Writers of Pro Football Prospectus 2008

08 Jan 2009

Bill James Takes On The BCS

Naturally, his take is pretty good. His main problems with the system:

1. That there is a profound lack of conceptual clarity about the goals of the method;

2. That there is no genuine interest here in using statistical analysis to figure out how the teams compare with one another. The real purpose is to create some gobbledygook math to endorse the coaches' and sportswriters' vote;

3. That the ground rules of the calculations are irrational and prevent the statisticians from making any meaningful contribution; and

4. That the existence of this system has the purpose of justifying a few rich conferences in hijacking the search for a national title, avoiding a postseason tournament that would be preferred by the overwhelming majority of fans.

Posted by: Bill Connelly on 08 Jan 2009

74 comments, Last at 13 Jan 2009, 7:48am by Sancho


by ChrisH :: Thu, 01/08/2009 - 8:25am

The one point that's driven me crazy over the years is #2 and 3, as it is always the outcry that "The computer must be wrong!" if they disagreed with popular opinion, not that the coaches or voters must be wrong, or have not looked at something in the same way the computers did. When USC was excluded despite being #1 in both polls, everyone got mad at the computers and reduced their importance without considering that, perhaps, they had screwed the computers earlier by forcing them to remove all margin of victory so that USC didn't look as good to the computers anymore. Looking back at those standings now, it looks like the major change after that year was also to remove the SOS as a separate component, and remove the quality win bonus.

I hadn't really considered point #1 much before, but that is a very good point that the polls don't account for. When we come here, we know what we're after when we look at weighted DVOA (the teams that are playing best now and therefore most likely to win), but we have no idea what the BCS Computers are trying to do. I don't know what I want the computers to rank, but I guess I want to know the teams that are most likely to win now, and I'm sure if that was the goal you gave to the statisticians, they would use margin of victory, strength of schedule, and record to determine that.

I wish I had enough interest in baseball to read more Bill James. I should probably just pick up a book and give it a read, I'm sure it would be interesting even if I don't care that much about baseball. Can anyone chime in on that point?

by Barrie Jones (not verified) :: Thu, 01/08/2009 - 8:40am

Just try Moneyball by Michael Lewis. He explains a lot of Bill James' ideas in it, and puts it in perspective of modern baseball practice. I don't have a lot of interest in baseball, but, I couldn't put the book down...

by Bill Connelly :: Thu, 01/08/2009 - 10:13am

I've always loved college and pro football more than baseball, but James' stats have far increased my interest in baseball. And most importantly, a lot of the concepts he uses can be applied to football as well. A lot of the (more and less successful) concepts I've used in Varsity Numbers have come from baseball stats/concepts...most of which were kick-started by James. His writing is a great starting point for looking at just about any sport from a statistical standpoint...

In other words, yeah, you should read what Bill James has to say whether you like baseball or not. Opens up a whole new world.

by zlionsfan :: Thu, 01/08/2009 - 10:44am

If you can find one, I'd recommend picking up a Baseball Abstract ... his last one was published in '88, so those may be hard to find (and no, I can't lend you my copy :) ). However, he's done some historical stuff recently, like his Historical Abstract (available on Amazon, I suspect if you start by clicking one of the links on the side and then find it, FO will get their finder's fee or whatever?). That might be a bit harder to digest, but it's still worth it IMHO.

One nice thing about baseball is that the application of statistics is so much further along than with football. Whether you find an older book or a newer one, you're likely to find some good reading material and maybe a look at what the FO folks (among others) are hoping to do someday.

Also, I completely agree with James, and I'm not watching the game tonight. I remember back in the day when Bill Walsh College Football (now known simply as NCAA) for the Genesis had the option to finish your season with either bowls or a playoff system.

IIRC, in either the next season or the one after that, the playoff option was removed. It hasn't been back.

by JimZipCode (not verified) :: Sun, 01/11/2009 - 3:46am

>> I wish I had enough interest in baseball to read more Bill James. I should
>> probably just pick up a book and give it a read, I'm sure it would be interesting
>> even if I don't care that much about baseball. Can anyone chime in on that point?

Bill James is worth reading for the clarity he brings to thinking about sports. Also for the pure acid of his humor. I'd suggest his book, "This Time Let's Not Eat The Bones". It's a collection of some of his best writing from the baseball abstracts, which are what made his reputation. That book purposely excludes statistical studies (the "bones"), so as to make his stuff more accessible. That's probably not a selling point for a fan of this site: yet it's still a good book.

He's the granddaddy of a certain way of thinking about sports. The writers here, and at CFF, and a ton of other places, owe a huge debt to Bill James. His influence permeates the sports-blogging atmosphere. He's really worth reading.

Some follow-on recommendations, if you enjoy that book, would be His "Whatever Happened To The Hall of Fame?" (aka The Politics Of Glory), and maybe his Historical Baseball Abstract. Somewhere in that book he displays his gift for asking the clarifying question, by noting that when people try to rank the best players of all time, they don't specify whether they are thinking of peak value or career value. And this is at the root of a lot of arguing. It was a stunning observation: so clear and obvious AFTER he articulated it.

James wrote a biographical profile of (I think it was) Ernie Lombardi, for the 2nd ed of the Historical Abstract, that was absolutely heartrending. And I think this illustrates a huge aspect of his work, that people overlook now. He wasn't a "statistician": he was a writer and historian and analyst. He looked at numbers because baseball is described in numbers; but he wasn't bound by them.

Great writer. By the way, Michael Lewis' Moneyball, while an excellent book, does a pretty crummy job of explaining Bill James' ideas.

by Karen R :: Thu, 01/08/2009 - 9:21am

I think the answer to # 1 is pretty clearly answered in #3. The problem is that it defeats the purpose of having "computer" rankings.

by Theo :: Thu, 01/08/2009 - 10:18am

If everyone would call it what it is, the computer programmers wouldn't have as much influence in the rankings.

by Dennis :: Thu, 01/08/2009 - 10:25am

What are you talking about? The computers are only 1/3 of the rankngs now. When the BCS started, objective data - computer rankings, strength of schedule, quality wins, etc. - was 75% or so of the rankings. Every adjustment has been to reduce the weight of objective data and increase the weight of the polls.

by Dennis :: Thu, 01/08/2009 - 10:24am

ChrisH has it exactly right, with James' point #2. When the computers (objective rankings) were different than the polls (subjective rankings) everyone immediately claimed that the computers were wrong so they adjusted the formula. Every change to the BCS rankings has been to remove or reduce the weight of objective data and increase the weight of the polls.

Where I disagree is that I want the computers (and the humans for that matter) to rank the teams based on what they have accomplished during the season, not by who they think would win if they played today. For example, IMO Utah has accomplished more than USC this season, but I think USC would beat Utah if they played today. So Utah should be ranked higher than USC.

by Dean (not verified) :: Thu, 01/08/2009 - 10:53am

Bill James is an absolutely brilliant man. However, he is wrong here. The bottom line is - no matter how often you hear otherwise - the BCS WORKS. What the BCS has is an image problem, and probably an unfixable one at that. Most people are simply too stupid to understand it.

