Writers of Pro Football Prospectus 2008

Most Recent FO Features

BrownMal15.jpg

» Futures: Texas RB Malcolm Brown

DeMarco Murray is the toast of the NFL, but injury and team issues clouded some observers' view of his talent. Texas RB Malcolm Brown might have the same problem this winter. 

30 Oct 2009

Varsity Numbers: Eight Turnovers!

by Bill Connelly

Fitting for Halloween, this week's big matchups all come out after dark. We've got undercard matchups like Cincinnati-Syracuse (we've got the feeling this one will be close for no apparent reason) and Ole Miss-Auburn at noon, and The Event Formerly Known as the World's Largest Cocktail Party in mid-afternoon, but it's all just a lead-up to the week's biggest games, Oklahoma State-Texas (projection: Texas by 5.9) and USC-Oregon (projection below).

But as always, before we dive into this week, let's take one more look at last week.

Box Score of the Week

For the most part, we will try to bounce around from conference to conference for the VN Box Score of the Week, or at least from division to division. But in following up Colorado's weird victory over Kansas last week, we had no choice but to take a look at Iowa State's downright insane 9-7 win over Nebraska, for very obvious reasons.

Iowa State 9, Nebraska 7

Nebraska: eight turnovers, seven points ... and a two-point loss.

ISU

NU

Close % 100.0%
Field Position % 16.4% 46.2%
Leverage % 58.2% 69.2%
TOTAL
EqPts 10.3 5.5
Close Success Rate 28.4% 41.5%
Close PPP 0.15 0.09
Close S&P 0.44 0.50
RUSHING
EqPts 4.1 2.4
Close Success Rate 29.8% 44.4%
Close PPP 0.09 0.09
Close S&P 0.385 0.535
Line Yards/carry
2.43 3.00
PASSING
EqPts 6.2 3.1
Close Success Rate 25.0% 39.5%
Close PPP 0.31 0.08
Close S&P 0.561 0.476
SD/PD Sack Rate
0.0% / 9.1% 0.0% / 6.3%
STANDARD DOWNS
Success Rate 38.5% 48.9%
PPP 0.20 0.12
S&P 0.580 0.613
PASSING DOWNS
Success Rate 14.3% 25.0%
PPP 0.10 0.00
S&P 0.239 0.248
TURNOVERS
Number 0 8!
Turnover Pts 0.0 34.1!
Turnover Pts Margin
+34.1 -34.1
Q1 S&P 0.221 0.759
Q2 S&P 0.756 0.464
Q3 S&P 0.560 0.754
Q4 S&P 0.222 -0.050
1st Down S&P 0.542 0.548
2nd Down S&P 0.340 0.488
3rd Down S&P 0.357 0.541
Projected Pt. Margin
+38.9 -38.9
Actual Pt. Margin
+2 -2

This game was strange in too many ways to fathom.  Nebraska turned the ball over four times inside Iowa State's five and were outgained overall in terms of EqPts, and their projected margin of defeat was almost 40 points.  And they lost by two.  If they had managed just seven turnovers and a Turnover Points Margin in the low-30s, they very well might have won the game.  Through three quarters, Nebraska had managed 327 yards, and the only thing preventing them from blowing the Cyclones out were their five turnovers.  But when it was time to suck it up and make a play in the fourth quarter, Nebraska imploded.  Their last four drives produced this:

Rushing: 3 carries, -4 yards
Passing (including sacks): 3-for-10, 33 yards, 2 interceptions, 1 fumble, six consecutive incompletions/interceptions to end the game

Unlike the Virginia Tech loss (a last-second loss on the road to a good team) and the Texas Tech loss (a bit of a hangover loss in which Tech got every conceivable lucky bounce), this one might sting Nebraska for a while.  Fans often say "We handed them the game," and usually it is an exaggeration.  This time, Nebraska almost literally handed Iowa State the game.

Because of this and Kansas' loss to Colorado, the idea of upper (Nebraska, Kansas, Missouri) and lower (Kansas State, Iowa State, Colorado) tiers in the Big 12 North has been blown out of the water.  However, in both of the North's big upsets, turnovers played huge roles.  Kansas handed Colorado 14 points on turnovers recovered or returned inside KU's 5; meanwhile, Nebraska took at least 14-21 points off of their own scoreboard with their turnovers deep in ISU territory.  It took a lot for these two teams to lose these games, but lose them they did, and they have both given up control of the division, at least for a small while.  The odds are still in favor of one of those three teams winning the North, but odds haven't meant a whole lot so far.

