BCS’ Computers Way Off in Finding Best Teams Suited For Championship Game

Note: This article was started Saturday morning, and finished early Sunday morning.  During Saturday’s action, both Iowa and Oregon (two top eight BCS teams included in this analysis) lost.  While these results affect the future BCS results, the said results do NOT affect the flawed philosophy of the BCS and its “computer rankings”.

The BCS claims to be fair and objective.  The BCS claims to have the best way to determine the two teams who deserve to make the National Championship Game.  In reality, the BCS couldn’t be any more misguided, and it’s not because the Division I Football Bowl Subdivision is missing a playoff system.  Rather, it’s an outdated and a far too simplistic system that the BCS has, a system far from being the best solution towards determining which teams deserve to compete for the national championship.

First off, the objectivity of the BCS gets thrown out the window when it includes the USA Today coaches’ poll, the Associated Press poll and the Harris Poll in its rankings.  That human element alone destroys the credibility of the BCS’ claim of being objective.  However, the polls are among the lesser concerns the BCS committee should have.  In fact, it’s their “computer rankings” and their decision to completely exclude margin of victory (or any other qualitative stat) that serves as the biggest problem in the BCS’ failure to fairly and correctly determine the championship contenders.

In a nutshell, the BCS “computers” focus on won-loss record and strength of schedule to determine which teams “deserve” to make the national championship and the at-large bids in the “BCS bowls”.  Right now, the only things the BCS has right are (1) having the BCS AQ (big six) conference winners in a BCS bowl game -and- (2) having the highest-ranked team from a non-BCS conference in a BCS bowl game (given the said team is in the top 12 of the rankings).  Furthermore, the BCS restricts a given conference from having more than two teams in the BCS bowl game, which is a relatively fair rule in most years.  These factors won’t be criticized.  However, what will be criticized is the lack of any factor in the “computers’” system besides won-loss record and strength of schedule.

The BCS’ criteria for finding the “most deserving” teams for their bowl games involve determining the teams that have beaten the “best” competition, without suffering too many loses.  The BCS logic is sideways at best, as the committee fails to factor in any qualitative stats into the formula.  The BCS more or less says the name of the game is to find “who beat the best”, when the true name of the game is to find “who is the best”.

To best discuss the ailments of the BCS “computers”, I’ll break down each “computer” system one-by-one:

Jeff Saragin NCAA Football Ratings
visit his ratings at USA Today
Sagarin makes it very easy understand his system.  He has two factors, ELO-CHESS and PREDICTOR, combine to make his final ratings.  One factor considers the results, while the other tries to add some qualitative factor.  Here’s exactly how Sagarin explains it:

In ELO-CHESS, only winning and losing matters; the score margin is of no consequence,
which makes it very "politically correct".  However it is less accurate in its predictions for
upcoming games than is the PURE POINTS, in which the score margin is the only thing that matters.
PURE POINTS is also known as PREDICTOR, BALLANTINE, RHEINGOLD, WHITE OWL and is the best single PREDICTOR
of future games.  The ELO-CHESS will be utilized by the Bowl Championship Series(BCS).

Whoops.  In attempt to mix quality of play (in whatever flawed play, it’s better than not using it at all), Sagarin has somewhat decent rankings.  However, the factor that only considers winning and losing is used to determine which two teams play for the national championship.  Why?  Because it’s “politically correct”.  Whoops.  As you learn in this post, if you didn’t already know, the BCS cares more about political correctness than actually doing its job and finding the two best teams in the nation to play for the championship.  Enterting today’s games, we had these numbers:

• Iowa: 2 ELO-CHESS, 15 PREDICTOR, 8 RATING
• Oregon: 4 ELO-CHESS, 4 PREDICTOR, 2 RATING

Anybody who’s anybody who watched Oregon trounce USC last week knows that the Ducks should be ranked higher than an Iowa team who trailed home last week entering the fourth quarter against Indiana.  Let’s take a look at the top eight in both the ELO-CHESS and RATING scores in the Sagarin rankings:

• RATING: Florida, Oregon, Alabama, Texas, TCU, Boise State, Cincinnati, Iowa
• ELO-CHESS: Florida, Iowa, Cincinnati, Oregon, Boise State, TCU, Alabama, LSU

Whoops.  Texas isn’t even in the top eight in the BCS portion, though they’ve played more like the fourth-best team in the nation after you consider the weak Big 12 conference the Longhorns play in.  To put it as simple as needed, RATING > ELO-CHESS when it comes to finding the best teams.

Anderson & Hester Ratings
visit their ratings at AndersonSports
Ah, nothing’s better than using the wrong system, except for pumping your chest when you use the wrong system.  This fool, Jeff Anderson, actually have the gall to say their right, and Chris Fowler is wrong for complaining about the BCS getting rid of margin of victory in their ratings.  Heck, the article itself is called “Why Chris Fowler is wrong”.  L-O-L.  Puh-leeze.  Let’s see how this article begins:

On ESPN SportsCenter on Saturday night, Chris Fowler criticized the Iowa Hawkeyes for their close wins and labeled the BCS decision not to include margin of victory in its computer rankings as “stupid.”

Couldn’t be more true.  Let’s continue:

The adjective applies, but not to the BCS decision.

Whoops.  Again.  Of course, Anderson follows up that bonehead statement with more stupidity: his “reasons” why he’s “right”.  He starts with:

The object of the game is to win.  Iowa may not be winning in the manner that Mr. Fowler wants, but that’s somewhat beside the point.