If you have 2 undefeated teams, you don't need the BCS computers to tell you which two teams to pick. And when the computers spit them out, people think "wow, it worked this year." But most years, there AREN'T two undefeated teams. There might be 3. There might be 1, or 0, or even more then 3. So how do you pick? We already know that people all have their biases. Teams they like, teams they don't like - those are obvious. But there's also games they saw and games they didn't see. So how can we possibly expect humans, with all our flaws, to pick the right two? We can't. But what we can do is turn it over to the machines. The computers don't have any biases. They take a genuinely objective, statistics-based analysis and determine who the top 2 are. And no matter what, there will ALWAYS be people who disagree - but that's because of the human biases, not because of flawed computers.

There IS one flaw in the computer. It's only as good as its human programmer(s). Garbage-in, garbage-out. So James does have a valid point that the computers are - in the real world - there to serve the good old boys. But that is not the fault of the computers. It's the fault of the programmers, and, in turn, the fault of the good-old boys who insist on it. And they have the power to insist on it, as they also have the power to turn off the computers and go home. So statisticians, like Bill James, want to take their ball and go home. Really, what he wants, is more power. The power to let statisticians make better computer programs - and I support his agenda even if I disagree with his tactics. You can't affect change if you're not in the room.

The fact of the matter is that The User - i.e. the good old boys - gives the programmer a set of spec, and the programmer in turn gives out a program which generates rankings. And those rankings are the most accurate in the land, without passion and prejudice. The goal of the program is to spit out the two best teams. And when people can't agree who those teams are, that's exactly when we NEED the computers the most, and when we respect their decisions the LEAST. None of that changes, however, the fact that at the end of the day, THE COMPUTERS WORK.

by ChrisH :: Thu, 01/08/2009 - 12:32pm

If you have 2 undefeated teams, you don't need the BCS computers to tell you which two teams to pick.

So, this year that would have meant a Boise State-Utah National Title game, as they were the only I-A undefeated teams?

And those rankings are the most accurate in the land, without passion and prejudice. The goal of the program is to spit out the two best teams.

No, that's incorrect on multiple levels. As Bill James says, there is no criteria as to what the two best teams are. Is it going undefeated in I-A? Hello BSU-Utah! Is it giving up the fewest points? Hello USC! How about scoring the most points? Here's Oklahoma!

Of course, the computers would never know that about USC and Oklahoma, since they aren't allowed to consider the score of the games, or if they are home or away (which I didn't realize until now). So, let's setup a theoretical example:

Notre Dame and Alabama manage to setup a schedule where they play every single game at home this year. They play a tough schedule, and they manage to win everything single game 49-48 by blocking an extra point every game. They come out undefeated. This same year, USC plays in a down Pac-10 with a lower SOS, and has to go on the road for every single game, winning every single game 55-0. If the computers were to see that each team went undefeated, and lets say ND and Alabama have the #1 and 2 SOS, and USC is #3, you would fully expect the computers to return USC as the #3 team in the country, due to the requirements.

This is because the computers are prejudiced due to the requirements given to them by the BCS schools that they totally ignore certain things that, I think we can all agree in this case, would make them far more accurate than if they didn't have them. Being able to look at the two sets of rankings that Sagarin still puts up (even though I am not a huge fan of his rankings, at least he still puts up the non-BCS version for comparison), you can see which teams are punished (USC, Florida) by not including these factors.

None of that changes, however, the fact that at the end of the day, THE COMPUTERS WORK.

The Computers work in the way that the BCS wants them to work. I think most people who like computer analysis were happier when the computer rankings made up far, far more of the BCS ratings to try to overcome the biases of the voters. What the BCS has done in the past decade is:

- Reduce the importance of strength of schedule by removing it as a component
- Reduce the importance of independent analysis (computers) but cutting their influence by over half
- Add a new group of voters (the Harris Poll) that, from many examples we have all seen, is not the best informed set of voters and inferior to the AP voters that removed themselves from the BCS
- Add a new bowl to placate the non-BCS conference schools while changing all other aspects of the BCS to continually reduce the chance of one of those teams making the BCS Title Game

The computers have moved from being the main area of the BCS, to being something that just agrees with the voters, most of whom seem to be ill-informed on college football, or the coaches of teams who come in with their own internal biases and should not be trusted in any way to fill out a ballot that determines their own fate.

by Dennis :: Thu, 01/08/2009 - 12:46pm

"So how can we possibly expect humans, with all our flaws, to pick the right two? We can't. But what we can do is turn it over to the machines."

This shows you clearly don't understand how the BCS works. 2/3 of the weighting in the BCS is the human polls and 1/3 is the computers. So the BCS actually does expect humans to pick the right two and doesn't turn it over to the machines.

by Pete (not verified) :: Thu, 01/08/2009 - 12:52pm

What is the purpose of a playoff (tournament)? It is not to find out which team is the best. It is to get a playoff (tournament) winner. Do you really believe that most of 2007 that NYG was better than New England?

What shows the best team? There are some computers that make an effort (FEI, Sagarin, etc.) and I think when they are allowed to work they can work well. However, the BCS insists that the computers do not include things like Margin of Victory (MOV) or Home Field. Here is some news for you, Dean. If a team wins by 20+ points (3+ scores) then this is likely a more compelling victory than winning by a single point. A single play (or bad call) would not have alterred the outcome of the game.

I believe Bill's point about computers was that the better analysis is not allowed. Please take a look at Sagarin's Predictor or Rankings compared to what is required: ELO-CHESS.

However, I also appreciate a subjective point of view. During much of the year Florida was blocking punts more often than some other teams. However, once Harvin, Rainey, and Demps were not playing (as much) on special teams then their special teams were slightly less dominant. In a matchup with Oklahoma (who rarely punts), great punt blocking and punt return teams may be less important. Computers may not realize that Alabama's best player was suspended. Computers may not have realized that other injuries would have an impact. However, I do like how some of the computer rankings reflect that a "good loss" can be worth more than a "marginal victory".

What I disagree with is how many people clamor for change without a realistic view of what can happen right now. A playoff is not going to happen. The current contracts are setup through 2014. Even the Bowl + 1 that the SEC & ACC/Big East tried to force through last year was not realistic since the Big 10 and PAC-10 are locked in Rose Bowl.

Something I could see happening without much change would be a Classic Bowl + 1. In this scenario USC would have played Penn State in the Rose Bowl. Florida would have played an at large team, probably Utah, in the Sugar Bowl. Oklahoma might have played against Alabama in the Fiesta Bowl (or maybe against Big Least/ACC in the Orange Bowl).