S&P+ Top 25

Tada! The moment we've waited for is finally here. Up-to-date S&P+ rankings can now be found on Football Outsiders each week. You can see the ratings for all 120 FBS teams as well as offensive and defensive ratings, with splits for rushing vs. passing and Standard Downs vs. Passing Downs.

When you click on that link, you will see that, indeed, two teams with losing records are in the Top 10. It is no coincidence that those are also the two teams who have played both Alabama and Florida. Their ratings will fall in the next month, simply by virtue of Tennessee playing Memphis, Vanderbilt, and Kentucky, and Arkansas playing Eastern Michigan, Troy, and Mississippi State; however, this does point to an interesting conundrum in terms of week-to-week calculations for S&P+.

As I mentioned last week, I have been relatively pleased with the final calculations for the past two years. However, the schedule strength adjustments do seem to cause rather interesting headaches in the middle of a season. You see plenty of teams ranked here simply because of the teams they have played, and thanks to the trickle-down effect -- Alabama and Florida are the top two teams, Arkansas and Tennessee have played both of them, and others have played Arkansas and Tennessee -- you have seven SEC teams in the S&P+ Top 25 and ten in the Top 35 (South Carolina is #29, Georgia #32, and Kentucky #35). Plenty of people would see nothing wrong with that -- it is, after all, the top conference this year; however, some of the rankings are still a bit extreme and should fall over the next month.

Before complaining too much about the rankings of Tennessee and Arkansas, however, ask yourself this: what record would the teams around Tennessee and Arkansas be if they played the same schedule? Certainly UT's home losses to UCLA and Auburn were questionable, but they did also kill Georgia and hang tough against Florida and (especially) Alabama. Would Miami or TCU have done the same? Maybe, maybe not. Meanwhile, Arkansas' losses have come at Alabama, at Florida (by just three points), at Mississippi and at home against Georgia. Maybe Oklahoma or Miami would be 4-3 with that schedule instead of 3-4, but probably no better.

Speaking of Oklahoma, yes, they are ranked fourth despite three losses. They have gotten there in much the same way as Arkansas and Tennessee -- with a road loss to S&P+ No. 6 Miami (by one point) and neutral-field losses to No. 9 Texas (by three points) and No. 19 BYU (by one point). They too are probably ranked a bit too high, but only a bit.

Biggest S&P+ Movers of the Week

After a relatively tame week of rises and falls, we had three teams make nice surges this week.

Strongest Rises

Mississippi State (26 spots, 102nd to 76th). Granted, the only offense they generated was through pick sixes, but the Bulldogs defense played its best game of the year (not surprising, since their head coach used to be Tim Tebow's offensive coordinator), and they moved up a lovely 26 spots thanks to both performance and strength-of-schedule adjustment.

Air Force (24 spots, from 91st to 67th). Play-by-play numbers do not seem to care much for grind-it-out running teams like Air Force and Georgia Tech (which we will get to below).  The Falcons were ranked in the nation's bottom quartile until last week's respectable showing against Utah.  They have now lost by a total of just ten points to TCU and Utah.

Iowa State (20 spots, from 83rd to 63rd). Nebraska's ranking did not suffer too much due to the loss to Iowa State (turnovers are taken into account in the ratings, but not too much, at least not yet)

Pittsburgh (13 spots, from 74th to 61st) and West Virginia (13 spots, from 64th to 51st). The S&P+ is clearly unimpressed with the Big East overall and has been slow to come around on both of these one-loss teams. But after Pittsburgh demolished South Florida and West Virginia knocked off S&P+ darling Connecticut, the rankings responded a bit.  They are probably both ranked too low, but we will see what happens when these two teams and Cincinnati start beating each other up.

Other notable rises: Ole Miss (36th to 24th), Texas A&M (54th to 42nd), Bowling Green (82nd to 70th), UTEP (89th to 77th).

Largest Falls

Washington State (15 spots, from 99th to 114th). The Cougars have been outscored 155-43 over the last four weeks, and any sign of slight improvement from the season's early games has begun to subside. There is at least a little bit of hope for the future in the form of freshman quarterback Jeff Tuel, but not in 2009.

