - 5:50pm

Late, but Eastern Michigan is a I-A (FBS) team. They're in the MAC and everything. They reportedly draw like a I-AA team and were worse than Div. II Grand Valley State in some recent seasons, but they're a I-A team.

26
by Bill Connelly :: Tue, 09/21/2010 - 3:34pm

Right, but the average FCS Tier 4 team is at the level of Eastern Michigan. That's what I was going for there.