Does this solve problems? Absolutely not. All it would do is keep the Bowls happy (who are shelling out a bit of money) and gives some good interconference play. However, the "best two teams" would then be decided after this. Two teams would play an extra game (not too much wear and tear, if that is a concern) and would play about a week later (they already are, so no overwhelming resistance there).

by Marko :: Thu, 01/08/2009 - 1:57pm

"What I disagree with is how many people clamor for change without a realistic view of what can happen right now. A playoff is not going to happen. The current contracts are setup through 2014. Even the Bowl + 1 that the SEC & ACC/Big East tried to force through last year was not realistic since the Big 10 and PAC-10 are locked in Rose Bowl."

It's not accurate to say that a playoff CAN'T happen because of the contracts that are in effect through 2014. Contracts can be modified, as long as all of the parties so agree. I do agree that a playoff is not going to happen in the near future because I don't think all the parties to the various contracts will agree to the changes that would be required to create a playoff system.

For example, the Pac-10 and Big Ten have made it quite clear that they are not in favor of a playoff system and are in favor of the current setup. So unless they completely change their position, there won't be a playoff (or at least there won't be a playoff that involves those conferences). But the existing contracts do not prevent those conferences and the Rose Bowl from changing their positions and agreeing to modify (or even terminate) their contracts, with the end result being a playoff system. Again, I don't foresee that happening, but it's incorrect to say that it CAN'T happen.

by Wanker79 :: Thu, 01/08/2009 - 4:31pm

I'd really like to hear some opinions of people who know more about the BCS than I do. But I can't come up with any reason why someone would argue against a 4 team playoff utilizing the current BCS bowls in a rotation. 1-vs-4 in one of the bowls, 2-vs-3 in another one, and then the winners play in the Championship Bowl.

Sure it leaves two of the bowls out of the mix every year, but currently 4 of them are left out. Might a 6 team, or an 8 team, or hell even a 16 team playoff be better? Arguably, but the more teams you add the more unrealistic the chances of it happening are. With only 4 teams, the season wouldn't be extended at all and the regular season would have every bit as much meaning as it does now. This year, USC and Utah would have still been left out in the cold, but you have to draw the line somewhere. And it would create two more must-watch games each year (presumably that could increase revenue which would make the schools and bowls happy), and it'd come closer to deciding a legitamate national champion on the field.

Besides the Pac10 and Big10 being dochebags, what could a counter-arguement to this system be? It seems so obvious to me that I'm having trouble finding one.

by sundown (not verified) :: Sat, 01/10/2009 - 1:26am

You've hit it on the head: there is no major problem with that proposal. And I fully expect that will someday be what happens. (All the major bowls will host playoff games, with each having the title game every few years.) From Day 1 that has been the obvious choice and I also was amazed as basically every big wig I saw interviewed the first several years of the BCS dismissed that idea outright, without expressing any problem against it other than a generic "it would never work." But, I've seen it growing in popularity, suggesting that the obvious answer may well eventually win out, after all.

by Temo :: Thu, 01/08/2009 - 10:54am

Actually, those 4 points were cited by James from an article by Dr. Hal S. Stern, the head of the Department of Statistics at the University of California-Irvine in a statistical journal.

by JMM :: Thu, 01/08/2009 - 11:06am

"Leaving the situation unresolved is unpopular because it's unnatural." Not sure I buy that statement.

The music industry has a structure which can be adapted to college football which would allow a set of better solutions. Kids who want to major in music have 2 options: they can go to a college and major in music or they can go to a conservatory. If they go to a college, music is a major which will represent somewhere around 50% of their course work. At a conservatory, their music workload would be closer to 80 - 85%.

Football continues as an extra-curricular activity because in its early days there were no legitimate career options for students who might want to pursue a career in football. Now, of course that is not the case. If schools could chose to offer a course of study modeled on the way music conservatories are established, kids who have the talent and desire to pursue a career in football can do it without figuring out how to fit it into a program that was structured to accommodate students who want a student-athlete college experience.

by Richie :: Thu, 01/08/2009 - 2:59pm

I've always thought schools should offer "Professional Athlete" as a major. Teach them how to do a decent interview beyond cliches (the better you are at this, the better chance of getting endorsements and a broadcasting career after playing), teach them how to deal with agents and invest their money, teach them how to be a coach (for after their playing career), etc. Teach them some actual skills that would be relevant if you become a pro athlete (just as an accountant is taught skills relevant to being an accountant), and most of these skills would also be somewhat helpful if the Pro Athlete career doesn't work out.

by tuluse :: Thu, 01/08/2009 - 5:09pm

One problem I can think of is that there are a lot fewer jobs available to football players than to professional musicians. Maybe if the UFL and AAFL can stay afloat.

by JMM :: Thu, 01/08/2009 - 11:30pm

Throw all of the coaching positions, high school, college and pro into the mix as well.

by Richie :: Fri, 01/09/2009 - 4:45pm

It's still probably more valuable than a guy who is a communications major, and barely goes to class and never graduates because he doesn't care. Maybe the Pro Athlete major will be more interested in his studies and graduate and learn some valuable skills along the way.

I'd be curious to see how the typical college athlete who doesn't make it to the pros does in his professional life after college.

by The Blow Leprechaun (not verified) :: Thu, 01/08/2009 - 11:36am

I say this every year, and I'll say it again. NCAA football isn't about what the fans want, it's about what's best for the universities that participate in it.

This is not the NFL, it is not supposed to be primarily a business. "Best product" is only one of many goals for the NCAA, and it hopefully isn't - nor should it be - the most important one.

It's pretty obvious that finding the national champion is just a footnote to the NCAA season, not the goal of it.

by cjfarls :: Thu, 01/08/2009 - 12:01pm

Except that it DOESN'T represent whats in the best interests of ALL universities... only those in the major conferences. Those schools who have the power... because of tradition & money - which comes from the fans.

If you're Utah or any other mid-major school, your consistently being screwed because of "what the fans want", as seen through TV revenues, etc...

Fixing the BCS makes sense because it:
A) Makes sense for the fans
B) Makes sense for the majority of schools

Unfortunately, because fans continually send mixed messages (what they say is different than how they spend their money), the "good ole boys" in the major conferences can continue to exploit the fans and other schools, and justify it by hiding behind lame excuses like "we don't want to add playoff games because it will interfere with the academics of the players."

A 4-team playoff, with the Bowls as opening round would solve 90% of the problem. An 8-team playoff would completely solve it. This year is tough, because there are legitimately 5 or 6 teams that could make the claim that they deserve in...

I've advocated a 4-team playoff because while there are OFTEN 3 or 4 legitimate contenders, years like this are rare and 1 extra game for 2 teams is not a big deal.

I'm less concerned about how you determine the 4 or 8 teams... either computers or polls will have biases... but in the end if you have a playoff for 4 or 8 teams, either method is likely to get the "best" 4 or 8 most of the time.

by Will :: Thu, 01/08/2009 - 1:14pm

"Fixing the BCS makes sense because it:
A) Makes sense for the fans
B) Makes sense for the majority of schools"

The top two rated bowl games, BY FAR, this year were USC vs Penn State and Texas vs Ohio State. The fans have emphatically stated that they don't care about Utah, they don't care about Cincinatti, and they don't care about Virginia Tech.


by Richie :: Thu, 01/08/2009 - 3:03pm

I think they don't care about those games/schools because the games were essentially meaningless and the schools have small followings. USC, Texas and Ohio State have huge national followings, because they are regularly in bowl games (because of the stacked system). Utah-Alabama was played on a Friday night, which probably didn't help anything.