Texas Tech (11 spots, from 41st to 52nd). Figure out Texas Tech and Texas A&M, we dare you. The transitive property suffered an epic breakdown this week, as A&M (who lost to Kansas State by 48) easily handled Tech (who beat Kansas State by 52) in Lubbock.

Georgia (10 spots, from 22nd to 32nd). The Bulldogs handled Vanderbilt easily, but it appears their ranking suffered just for playing the Commodores in the first place.

N.C. State (9 spots, from 78th to 87th). The Wolfpack got a week off after giving up 101 points combined to Boston College and Duke the previous two weeks, but their ranking fell as they apparently allowed 450 yards and 40 points to Bye.

Other notable falls: Ohio (84th to 101st), Akron (85th to 102nd), Northern Illinois (52nd to 65th), Toledo (70th to 82nd), UL-Monroe (65th to 75th), Southern Miss (62nd to 72nd).

F/+ Top 25

Featured inside this year's Football Outsiders Almanac was a combination FEI and S&P+ rankings called F/+. Now that S&P+ is updated weekly on the site, it might be a good idea to get this one going too. Here are updated F/+ ratings after eight weeks.

F/+ Top 25 (After Eight Weeks)
F/+ Rk Team Record FEI Rk S&P+ Rk Off F/+
Rk
Def F/+
Rk
1 Alabama 8-0 1 1 14 2
2 Florida 7-0 7 2 18 4
3 Miami-FL 5-2 6 6 2 17
4 TCU 7-0 3 8 8 14
5 Iowa 8-0 2 13 31 3
6 Virginia Tech 5-2 5 15 5 16
7 Oklahoma 4-3 14 4 45 1
8 Penn State 7-1 17 3 16 7
9 Boise State 7-0 10 12 36 10
10 Texas 7-0 12 9 6 25
F/+ Rk Team Record FEI Rk S&P+ Rk Off F/+
Rk
Def F/+
Rk
11 Oregon 6-1 11 17 35 8
12 Clemson 4-3 13 18 52 5
13 Cincinnati 7-0 9 20 7 28
14 LSU 6-1 16 14 47 12
15 Ohio State 6-2 19 16 48 9
16 Georgia Tech 7-1 4 43 1 55
17 USC 6-1 8 26 17 18
18 Wisconsin 5-2 27 10 34 15
19 Tennessee 3-4 37 7 20 21
20 Arkansas 3-4 49 5 11 45
F/+ Rk Team Record FEI Rk S&P+ Rk Off F/+
Rk
Def F/+
Rk
21 Nebraska 4-3 30 11 86 6
22 Oregon State 4-3 22 28 21 61
23 Arizona 5-2 15 50 37 32
24 BYU 6-2 55 19 9 54
25 Ole Miss 5-2 32 24 62 13

One major offseason project will be to take a look at some of the most notable disagreements between FEI and S&P+ (Georgia Tech, Arkansas, Arizona, and BYU, for instance) and drill into how there can be such a large difference between ratings based on possession data and play-by-play data.

For now, you've got an interesting set of ratings that combines each system's strengths and weaknesses. The case made above for how teams ranked around Tennessee and Arkansas would have similar records given their schedule? It's even easier to make that case when the teams are ranked 19th and 20th instead of fifth and seventh. Do you really see Wisconsin or Oregon State going better than 3-4 or 4-3 with Tennessee's or Arkansas' schedule?

Random Golf Clap

Words cannot express the moving scene that preceded last week's West Virginia-Connecticut game. All of Bill Stewart's strengths as a human being were on display in his lengthy hug of Randy Edsall before the game, and the West Virginia crowd reacted to Connecticut's emergence from the tunnel in exactly the way one would hope that every fanbase in the country would react. In Morgantown on Saturday afternoon, you saw exactly why football is both a) only a game, and b) something that can be used to bring people together.

Meanwhile, a somewhat belated golf clap for Oregon, who has demonstrably turned their seasons around after the worst possible start. Seriously, the only way Oregon's season could have been any worse after one game was if LeGarrette Blount actually did make it into the stands after the game, and if the entire defensive line contracted polio. One week into his head coaching career, people were already wondering if Chip Kelly had lost his team and was doomed. One week!  And now he's looking like the favorite for Pac-10 Coach of the Year.

(Of course, we should maybe hold off on the applause until after the USC game.)

Random Mini-Rant

A whopping three mini-rants this week!