Somewhat beside the point?!  How so?!  BCS has acknowledged that there’s not enough sample size to determine who’s “most deserving” based on wins and losses alone, yet quality of victory is “somewhat beside the point”?!  Are Temple and Idaho better than FSU and VaTech?  No, you’ll probably point to strength of schedule, so how about this: are Temple and Idaho better than Central Michigan and Fresno State? Wins and losses aside, it’s an easy NO.  Okay, continue:

The BCS Standings are not trying to mirror Las Vegas in projecting who would win in a (hypothetical) future matchup. They are trying to reward the most deserving teams.  In terms of accomplishments on the field, to date, Iowa has achieved quite a lot.  The Hawkeyes are the only team in the BCS top-10 that has beaten two current BCS top-25 teams on the road, and they’ve also beaten #18 (in the BCS) Arizona….There’s no denying their success to date.

Okay, if the BCS are trying to “reward the most deserving teams”, then why is a team that struggles to defeat Northern Iowa and Arkansas State at home in second place, according to the computers?   We’re talking the team that needed to beat Michigan State on the last play in one game and a 28-point fourth quarter at home against Indiana in another.  Yes, yes there is denying their success, BECAUSE IOWA HAS BEEN AVERAGE IN TWO-THIRDS OF THEIR GAMES.  Hey, what about Oregon?  The Ducks are 3-0 against top 25 teams, including that killing of USC, yet Iowa is five spots ahead of Oregon.  Go figure.  Next:

Unquestionably, including margin of victory in the computers would encourage teams to run up the score in an unsportsmanlike way.  If teams instead showed good sportsmanship, they would be penalized.

Okay… sure.  Eliminating margin of victory from the BCS has definitely helped sportsmanship, seeing as how 14 games were decided by 24 points or more.  My school went through a 254-41 stretch in four games, and nobody called our radio station calling the football team classless.  However, if that’s not good enough, I’ll show a great solution, one that won’t penalize consistent, efficient teams like the BCS does.  Here’s how Anderson closes:

Far better to reward teams for winning against good competition and leave the style points to diving and gymnastics.

This deserves the award for “FAIL Comment of the Year”.  Hey Bozo, guess what other sport uses “style points”?  COLLEGE BASKETBALL.  Guess which sport has the most popular tournament? COLLEGE BASKETBALL.  I’ll bury the Anderson grave by saying this: far better to reward the best teams and leave the “beating good competition” argument to the morons.

Richard Billingsley ratings
visit his ratings at College Football Research Center
After finally finding Billingsley’s explanation, I give him credit for being the most articulate with his system.  He’s all fine and dandy in my book (besides missing any sort of qualitative stats), but then closes with this sentiment.

It’s no secret that I’m a big fan of the BCS….Thanks to the BCS we no longer have to deal with “mythical” national championships. The BCS is not standing in the way of a playoff. The regular season in college football is a playoff.

Sorry, Richard, but this is a still “mythical” championship the BCS has, because the system doesn’t devote its resources to determining the best teams, without any qualitative stats and all.  You mean well, and you realize the BCS is flawed, but you don’t realize the true flaw of the BCS.  One that still makes the championship game “mythical”.

Colley Matrix Ratings
visit the ratings at the Colley Matrix website
Back to poking pancake-sized holes to this computer’s claims.  Colley tries to persuade viewers, in his abstract circa 2002, mind you, that his matrix achieves seven goals.  Just to name two, the matrix:

3. uses a minimum of assumptions,
7. produces common sense results.

First, Colley is using the assumptions that having a margin of victory is “running up the score” and that “running up the score” is “unsportsmanlike” and that “unsportsmanlike” behavior is a more important factor than finding the best teams.  Second, by having such assumptions, Colley is ignoring the common sense that some sort of qualitative factors (besides strength of schedule) needs to be used to determine the best teams.  After all, what determines who has the strongest schedule?  Win-loss record?  Well, if so, it simply builds a web of misguided SOS figures skewed by a small sample size.  Next victim.

Massey Ratings
visit his ratings at his website
Nice try.  You try to eliminate the random factors of the game, yet… you guessed it.  No qualitative stats.  Score.  Venue. Date.  That’s it.  And I bet you “score” = win or loss.  Massey claims to have a goal of differentiating a win, a big win and a loss, but he can’t do that with only the score, venue and date.  Massey at least tries to eliminate the random factors, he needs to get in this decade with his football philosophy (the version he uses was last updated in 1999).

Dr. Peter Wolfe Ratings
visit his ratings at his website
Yet another claim of a “reasonable system” without any qualitative stats.  Wolfe fails to provide that, or an example as to how his “formula” works.  There just seems to be no substance, and it’s getting tiresome pointing out the flaws here.

Conclusion: What should be the solution?
Let’s face it, BCS, your system is worse now that margin of victory is completely excluded.  You made steps to include the smaller schools in the fight, but you still fail to find the best teams.  You have “computer” systems that nothing more than “brains” that either have no clue about how quality football teams play or no chance to make a quality system thanks to your anti-MOV rules.

My solution, a system that includes all of the above:

1. Quality stats, such as passing efficiency differential, non-garbage margin of victory, turnover margin, yards per attempt differential, third down differential, etc.
2. A drive-by-drive ratings system: similar to the FEI ratings
3. A possession-based ratings system: the time led by two possessions has a higher value than the time same led by one possession (the max value would be four possessions, so any lead by 25+ points would be equal)
4. Win-loss record and strength of schedule factored as before
5. NO influence by the polls

To weed out any doubt, yes I DO believe that this system would be much better than the current broken BCS system.

In ELO-CHESS, only winning and losing matters; the score margin is of no consequence,
which makes it very "politically correct".  However it is less accurate in its predictions for
upcoming games than is the PURE POINTS, in which the score margin is the only thing that matters.
PURE POINTS is also known as PREDICTOR, BALLANTINE, RHEINGOLD, WHITE OWL and is the best single PREDICTOR
of future games.  The ELO-CHESS will be utilized by the Bowl Championship Series(BCS).

Advertisement