If the Utah-Alabama game actually meant something, and led to a round 2 matchup of USC-Utah (or whatever), I am confident that it would have had better ratings.

by Will :: Thu, 01/08/2009 - 5:44pm

It's absolutely not true. Even in the NCAA tournament when George Mason went on one of the greatest Cinderella runs of all time, their ratings were consistently significantly worse than games between traditional powers. Their Final Four game had about 10% less viewers than the equivalent game the prior year.

It's true for all sports at all levels - it's why the TV execs hope like heck that the Yankees, Red Sox, Dodgers, and Cubs make the World Series or the Lakers, Celtics, Bulls, or Knicks make the NBA finals.

The BCS makes a lot of money for itself by making sure the traditional powers are always in the biggest bowls. If they stopped, then the Rose Bowl, Pac 10, Big Ten, and Notre Dame would go back to their old system when they snubbed the "Championship Game". What good would a playoff be if the Pac 10 and Big Ten didn't participate?


by The Blow Leprechaun (not verified) :: Thu, 01/08/2009 - 3:27pm

No system will benefit all the universities. Bowl games benefit /more/ universities than a playoff would, because it involves /more/ teams. Unless you have some kind of revenue-sharing, a playoff is only going to produce added revenue for, let's say 16 teams. Bowl games provide revenue to a greater number of schools with a lower goal for qualification (don't have to be a top 16 team).

Further, even a minor dilution in the importance of the NCAA regular season would damage revenues for virtually all teams. It's an easy argument to make that college games are so well-attended because "every game matters." I think this would be a minor effect in each instance, but spread out over so many universities and teams, it could be a lot of money.

No, I maintain that the bowl system is a more utilitarian solution for the NCAA. It benefits more teams better than a playoff would. A playoff benefits the very top more, and many fans think it would make them happier (a proposition I don't really buy, I think there would still be just as much controversy), but it makes less utilitarian sense for the NCAA as a whole.

by Richie :: Thu, 01/08/2009 - 3:32pm

I always feel the "every game matters" argument for the CFB regular season is bogus. It only matters until you lose 1 (or 2) games. Once your team has lost a couple of games, the rest of the season is mostly meaningless because the chance of making the championship game is almost nil. As a UCLA fan, the only postseason games I currently would care about is a championship game or a Rose Bowl. I don't give a crap if they make the Las Vegas bowl. But if they were the 16th seed in a tournament, you can bet that I'd care about that round 1 matchup.

by The Blow Leprechaun (not verified) :: Thu, 01/08/2009 - 4:28pm

Right, but if it only matters after the first loss now, the issue is if a playoff shifts that until now only the second or third loss matters.

It wouldn't make the regular season meaningless, the issue is more whether or not it devalues it.

by S.K. (not verified) :: Thu, 01/08/2009 - 11:55am

Is anyone else getting just a blank page with the Slate logo? It's not just the link here, I can't get to the story from the main Slate page either. Maybe it's a Canada thing.

by ernie cohen (not verified) :: Thu, 01/08/2009 - 12:08pm

I generally love James' writing, but he's off the mark here, in fundamental ways.

1) He completely misunderstands why the BCS needs computers. The BCS doesn't need computers for accuracy. The BCS needs computers for fairness. Unlike people, computers will not vote for a team because their school is named Texas or USC. The use of computers is exactly analogous to the use of rules of evidence in lawcourts.

2) Obviously, he is not an economist. The reason for keeping victory margins out of the ranking algorithms is, again, not accuracy. The reason is to avoid giving teams an incentive to run up the score. If any analyst wants to boycot because he thinks that ranking accuracy is important enough to encourage poor sportsmanship in collegiate sport, I say good riddance.

Frankly, I would expect competent statisticians providing computer rankings to propose some alternative way to get information about blowouts without relying on the final score. For example, record the earliest times at which the victor was ahead by 10, 14, 20 points, and whether the final score exceeded 14. This would probably provide better information about domination than the final score, without incenting blowouts.

by Tom Gower :: Thu, 01/08/2009 - 12:50pm

Nice ideas, but these are actually post hoc explanations of changes made for other reasons, namely that the then-existing BCS formula produced the "wrong" result, and making the changes w/r/t margin of victory produced the "right" result. The BCS uses Billingsley, for cripes sake, they're not interested in using the highest quality computer rankings unless they produce the "right" result.

by Jed Dawson (not verified) :: Thu, 01/08/2009 - 12:55pm

1. Motivation is belied here by the historical record. If computers are consistently being recalibrated to chase opinion - as we've seen over and over again - then they're not really objective, are they? And that's not really like evidence in any court, except a Stalinist one, is it?

2. Under the current system, the computers account only for a fraction of the results. So even if you invented a gee-whiz bright shiny "ranking algorithm," it would struggle to balance out against human input - read: polls - in which scores do matter a great deal, including great big scores not tied to any arbitrary timetable.
2a. And even if they employed the gee-whiz algorithm, it would diverge so greatly from the conventional thinking of the polls - which, fairly, expect scores to matter - that it would be neutered overnight under the "computers is broken!" clause we've seen repeatedly.

Frankly, I would expect competent statisticians to run for the hills when it comes to this massive shell game called the BCS. The inclusion of "computers" are intended to lend the proceedings an air of legitimacy, but an eminence gris it most certainly is not. It's the exact opposite, really - the "advisor" in a bright shiny robe who is paraded in the throne room, briefly consulted, then yanked offstage by a big hook to make room for the status quo.

by DGL :: Thu, 01/08/2009 - 2:03pm

"The reason for keeping victory margins out of the ranking algorithms is... to avoid giving teams an incentive to run up the score."

Well, that's worked really well. Clearly, since the BCS ordered the computer rankings to disregard margin of victory, college coaches nationwide have stopped running up the score, and a wave of sportsmanship has spread over Div 1-A football.


by Bad Doctor :: Fri, 01/09/2009 - 3:34pm

"Frankly, I would expect competent statisticians providing computer rankings to propose some alternative way to get information about blowouts without relying on the final score. For example, record the earliest times at which the victor was ahead by 10, 14, 20 points, and whether the final score exceeded 14. This would probably provide better information about domination than the final score, without incenting blowouts."