1. Why Neuter the Computer Rankings?

A lot of people were shocked that Iowa got such a boost from the computer portion of the BCS formula, and while others have already gone into detail about why they were ranked so high, it brings to mind a rant I am almost always more than willing to make: if you are going to use computer rankings as part of your formula, let the computer rankings do what they do. When the BCS higher-ups (whoever those nameless people actually are) took scoring margin out of the equation, they completely and totally neutered their effectiveness.  Not every computer rankings system takes scoring margin into account, but those that do (for instance, Jeff Sagarin's system) had to create a slightly different rankings system that removed margin from the equation.

This is insane for a number of obvious reasons. First of all, scoring margin matters. If you want to put a cap on a given game's point differential -- say, for instance, that any victory of over 24 points counts as a 24-point victory, whether it was 24 points or 66 points -- then that's fine. Oklahoma high schools have been doing that for years. One could envision a scenario where a highly-ranked team beating an opponent by only 21 points would keep their starters in for one more drive to try to get to a full, 24-point victory; but that is still a more acceptable scenario than running up a score to 50, or even 60-70 points. But Iowa is getting credit for simply beating Arkansas State and Northern Iowa, while not at all getting penalized for only beating them by a total of four points. Again, scoring margin, how badly Team A actually beat Team B -- tell a piece of the story that the BCS ratings refuse to tell. Would Iowa be as highly-ranked if their 8.9-point victory margin was taken into account?  Of course not. Everything in this world is a shade of grey, and limiting the discussion to black-and-white wins and losses is simply not a great idea.

Second of all, there are hundreds of ways to create a computer rankings system (including, ahem, those you find at Football Outsiders). If you don't want Jeff Sagarin using point differential, then don't use Jeff Sagarin's system. Use only systems that use nothing but wins and losses if that's what you so choose. But don't neuter the ones you have. You chose them because, in theory, they were the best. Use them in the way they were intended, or use others that fit your qualifications.

(Note: I like Sagarin's system quite a bit, so do not in any way construe this as anti-Sagarin. I'm just making a point.)

2. Why have award watch lists and semifinalists anymore?

There was a decent amount of justifiable outrage this week when the semifinalists for the Davey O'Brien award were announced, and Florida State's Christian Ponder, among others, did not make the cut, while less stat-friendly quarterbacks like Iowa's Ricky Stanzi did. The Ponder supporters have an excellent case, but here's the deal: this is 2009. Potential voters can keep up with all 120 FBS quarterbacks pretty easily and effectively nowadays. Why in the world do we still need semifinalists? Would anybody complain if they just skipped this step (and the preseason watch list, which is an even more incredibly insane concept) and announced finalists later? Would anybody think their guy didn't get enough exposure because he didn't make the semifinals?

3. Why not have cameras directly on the goal lines at all times?

In all of the hubbub over the latest controversial call in the SEC, Dustin Doe's iffy pick six in which he may or may not have fumbled before crossing the plain of the goal line, one very important point wasn't made: if we had cameras permanently eying said goal line plain, we would know for sure when fumbles happened or if teams scored. They end up in the right place in goal-to-go situations, but not in unexpected plays like Doe's interception.

Random Reasons to Love College Football

Signs, rocks, horns, gators, light switches, and all the other college touch traditions.

Also: "There are no Brett Favres in college."

And Mike Leach.

Random Playlist

In honor of the officials and refs at the heart of seemingly every storyline this past month:

Angry People, Barenaked Ladies
Blind, Company Flow
Blinded by You, The Starlight Mints
Eyesight to the Blind (The Hawker), The Who
I Got Stripes, Johnny Cash
Imagination Blind, Dinosaur Jr.
The Official, J Dilla and Madlib
Whistles in the Wind, Flogging Molly
You Turn the Screws, Cake
Your Cheatin' Heart, Hank Williams

Upset Watch

Pulled off an 0-fer last week in the Upset Watch, which was probably as much regression to mean as anything.  In four weeks of Upset Watch, we're now sitting at 7-9 straight-up, which still isn't bad, but it looks disappointing after the 6-2 start.