Totally agree, and I think James totally agrees as well. There would be reasonable ways that statisticians could use the score to indicate the quality and thoroughness of a victory without encouraging coaches to run up the score (such as capping the margin of victory and/or using margins at earlier points in the game). Any statistician that was asked to put together a calculation to determine the best college football team (whether best means "most accomplished to this point" or "best going forward"), I would think, would absolutely insist on using such data. But the BCS statisticians have been forbidden to use any score data or brainstorm any ways to use it without promoting poor sportsmanship, because, as noted by one of the statisticians and quoted in James's article, the BCS itself refuses to use any statistical ranking that uses any sort of scoring margin data in its calculation! This is one of the reasons that James suggests that these more than "competent" statisticians should protest against this sort of intellecutual handcuffing by simply refusing to take part in the process, given that the process relegates them to mere enablers rather than active participants.

by johonny (not verified) :: Thu, 01/08/2009 - 12:33pm

love me some Bill James

by Tom Gower :: Thu, 01/08/2009 - 12:40pm

I mentioned this in my book review guest column, #1 is the underlying basic problem. The BCS was created to match the two best teams, with the definition of "best" being the top 2 in the AP Poll, in order (including the thorny problems of Colorado-Georgia Tech, Washington-Miami, Nebraska-Penn State, and Nebraska-Michigan). Oh, but you can't just use the AP Poll, because then the coaches will have a hissy fit. So, there enter the computers, to make sure you get the top 2 teams in the AP Poll.

Ah, but the problem is AP voters aren't necessarily consistent or logical with how they evaluate teams. And, lo and behold, a system designed to spit out the top 2 teams in the AP Poll that uses inputs other than "who are the top 2 teams in the AP Poll" sometimes spits out teams other than the top 2 teams in the AP Poll. When that happens, it's back to the drawing board, capping MOV at 21, removing MOV altogether, bonus points for quality wins, and all that other rigmarole. I think by now they've unofficially dropped the requirement that any system tweaks produce the "right" past results, either BCS-era or the 90s title confusion that spurred the creation of the BCS, but the underlying problem is still there-the raison d'etre for computers in the BCS has not been and is not now because people (sportswriters in particular) accept they're as good as (let alone better than) sportswriters at evaluating team quality.

by Kevin 11 (not verified) :: Thu, 01/08/2009 - 12:58pm

James says:

"The smaller schools, which outnumber the big football powerhouses about 5-to-1"

This simply isn't true. College basketball yes, football no.

by Joexxx (not verified) :: Thu, 01/08/2009 - 3:21pm

Right. My guess is he was including the other NCAA divisions in that ratio. He makes a comment about how if the football powers left the NCAA, then the NCAA would be in between whatever organization they formed and the NAIA. So it seems like he isn't distinguishing between the 4 levels of NCAA football (FBS, FCS, D-II and D-III).

by Lance :: Thu, 01/08/2009 - 8:33pm

Huh? There are about 120 or so D-1A schools. I don't think I can name more than 20 "big football powerhouses" though. Sure, the Big XII may be a power CONFERENCE, but how many schools in are honestly powerhouses? Two? Three? Baylor, Kansas, Iowa State-- those aren't powerhouses. I don't even think that Oklahoma State and A&M are powerhouses. The Big 10 has some powerhouses, but I don't think anyone is going to include Northwestern or Indiana in that list.

Perhaps we have a different understanding of the term, though. I don't know where I'm going with this, except to say that there are really only a handful of schools-- maybe 10-- that are year-in and year-out (with an occasional down year or two) powerhouses. At least, for the current BCS "incarnation" of CFB. What-- OU, Ohio State, Texas, USC... and then? Kansas State hasn't been around since 2003 or so. Virginia Tech has been a top 10 team more than a few times, but not so consistently. LSU? Florida? I don't know.

by Will :: Thu, 01/08/2009 - 1:17pm

I always hated running up the score as it tends to make a 57 - 17 game more impressive than a 17 - 0 game. I would argue if these same matchups were played 50 times, the 17-0 winner would win more often than the 57-17 winner.


by GlennW :: Thu, 01/08/2009 - 1:19pm

> The BCS doesn't need computers for accuracy. The BCS needs computers for fairness.

I think for most of us there's an intersection between accuracy and fairness, and if a system is markedly inaccurate then it's not really fair. I mean, we could determine the BCS championship game participants via a series of coin flips and the system would be completely non-subjective, but it'd be arbitrary and inaccurate and therefore "unfair", at least to the teams which were truly the best in the country.

Accuracy should be a primary goal, in my opinion, and similar to the NFL passer rating formula, throwing together components A, B and C in equal weights and then pretending that the output is particularly meaningful doesn't make much sense. James nails this problem from its starting point-- the lack of meaningful goals for the system.

by Goran (not verified) :: Thu, 01/08/2009 - 1:31pm

I'm not really an advocate of BCS, but one of the things that completely bug me, when talking about a playoff format, is mentioned in this article:

"...avoiding a postseason tournament that would be preferred by the overwhelming majority of fans."

Who says? Where does this say and we know it how exactly?

by GlennW :: Thu, 01/08/2009 - 1:50pm

> "...avoiding a postseason tournament that would be preferred by the overwhelming majority of fans."

I believe this is almost certainly true. ESPN ran one of its online polls in November and about 75% of the respondents said they'd prefer a playoff system to the current system. I know such polls are unscientific, but the 3-to-1 margin is difficult to ignore and is consistent with my subjective perception of public opinion.

by Alex51 :: Fri, 01/09/2009 - 12:22pm

I know such polls are unscientific, but the 3-to-1 margin is difficult to ignore

No, the 3-to-1 margin is not difficult to ignore. An online poll has so many critical flaws that its results are virtually meaningless. For instance, if people who want a playoff are three times more likely to take the time to answer the poll, then you would expect a 3-to-1 margin if the reality was more like 50%-50%. I could make an online poll that would get completely different results if I were given control of the exact wording of the question and the location of the poll. There has not, to my knowledge, been any scientific poll that confirmed that an overwhelming majority of fans preferred a playoff system.

by Robo-Hochuli (not verified) :: Sat, 01/10/2009 - 12:49am

I would like to add that ESPN isn't exactly balanced in its treatment of playoff vs. BCS debates, and it would probably stand to gain some of the spoils if a playoff format were ever adopted.

by Thok :: Thu, 01/08/2009 - 2:18pm

It seems that the main reason the BCS uses a formula to determine who is in the national championship is so they have an automatic justification of their selection process that allows them to avoid taking the blame for mistakes.

Ironically, they'd probably be better off from a PR point of view of just having a BCS selection committee that picks the teams: they'd still get criticized for mistakes, but there would at least be some recognition that somebody is attempting to do the difficult task of assessing that year and that the people in charge are doing some thinking and have some perspective.

(In fact, if they ever go to a playoff, I suggest that something like the BCS formula is kept around, but is used only like the RPI: as a guide to help judge the best teams, not as an absolute criterion to pick 4/8/16 teams and nobody else.)

by DFJinPgh (not verified) :: Thu, 01/08/2009 - 2:22pm

The purpose of the BCS:

"That the existence of this system has the purpose of justifying a few rich conferences in hijacking the search for a national title, avoiding a postseason tournament that would be preferred by the overwhelming majority of fans."

(I tried to use strikethrough to eliminate the second clause - since it's really just speculation - but that's not allowed html. :-( ) All the other points, both in the article and the comments above, are subsidiary to this. As usual, simply following the money explains the most.

by Led (not verified) :: Thu, 01/08/2009 - 2:24pm

I'm anti-playoff. I'm one of the few people who think we should go back to the old bowl format and the leave the "National Championship" as the unofficial, media-created title it always was. When winning your conference and then winning your bowl are the main goals for a team, it's possible for many different teams to have successful, rewarding seasons. What's wrong with that? Why do we have to have one champion in I-A college football? The disparity between teams and conferences is so great and the number of games is necessarily so small compared to the number of teams that any system for naming a single champion will be arbitrary. A playoff would be less arbitrary, but I think it's a mistake to view it as a panacea. In some ways, the BCS is the worst of all possible worlds because it breaks down the traditional links between conferencs and bowls in return for a championship that is no more legitimate than it was when it was just a creation of the Associated Press.

by Bowl Game Anomaly :: Thu, 01/08/2009 - 3:17pm

Bingo. You nailed it. The BCS system took away what was good about the old system without adding any real improvement. They keep trying to "fix" the BCS but as someone noted above fixing the BCS means making the 2 top teams match the top teams in the AP poll, and nothing more. That being the case, it can't be fixed (regardless of the fact that each change to the computer contribution actually makes it poorer at accurately and objectively identifying the best teams by removing important input).

Instead of trying to use computers to legitimize the AP poll, why not just accept the AP poll? I would do away with the championship game entirely, but that's not necessary if you still want to end with a #1 vs #2 game. So how about returning to the traditional bowl system, and then having a "plus-one" game afterward with the top 2 teams in the AP poll? Would anyone really be upset with that?

(Formerly "The McNabb Bowl Game Anomaly")

by Will :: Thu, 01/08/2009 - 5:19pm

Agreed 100%


by Wanker79 :: Thu, 01/08/2009 - 5:32pm

I've always been a staunchly pro-playoff system guy. As someone who grew up in Jersey and never went to a 1A school (and nobody in my family strongly supports any 1A schools either), I've never had much interest in college football. As they say, I don't have a dog in the race. But I've been raised on a steady diet of the NFL, where winning the SuperBowl is the only thing that really matters. So the idea of any system besides a playoff is so foreign that I just dismiss anything else.

But you may have just converted me, Led. For some reason I don't quite understand, the idea of getting rid of the BCS entirely (instead of trying to fix it) is really resonating with me. I'm going to have to chew on that for a while, but getting rid of the championship game is coming at the problem from a completely unexpected place (at least for me it is), and my first reaction is strongly positive.

by Packer Pete (not verified) :: Thu, 01/08/2009 - 2:37pm

Obviously, the key point is the formula run through the computer that determines the rankings. Does the BCS reveal their formula to the public, and where can that be found?

Recently, I spent some time checking into the NFL's passer rating formula. It's based on completion percentage, yards per attempt, touchdown pass percentage, and interception percentage, all weighted differently, and all subjected to divisors and multipliers that I couldn't find explanations for.

If you really wanted to rank NFL passers, wouldn't you consider 3rd down conversions? Wouldn't you weight a 300-yard passing game against the Steelers differently than a 300-yard game against the Lions? Why are touchdown passes important? (If a QB passes his team 79 yards downfield and the fullback runs it the last yard into the end zone, why doesn't the QB receive partial credit for this score? If McNabb dumps a short screen to Westbrook who promptly runs 70 for a score, why should McNabb get full credit for a TD?)

The passer rating, like the BCS rating, is some vague yardstick that sportscasters, writers, and fans can glom onto. The ratings might be vaguely correct--Manning is a better QB than Kitna; Oklahoma a better team than Murray State -- but the formulas don't dial in well enough for the fine judgments such as who is the number one college team and who is the number two, and should meet in the championship game.

Add me to the tournament list.

by Joexxx (not verified) :: Thu, 01/08/2009 - 3:15pm

"Obviously, the key point is the formula run through the computer that determines the rankings. Does the BCS reveal their formula to the public, and where can that be found?"

There are six different computer rankings that are averaged. I don't know how many of them, if any, disclose their formulas.

by Tom Gower :: Thu, 01/08/2009 - 3:39pm

Here is Richard Billingsley's explanation of his "formula." Fun quote: "But my system is not about mathematical algorithms. It’s about rules created to compliment a common sense human response to a football game." Even more fun, his old explanation, since revised:

"I wrote the program myself and it's not written using fancy math equations, just simple addition, subtraction , multiplication and division. It's the RULES that make the system unique and the rules are MY RULES. Rules that make sense from a fan's perspective. Rules that come from 32 years of experience in which I researched the ENTIRE 132 years of College Football."

For something slightly more sane, check out the Massey Ratings explanation. My understanding is that Sagarin's rating is also MLE-based, as is Wolfe's. People with more statistical competence than I can probably tell you more.

by Pat (filler) (not verified) :: Thu, 01/08/2009 - 6:37pm

Elo is equivalent to a stepwise maximum likelihood estimation.

Billingsley is insane. Anderson-Hester might be too: they don't include any information. Massey's time-weighting is also insane. Wolfe and Colley are the two most statistically sound rankings out there. Colley's treatment of IAA is a bit weird, but in the end it works primarily because you don't give a crap about ranking IAA teams - just their effect on IA teams.

by morganja :: Thu, 01/08/2009 - 4:47pm

Great article by James, as always, but we really need to consider the root of the problem. Why are publicly funded Universities in the sports business?

James accurately describes how the problem stems from all the universities trying to grab as much of the pot as possible. Why are public universities primarily about generating revenue?

The answer is that in every state in the union, the public university system is run as a patronage machine to generate illicit cash for politicians and their friends.

Check out who the head of the university system is in your state, what qualifications he has, and how he got the job.

In North Carolina, Erskine Bowles is a long-term political hack whose family has been heavily involved in the North Carolina political machine for generations.

Every University president seems to spend almost his entire focus on what they euphemistically call "fund-raising". They are 'run' by Boards of Trustees which consist entirely of local political appointees who make a ton of money for themselves and their friends.

The whole system is corrupt at the core. That is why the BCS is such a clusterfu*k. It has one function. To generate cash for political machines.

There is only one reason why public universities don't compensate athletes according to labor laws that apply to every other profession. Because each dollar paid to the athletes is a dollar diverted from the political machines and there is never, never, never enough money to go round when were talking about political patronage.

I don't follow college sports because I refuse to support corruption in government. I understand how others don't care, they just want entertainment, but people should understand what this entertainment actually costs them in misappropriated taxes and the hidden taxes of corrupt state government.

If University sports were governed by the same rules that apply to a free and open market, they would have worked all this out years ago, fairly. But you can't expect a collection of thieves dividing up their booty to follow any free, open market principles.

Anyhow, I'll return you to your discussion about the BCS.

by Kevin Eleven :: Thu, 01/08/2009 - 9:05pm

As soon as you start selling tickets, you're running a business. It's not too much to ask, in my opinion, for a sport to support itself.