Auburn over Ole MissSpread: Auburn +3.5 | Projection: Auburn by 8.1

Recent members of the Auburn bandwagon are jumping off so fast that they are skinning their knees on the fall, but without a momentum factor, Auburn still looks to match up pretty well against an Ole Miss team that has only recently begun to look like a solid squad. How much does momentum matter? Consider that another off-season research project, but for now all we know is that Texas A&M beat Texas Tech handily last week despite the teams' momentum suggesting an easy win for Tech. That's a sample size of one game, of course, but do not overlook the Tigers simply because they have struggled. Momentum turns quickly.

Colorado State over Air Force. Spread: CSU +3.5 | Projection: CSU by 11.9

Speaking of momentum, CSU has looked positively dreadful in recent weeks, losing five in a row by mostly increasing margins.  A 2-point loss to Idaho led to a respectable 7-point loss to Utah.  Then came a 44-6 debacle against TCU and, the biggest kicker of all, a 42-28 home loss to San Diego State.  Now CSU faces an Air Force team that has seemingly improved through losing.  The Falcons have lost three of four, but the losses were to Navy, No. 10 TCU and No. 18 Utah, and all by less than a touchdown. The numbers like CSU in this one ... for some reason.

Minnesota over Michigan State. Spread: Gophers +3.5 | Projection: Gophers by 10

Two more desperate teams meet Saturday night in Minneapolis. (Really? A night game in late-October? In MInnesota's new outdoor stadium? Needless to say, the weather doesn't look promising.) These projections, of course, do not take into account the loss for the season of star Minnesota receiver Eric Decker to a sprained foot, but even with an injury adjustment, they would still be projected to win. At 4-4 with three home games remaining (including two against Illinois and South Dakota State), a bowl still looks likely for the Golden Gophers, but after back-to-back blowout losses to ranked teams (Penn State, Ohio State), they need to build some positive momentum. Meanwhile, Michigan State looked to have turned their season around with about a minute left in last week's Iowa game. After a 1-3 start, they had won three straight conference games to move to 4-3 and had taken a late lead on Iowa. Then Iowa scored on the last play of the game, keeping their BCS title hopes alive and sending the Spartans reeling.

Oregon over USCSpread: Oregon +3.5 | Projection: Oregon by 11.7

The rankings for this matchup are almost the direct opposite of what you would expect. Currently the S&P+ rankings have Oregon at 39th on offense but 9th on defense. Meanwhile, USC is a staggering 55th on defense but 17th on offense. Oregon's season seemed on the verge of completely falling apart after just one game, but the Ducks have slowly turned things around, and a win on Saturday would give them very good odds of ending USC's Pac-10 conference title streak.

Closing Thoughts

We have been tinkering with Varsity Numbers for the past few weeks, looking for an enjoyable format that highlights both stats and love of college football. Is there anything in particular that you would like to see beyond or instead of other Varsity Numbers features? This has turned into a pretty long column, but it is enjoyable to write, and it needs to be enjoyable to read as well. Feel free to share in the comments or my e-mail using the FO contact form.

Posted by: Bill Connelly on 30 Oct 2009

17 comments, Last at 02 Nov 2009, 3:31pm by Pat (filler)

Comments

1
by Will :: Fri, 10/30/2009 - 6:53pm

"all just a lead-up to the week's biggest games, Ohio State-Texas (projection: Texas by 5.9) and USC-Oregon (projection below)."

You got the wrong OSU.

Will

2
by Lance :: Fri, 10/30/2009 - 6:57pm

As A Stillwater native and fan of the school (even if I didn't go there-- it's just part of your blood), this is the most frustrating thing in college sports. Yes, there is MORE THAN ONE OSU.

3
by Will :: Fri, 10/30/2009 - 7:03pm

Many have come to abbreviate the Buckeyes as "tOSU" (i.e. "The" Ohio State University), leaving Oklahoma State and Oregon State to battle to the death over "OSU".

Will

6
by Bill Connelly :: Fri, 10/30/2009 - 8:36pm

Being that I grew up in Oklahoma (Weatherford) and have been to quite a few games at Lewis Field, you can consider this one an honest error and not part of a national conspiracy.

7
by Aaron Schatz :: Fri, 10/30/2009 - 9:18pm

In fact, it's my fault. Bill put "OSU," and I changed it because it's not a school we generally use an acronym for. (Either Ohio OR Oklahoma State, actually.)

4
by KJ@theonlycolors (not verified) :: Fri, 10/30/2009 - 7:33pm

I apologize if I've missed this somewhere, but I'd love to see a detailed explanation of the differences between the various FO college football ratings: FEI, game efficiency, S&P. Hard to figure out what exactly what each is measuring intuitively.