Since the schools are "in business" anyway, they might as well make the most of it.

by Pat (filler) (not verified) :: Thu, 01/08/2009 - 6:32pm


James is wrong regarding margin of victory. It's a bit depressing, unfortunately.

A system used to select teams for a championship requires one thing: minimum bias. Accuracy is far less important - if you're wrong, at least you're just randomly wrong, rather than predefining new rules for a sport. Margin-of-victory is by definition biasable, since football is a timed sport.

His argument is that "it's like a surgeon not using a scalpel." Not really. Using margin of victory is the same thing as a small town doctor killing infants that don't meet his definition of "healthy" hoping that that will make his job easier in the future.

by AnonymousPerson (not verified) :: Sat, 01/10/2009 - 6:04am

What's the beef with margin of victory? Is it a retrodictive vs. predictive thing? Because I know MOV has been proven to be a better predictor of future games than straight-up Win%.

by Pat (filler) (not verified) :: Sun, 01/11/2009 - 10:38pm

Essentially, it's predictive vs. retrodictive, yes.

This goes back to James's "point #1" - the BCS doesn't say what it's trying to do. But we can correct that a bit, and say "what should the BCS be trying to do?"

What it should be trying to do is say "these two teams have had the two best seasons in all of college football." That is the only thing that makes sense: you can't say "who's the better team right now." What happens if in the last win before the bowls, the #1 team loses their starting QB? Do they not get into the NC game? What happens if they lose him the day before the game? The week? The month?

Margin of victory is similar: if you have two teams, who by some fluke play identical schedules, and all of their game scores are identical except one, where team A wins 21-17 and team B wins 21-16, without knowing the details of the game, how can you possibly suggest that the 21-17 team won better?

Remember, the BCS is picking teams for a playoff. A two-team playoff, yes, but still a playoff. Now look in the NFL, and look for where "margin of victory" is on the tiebreaker list. Tenth. They rank "margin of victory in common games" above margin of victory overall. They rank "strength of victory" (meaning record of opponents that the team has beaten) above margin of victory.

That's right: in the NFL, which most people can agree is a decent playoff system, they value a ranking which completely discounts margin of victory over one that doesn't.

Margin of victory is a decent predictor of future performance. But it's a biased indicator of past performance - primarily because the goal of a football game is not to score tons of points, but to win.

by Will Allen (not verified) :: Thu, 01/08/2009 - 8:47pm

Morganja is entirely correct that the BCS is simply a subcomponent of a larger system in which taxpayer funds are used to enrich politically connected people under the guise of serving higher education. I'd personally be perfectly happy if all the University presidents saw the ghost of Robert Maynard Hutchins tonignt, and got their schools out of the football business tomorrow.

Since that ain't going to happen, I'd just as soon see college football maximize profits, and not just for the schools in six conferences; cartels are really, really, really obnoxious when they are taxpayer supported. Thus, I'd like to see the regular season start on the last Saturday in August, which would leave a minimum of 13 Saturdays for all the conferences to identify their champion. Then, on the first Saturday in December, I'd like too see the champions of at least eight conferences begin a playoff on the home fields of the top seeds. I'd play the semis on Jan. 1 at two of the traditional bowl sites, with the big game a week later at a third traditional bowl site.

I think such a format, if marketed well, would strike a de well at diluting the regular season, since finishing 2nd in one's conference would mean being left out. If the home filed in the first round was awarded purely on the basis of the highest quality nonconference wins, it would discourage the deliberate scheduling of creampuffs, the worst feature of college football. I also think it would be a cash monster, eventually rivaling the Super Bowl.

by jimmy oz (not verified) :: Thu, 01/08/2009 - 10:05pm

Strength of victory needs to be done by humans who can gauge when a team sends in its scrubs and starts coasting. Otherwise beat downs are the order of the day.

The Massey rating contains flawed philosophy:
Game Outcome Function (GOF)
Given the score of a game, GOF(pA,pB) assigns a number between 0 and 1 that estimates the probability that team A would win a rematch under the same conditions

There is no valid reason to ignore the empirical evidence of Team A's win and assign a probability to it. Under the game conditions, Team A has won 100% of the time and any re-match would be under new conditions.

If the probability scores are taken from games through history which have been re-matches of contests earlier in the season, this still has nothing to do with the teams' skill level.

by Key19 :: Fri, 01/09/2009 - 1:56am

I honestly feel that the previous system was way better than the BCS is now. In the previous system, there was some lee-way in terms of who would be named National Champion. I think that the Bowl Games were a heck of a lot better when you had the scenario of "whichever team impresses the most will probably get the National Championship." If it's really tight, then name Co-National Champs. I know, everyone always wants them to play if that happens. But sorry, it just is not going to happen. There is no way to possibly determine a true Champion in my opinion. There are just too many teams, too many styles of play (with certain styles trumping others, therefore making it even harder to determine which team is the best), and not enough time to play the proper amount of games. I honestly think that college football is a joke. Sure, it's fun to watch. But I'll never be able to take it seriously because it's so arbitrary in a lot of ways and so ridiculous in others.

My solution to making college football more meaningful is this:

1. Break up Division 1 into two conferences. The legit conference and the pointless conference. The legit conference is full of let's say the top 50 football schools in the nation. This will be determined based upon their historical place in the game, what they have done lately, and how much talent they have. Any team in the previous year's top 25 is automatically in. The other 25 spots are up for grabs. Honestly, this number does not have to be 50, but it seems convenient so I am using it. Whatever number ends up being necessary to accomplish the goal is fine. Anyways, once the field is complete, these 50 teams play only each other for the duration of a regular college football season. This ensure quality matchups every week. The current format is a sham. How do you expect to be able to determine the best teams when the best teams almost never play each other? Sure, the SEC has some perennial powerhouses that clash every year. But overall, USC does not play the majority of the elite teams. Neither does Ohio State (or Michigan when they were good). It's always coming down to "well, since the Big 10 Champ isn't in the SEC, and their record is not better than 2 or 3 of the top SEC teams, the Big 10 team must be worse because their schedule was easier." This happens simply because each non-SEC team plays approximately 3 meaningful games all season. The rest are just "let's see how much we win by." You CANNOT possibly determine the best team with this small sample size of meaningful data. Once again, my field of the 50 best teams playing each other every week fixes 75% of the problems with the system. If good teams play other good teams every week, you find out who the best teams are. Therefore, you still have the "every game counts" thing that pro-BCS-ers love so much, and you also have the sort of tournament flare that tourney-fans like. If you're playing a legit team every week, it's pretty much a tournament! There will be no more undefeated teams. It just would not be possible with the schedule. If there is one, they are pretty much going to be the national champion.

This format has the benefits of:

No more running up the score arguments. If a team runs up the score, it means that they legitimately outplayed their opponent, rather than "out-recruited" them. This would be just like a NFL game.

The good schools matter, the bad ones don't. Who cares if Ohio State can go undefeated or have 1 loss every year in the Big 10? It never tells us anything about them and they are always getting the shaft for being in a bad conference! Now, Ohio State's schedule includes USC, Boise State, Oklahoma, Florida, Texas, Utah, etc. And of course, Michigan. I guess Michigan gets into the Top 50 automatically. *note: this rule doesn't have to make it into the final version of the system*

The system will be fluid. The teams in the Top 50 change every year. The 10 (or so) best Pointless Conference schools get to play the 10 worst Meaningful Conference schools, winner take all. If the Pointless Conference school wins, they are in the Top 50 and their opponent is out. Better luck next year. This ensures that programs who are smaller right now but really do give a crap about having a good football program can actually make their way up the food chain. One flaw in this is that the best players will always want to play for teams in the Top 50, therefore making it nearly impossible for teams out of the Top 50 to build significantly, but that's how it is already anyways. Florida always beats out Appalachian State for the best players. Nothing different about my system.

Overall, I'm tired of this entire college football system. If you truly want to find the best team, then stop letting teams play horrible schools for half of the season every year. It is just a waste of time and stats. The NFL is great because every game is legit (well, I guess Lions/Rams would be an exception, but that's the exception that makes the rule). The NFL ALSO has a playoff system on top of the fact that all of the games are legit. That's why the NFL is so good and so popular. The 07 Patriots are the closest thing to a college football team as you'll ever see in the NFL. They literally looked like they were Florida playing their non-conference games. Going undefeated is ok. I feel fortunate to have seen the greatest regular season in NFL history in my lifetime. But going undefeated loses its luster if in any given year, multiple teams can accomplish that feat.

In the end, if you make the games more legit every week, everything works out better. The Playoff/No Playoff argument becomes nearly moot and life goes on and we have great football every week in college.

Honestly, I'm partially joking here, because some of these points are a little too over-the-top to ever work. But I'm also very serious. I am not a genius. I don't think I'm one. I don't even follow college football that much so I really am not the person who should decide the system. These are just some thoughts I have on how I personally would enjoy college football more. If you like it the way it is, fine. Disagree with me. It's ok. Just be dignified about it. I'd love to hear some thoughts on how to improve my model from some other creative minds. Or how to make it slightly less ridiculous from the die-hard BCS supporters. Overall, I just want college football to be better. This is how I would do it. Thanks for reading and have a great evening.

by jimmy oz (not verified) :: Fri, 01/09/2009 - 2:03am

i went and asked Ken Massey. Its not win probabilities at all. All this method does is assign an arbitrary value on winning margin based on final score, but doesn't reward teams running up the score as i expect 35-0 is similar to 70-0 in the model.

Doesn't note whether starters are involved, or whether a heavy dose of fullback dives and base defence were used. this is why winning margin should stay out of computer calculations.

From: Kenneth Massey []
Sent: Friday, 9 January 2009 04:04 PM
To: James S
Subject: Re: Flawed philosophy

It "estimates" the probability with a mathematical model. Obviously the winner in a 21-20 game is not necessarily the better team (a rematch would be close to 50-50), whereas the winner in a 63-0 game almost certainly is.

On Thu, Jan 8, 2009 at 9:10 PM, James S <> wrote:

Hi Ken

About your ratings...

Game Outcome Function (GOF)
Given the score of a game, GOF(pA,pB) assigns a number between 0 and 1 that estimates the probability that team A would win a rematch under the same conditions
I can't see a valid reason to ignore the empirical evidence of Team A's win and assign a probability to it. Under the game conditions, Team A has won 100% of the time and any re-match would be under new conditions. If the probability scores are taken from games through history which have been re-matches of contests earlier in the season, this still has nothing to do with the teams' skill level, and is only a statistical probability.

Are you able to elaborate further?


by Paulo Sanchotene, RS, Brazil (not verified) :: Fri, 01/09/2009 - 9:08am

Come on. Let me say a truism. It's just impossible to organize a "league" of more than 100 teams without changing the bases that built college football. You must change the system to something similar to "College Basketball March Madness" or keep the bowls. BSC, somehow, manages to keep bowls alive. Some people thought the old bowl system needed a change, and they did without a revolution. Every conference is a different league, bowls are a totally different championship. The only big difference is that now there is "National Championship Bowl".

That said, #1 is easy to answer: the goal of BCS system is to pick two teams that are closely the best duo after regular season and make them play against each other. That's why they change the power of computers on ranking. If the problem is perception - afterall, the teams don't need to be, but must be seen as the best - the polls just work better! Bill James may like it or not, but people still has more confidence on the authority of the polls over statistics or computers.

by Paulo Sanchotene, RS, Brazil (not verified) :: Fri, 01/09/2009 - 7:27pm

BTW, I prefer the old system...

by Botswana Meat Commission FC (not verified) :: Fri, 01/09/2009 - 11:40am

As as a non-fan of the college game, I read stuff like this and laugh and laugh and laugh. At least Bill James is enough of a radical to suggest that the system can NEVER be fixed within the confines of the BCS, the bowls and the athletic conferences.

by Fox Executive (not verified) :: Fri, 01/09/2009 - 4:30pm

FOX did its share to stop the bowl system - with announcers like NC game and Fiesta Bowl, it was trying to make the games unwatchable.

Now just tell the fans of the schools to stop buying tickets to the games, and you'll see some changes.

by Lew Alcindor (not verified) :: Sat, 01/10/2009 - 12:21am

Baseball stats are limited in their application to football. Football is too fluid, the injuries skew the ability to predict, the game is timed so the number of plays and how the game is played changes.

That doesn't mean you can't narrow the possibilities, but there is a lot more guesswork. The NFL even thrives on parity and unpredictability, making the margin of error even smaller.

BCS solution, there's 12 conferences...have a 16 team playoff and you'll have a true champion by taking each conference champion.

by ernie cohen (not verified) :: Sat, 01/10/2009 - 2:56pm

Remember, this is college football; why are we giving the A students extra homework? Here's the postseason system I prefer. Teams play extra games only if they have not yet won, say, 6 games. As teams win their 6th game, they drop out of the tournament, until there is only one team left.

by dave in san mateo (not verified) :: Mon, 01/12/2009 - 2:29am

I know I'm late to the game here, but I'm surprised that a discussion of the BCS among relatively intelligent individuals is taking place without addressing two fundamental problems of the bowl system (in my view, the two most important problems). (1) Time between games. If more than two weeks lapse between the last game and the next game, your system is officially a joke. Thirty days? Please. (2) Game location. Why do midwestern teams continue to accept a system that values games at exclusively warm weather sites (often a bus ride or less away from the opponents's home field) higher than all other games? Would Big Ten teams fare better if the Rose Bowl was in Happy Valley every year? Of course. The current system's problems are not limited to tournament size (we have a two team tournament now, for those of you who haven't noticed).

by Sancho (not verified) :: Tue, 01/13/2009 - 7:48am

Every conference is a different league. Bowls are a totally different championship.