5
by knappster :: Fri, 10/30/2009 - 7:39pm

Also:

"and thanks to the trickle-down effect -- Arkansas and Florida are the top two teams, Arkansas and Tennessee have played both of them"

I believe that should be "Alabama and Florida".

8
by Kevin 11 (not verified) :: Sat, 10/31/2009 - 2:15am

Bill...or whomever is responsible for this.

4-3 Oklahoma is ahead of 7- 0 Texas?

After Texas beat Oklahoma?

Is that evidence in itself that YOUR SYSTEM DOESN'T WORK AND NEEDS TO BE FIXED?

9
by Bill Connelly :: Sat, 10/31/2009 - 9:54am

Okay, one more time: this ratings system counts every (close) play of every game. The fact that Texas beat OU and killed Missouri matters the same as the fact that they also let Colorado and Wyoming stick around 2-3 quarters too long. As I mentioned, OU is ranked (too) high because they have played a series of good teams (Texas, Miami, BYU) statistically even. Simply by playing K-State, Texas A&M, etc., their rankings will fall over the next month.

The whole season matters, not just one game. Should UTEP be ranked over Houston because they beat them? N.C. State over Pitt? Texas A&M over Texas Tech? Texas Tech over Kansas State? Kansas State over Texas A&M?

And again, the schedule strength adjustment seems to over-adjust with less data. This really young ratings system has shown that at the end of the year, the rankings it produces will be strong and logical, but until then there is simply not enough data. And as I mentioned last week, consider them a suggestion instead of a hard-core "Team A is ranked 3rd and is therefore demonstrably better than Team B, ranked 4th." This suggests that OU is probably better than we're giving them credit for, the end.

10
by zlionsfan :: Sat, 10/31/2009 - 1:43pm

I think you mean plane and not plain.

Amen on the computer rankings. One of the many complaints I have about the BCS is that when games do not play out as the BCS people expect, it is blamed on the objective part of the ranking system, and so they attempt to change the entire system to adjust how a few teams are ranked, much as people do here (pointing out one or two teams as "proof" that the system doesn't work).

But no one ever mentions the elephant in the room: is there a single person anywhere in college football who can accurately rank the best 25 teams in the game in less than one day after games have been played? It's not the polls, it's never the polls, the polls are always right, nothing's wrong with the polls.

The polls were never meant to be used for this purpose. They're fine as a talking point in the same way that subjective systems are fine for that purpose. They are terrible as a means of determining who gets to play in the final game.

RPI isn't nearly as much of an issue, because it really only comes into play for choosing between, say, the 40th- and 41st-best teams in college basketball. There's never really a chance for the best team to be excluded from the tournament. With the current system in college football, though, the subjective parts of the system can exclude the best team from the final game, and it's the one part that nobody really seems to want to change.

11
by dvdburns@yahoo.com :: Sat, 10/31/2009 - 2:32pm

My favorite upset specialist strikes again. Congratulations on your fantastic success this year. I hope someone out there is profiting from your wisdom. The rankings may get a little hokey from time to time, but the upset specials have to make you happy.

12
by Alexander :: Sat, 10/31/2009 - 4:48pm

I could not have said it better myself about the computer rankings Bill. The BCS cannot have it both ways: Say we don't want PD in the system which encourages people to run up the score and then turn around and say, "THE COMPUTERS ARE WRONG1111!" The computers are never wrong, the people who program them might be wrong, but the computers are not.

The capping idea makes perfect sense(I thought about a 30 pt cap the other day) and makes very good sense.

13
by Pat (filler) (not verified) :: Sun, 11/01/2009 - 2:42am

Oh, Bill, now you've done gone and annoyed me.

Not every computer rankings system takes scoring margin into account, but those that do (for instance, Jeff Sagarin's system) had to create a slightly different rankings system that removed margin from the equation.

Sagarin's "non-margin" ranking system is just an Elo system. He didn't have to "create" anything. Elo's been around for decades.

This is insane for a number of obvious reasons.

No, it's not. It absolutely, 100% is not insane. Why? Because there is no unbiasable state function for football other than winning. Period. The end. This means game over for anything other than pure wins/losses.

Why?

1) Because if you base the statistical rankings on anything other than pure wins/losses (and situational non-game related factors, like game location), you change the game itself. The point of football is to win the game, and by extension, to win the national championship. If you change the way to get to the national championship to not "winning", but something else (scoring margin, highest DVOA, etc.) then the point of the game is no longer to win. It's to score the most points.

And that is not football. See, for instance, Brian Westbrook kneeling at the 1 yard line. Smart football. Not smart for scoring margin. "Wait," you say, "that doesn't happen that often," right? Wrong - if it happens once in a team's season, that's 8% of their total season. It may not affect many teams, but when it does affect a team, it affects them a lot.

2) Because winning a game kinda needs to mean something. Ideally, if a game goes to 3OT, and one team wins, an idealized "game output" would say "this game was a tie" because, well, it essentially was - through regulation, it was, at least. But one team won! That's the point of having overtime - because we want winning to mean something.

3) Finally, from a statistical point of view, using margin of victory is stupid. Margin of victory is useless with anything less than a margin of ~14 points - the difference between a 10 point victory and a 3 point victory is statistically nothing. In fact, the 3 point victory actually might be more solid than a 10 point victory if the team with the 3 point victory closes out the game on its own by kneeling, whereas the team with the 10 point victory scored the TD and handed the ball back to the other team, thus giving them an (unlikely) chance to win.

But you also want to have a "points cap" as well, to eliminate the biasability of the metric (which you can't do, as stated above, but let's ignore this). Which means in actuality, all you really care about is "greater than 14 point victory" and "less than 14 point victory". Great. Now you've changed the entire point of the game for one (literal) bit of information. So now instead of just "W/L", you get one more binary variable. Guess what? This won't improve the accuracy much at all.

And you know what? The BCS knows all of this. Because statisticians are the ones who pulled for removing MOV from the rating systems because it simply wasn't worth the tiny gains you got. It improves the predictability of the rankings, but does so by biasing the final result.

(As a quick aside: it's easy to see why using MOV is stupid because football is a paced game. Degree of victory changes depending on the length of the game.)

15
by Bill Connelly :: Sun, 11/01/2009 - 10:20am

Even though I disagree with a majority of this comment (obviously), I really liked it--you made a lot of good points, but let me simply ask you this: in a system in which only the top two teams get a chance at the national title, how exactly do you suggest differentiating between seven undefeated teams? Florida, Alabama, Iowa, Texas, Cincinnati, TCU and Boise State have all done everything asked of them...so do we just give them ALL a spot in the title game? We can't do that. We have to differentiate between them somehow (and really, we'd have to do the same thing if there were a 4-, 8-, or 16-team playoff as well). With 120 teams playing different schedule, simply looking at who wins games and who loses them simply does not differentiate between teams enough.

Iowa is living a blessed life right now, and it's a really neat thing to watch, but how exactly can you pick them above Cincy or TCU right now when they've trailed in all but one game this year and have trailed in the fourth quarter of, what, four games now? Cincy, TCU, Boise, etc...all of these teams "just win games" too, and they do it in a much more impressive fashion. None of those teams would have trailed Indiana and Michigan State in Q4, beaten Northern Iowa by 1, and beaten Arkansas State by 3. They may have done ONE of those things, then won in blowouts in the other games. (Here's where you can absolutely make the case that those teams also wouldn't have beaten Penn State in PA, and you would have a very good point.) If Iowa is one of two undefeated teams at the end of the season, they absolutely deserve a shot at the title. But if there are more than two, then it's hard to make a case that they're one of the top two.

16
by xLittleP :: Mon, 11/02/2009 - 2:14am

First, I want to say that the Elo algorithm, invented to rank chessmasters, is an excellent way to rate two opponents. (Not to get booed off the stage here, but) I prefer to use the World Soccer Elo Ratings over the ridiculous FIFA ratings and even UEFA "coefficients". However, that rating method requires a much larger sample size to be effective. Given the ever-so-short length of the college football season, like the S&P+, it's merely effective if not entirely accurate, until the end of the season.

Back to the subject at hand, yes, the Iowas need to be separated from the Floridas somehow, and if not on the football field, than in some subjective manner. But how do you take away from Iowa a close victory over an inferior team? Aren't a never-say-die attitude and an ability to come from behind valuable measures of a football team's greatness? Isn't this why we include feeble humans in the BCS formula?

Sure, looking at who wins and who loses isn't enough, which is why we have to factor in strength of schedule. I believe that's what the majority of the BCS Computers are trying to perfect, the SOS calculation. If you look at the Massey Ratings, of the sites that actually divulge their methodology, you see some very wild metrics to determine SOS (one takes the "RPI" method to an infinite level; the Elo method uses it inherently). Since scoring margin is useful only for predicting future outcomes, and strict wins and losses is useful only for rating a team's prior success, I think it's right that the BCS eschews margin of victory.

No system is going to be perfect, which is why the BCS is the marriage of three systems, and in my mind all of those complement the others strengths and weaknesses well. Short of completely restructuring College Football as we know it, as much as it hurts to say it, I think the BCS is as good as we're going to get.

17
by Pat (filler) (not verified) :: Mon, 11/02/2009 - 3:31pm

Bill, there's an important point to note: if you use scoring margin, what you end up tending to do is 1) rank teams by their easiest games.

The easiest way to see this is to realize that what using MOV does is essentially replace close games with a tie, or much closer to a tie than a normal game. PFR, in their blog, had a mathematical example of how most statistical rankings work, and what MOV does is replace the full "1/0" weight for a win with a reduced value (down to "0.5/0.5" in the case of a very-near tie).

in a system in which only the top two teams get a chance at the national title, how exactly do you suggest differentiating between seven undefeated teams?

The statistical rankings use Bayesian inference, which is a fair thing to do.

Determining how good a victory was has to be done by humans. Humans invented the game, and have an idea of what the game is supposed to be (hence the reason I really support the idea of a coaches' poll. There certainly needs to be better oversight, but fundamentally the idea is fantastic). If you use anything other than a system based purely on wins/losses, you're ranking teams based on an idealized version of football which is not the game they're actually playing.

Margin-of-victory works on average for predicting future performance. The problem is that that cannot be what the statistical rankings are supposed to do - they have to be retrodictive evaluations of the games. And margin-of-victory does not work for that. It flat-out doesn't: nearly every week there is a team that sacrifices margin-of-victory for an improved likelihood to actually win the game.

Last week had a great example of a mistake where a team minimized their margin-of-loss and nearly guaranteed their loss probability.

UNC vs Va Tech: UNC recovered a fumble basically in field goal range, and promptly ran the ball as many times as they could, and kicked a 21 yard field goal - closer than an extra point.

Virginia Tech shouldn't've let that happen. They should've simply let UNC run in for a TD, quite possibly on the first play, definitely after they gave up the first down. That would've given them the ball back with a chance to tie the game. They might not have won, and they would've lost 24-17, but that game could've been closer than the 20-17 loss because Va Tech would've had a chance to tie the game. They also might've tied the game.

It may seem like I'm finding strawman-type examples, but that's the point: a statistical ranking has to be perfectly fair. Right now, it is (if you exclude the insane rankings, like Billingsley).

I know it sounds terrible to say: "the polls should decide" because everyone thinks that the polls are horribly biased. The solution is to fix the polls, not try to replace them with a system that's inherently biased.

Incidentally, the whole "the polls are horribly biased!" thing? Completely overrated. There've been studies on bias in the polls, and the bias goes almost exactly the opposite way that you would expect. There's no evidence for preseason bias. There is, in fact, a bias for small schools. Yes, that's right: the biggest bias in the polls? For the MAC.

Scary as it seems, the polls are, for the most part, very good. Let them decide the difficult questions, and let the statistical rankings correct for the minor biases present in the polls. That's the fairest method.

Iowa is living a blessed life right now, and it's a really neat thing to watch, but how exactly can you pick them above Cincy or TCU right now when they've trailed in all but one game this year and have trailed in the fourth quarter of, what, four games now?

You realize that Iowa just won a game 42-24, and you're including it in a narrow victory ("trailed Indiana in Q4"). Which a MOV-system wouldn't know. You're helping my point, here.

How can you justify ranking Iowa above Cincy or TCU? Because they won games against better teams. If you think the fact that they won those games narrowly means something, that's for the human voters to decide.

14
by TV_Pete (not verified) :: Sun, 11/01/2009 - 10:04am

Great points!

I would love to see goal-line cameras always there (how much does it cost to have them lined up from each side at all times during plays?).

I completely agree about the Computer rankings being neutered. Let them do there thing and they are worth considering.

That being said, I think a subjective view still has something to offer as a piece of the puzzle.