There’s been a lot of buzz about Oregon sliding the last two weeks, and fans are decrying the system. It almost makes me feel nostalgic for 2001 and 2005. Unfortunately, a lot of these cries don’t seem to actually hold an understanding of how the system works. It’s actually extremely simple to understand.
There are basics everyone knows. The top two teams in the BCS rankings play each other. Everyone also seems to be aware that there are three primary components, with equal value:
2 Human Polls (The USA Today Coaches Poll, and the Harris Poll), and 1 Computer Poll.
The nitty gritty details from these polls seem to be amiss however. All polls, including the Computer Poll, are aggregated into a score determined from ballots entries.
Let's start with the worst of the 3 polls.
The USA Today Coaches Poll.
Why is it the worst? The coaches don't pay much attention to college football outside of their own team and opponents. In fact, they have so little time to actually participate, many coaches just leave it to their assistants to fill out their ballots. The next issue is that the ballots are unpublished (Unlike the Associated Press Poll). You can find out who the voters are, but not how they each voted, except the season's final ballots. This makes the voters unaccountable for their ballots.
The ballots are submitted (For 2012, there are 59 voters) and 25 points are given to the first place team, 24 for the 2nd place 23 for 3rd all the way until 1 point for 25th. Each team's "BCS Score" is totaled up, and then divided by the maximum points (1475 for the USA Today Poll). Here are a couple examples:
Alabama has every 1st place vote. This gives them 1475 total points (25 points, 59 times; 25*59=1475). Their BCS score is 1.000 (1475 total vote points divided by 1475 total points; 1475/1475=1.000).
Oregon has 1329 points (The 2nd most). Their BCS score is 0.901 (1329/1475=0.901).
The actual ranking is determined, in descending order, by who has the most points.
The Harris Interactive College Football Poll
This poll is actually ran by Harris Interactive, a Market Research firm. They replaced the AP poll in 2005 after the Associated Press protested the BCS system, and didn't want to be included in the formula any more.
Harris Interactive is primarily known for their polls that cover a wide variety of topics. However, the voting participants for the College Football Poll aren't Market Researchers. It includes 115 former coaches, players, administrators and current and former media members. This poll actually suffers from some of the same issues as the USA Today Coaches Poll. You can find out who the voters are, but they don't publish any ballots. In fact, unlike the Coaches Poll, the Harris Poll voters don't even publish their final ballot. However, the people comprised of the poll are people that would have the time and interest to actually research and follow the entire scene of college football. Another thing that it actually does that is an improvement upon the Coaches Poll is that it doesn't receive ballots until 6 weeks of college football have been completed. So before assuming which team is the best, it actually looks at a total body of work.
The Harris BCS score is calculated in a similar method as USA today. The only real difference between the two is that there are 115 votes instead of 59 votes. So the maximum points is 2875. We'll use the same teams as an example.
Alabama is actually NOT a consensus #1 in the Harris, with 5 first place going to Oregon, and 1 first place vote going to Florida. However, Alabama still is receiving the other 109 first place votes, and is #1 because they have the most total points, with 2868. Their BCS score is 0.997 (2868/2875=0.9976)
Oregon has 2727 points (Again, the 2nd most) in the Harris Poll, which gives them a score of 0.948 (2727/2875=0.948).
Before moving on, let's be clear about what this means. 2/3rd of the BCS poll loves Oregon a lot. Despite sliding 2 spots over the last 2 weeks, Oregon is still very popular. If they continue to play well, the other factors will take care of themselves. Saying the BCS is trying to team up two SEC teams right now is a statement of ignorance. No other factors within the BCS have actually changed.
The Computer Poll
For the Computer Poll, there are very few aspects in common with the human polls. Instead of 50+ or 100+ ballots, there are 6. Instead of being able to change methodology, agendas or "going by the gut" throughout the year, the computers apply a consistent calculation for their results.
There is a significantly larger amount of mystery when it comes to the computer poll. The computer formulas themselves are unknown to the public. There are a few sparse statements from those that run the computer polls about the methodology, and some, like Sagarin are used often as a baseline for a team's strength of schedule (Partially because Sagarin actually ranks all Division I teams. So he includes FCS teams with FBS teams. Massey actually includes Division I, II and III teams, over 700 total teams). What is known is that the computer polls are no longer allowed to use Margin of Victory (That ended after 2004).
Getting to know your college football computer system.
(Note: A lot of this is found on Wikipedia, so take that either as a source or a warning):
Anderson & Hester (A&H)
A&H has always used won-loss records and never used Margin of Victory. It's important to note that it's never used the win-loss record. Unlike some other computer systems, it hasn't had to adjust it's methodology from it's inception. A&H judges the strength of schedule through the opponents and the opponents' opponents win-loss record as well as the conferences strength. The conference strength is calculated by a win-loss record and the difficulty of the non-conference schedule. It also doesn't do anything until week 5 of the season. It doesn't use a previous season's rankings unlike other computer systems.
RB's factor is Win-Loss, the strength of the opponent and has a stronger emphasis on recent games. It use to factor in the margin of victory. It also has a little bit of influence from defensive scoring and where the game's location is. The actual values of these components are unknown.
Edit: As pointed out by AtQ contributor TheOcean, RB uses previous year's rankings as a starting point, regardless of the changes to the team. Billingsley rationalizes this by claiming it's provided a more accurate measurement of strength of schedule.
Colley Matrix (CM)
Like A&H, CM doesn't use any pre-season polling, but it does start it's polling early in the season when it treats all teams as equals. It's formula has changed a little bit over the years. It use to ignore games against FCS teams, but when the college football season expanded to 12 games, it started to include them, but with very minor influences. It calculates it's own strength of schedule, and puts a heavy emphasis on it.
There are a few factors that go into the KM system. There are essentially two main factors: A Massey rating and a power ranking. The Massey rating is score, venue, and date of each game. Then the power ranking is a formula that combines offensive ratings and defensive ratings. That makes an overall team rating after applying a win-loss correction.
Sagarin's BCS computer system uses a pretty common system called "Elo Rating System". This is probably the most transparent system, because it's a formula used in several other performance rating systems. Elo Rating was actually developed by Physicist & Chess Player, Arpad Elo. It was used to measure a player's performance rank him against several other players. It's used in many other sports in addition to Chess and College Football. The basic application is that a team has a performance score. This also includes expectations against other opponents. If a team actually does better than their expectations, their ratings improve. JS also includes his predictive system, which he openly prefers and says is more accurate, but it includes the Margin of Victory, so it cannot be used in the BCS systems.
PW also uses an established system, like Sagarin. Only it is the Pairwise comparison. It includes a lot of transitive logic, with a few other factors, such as location.
All of these computer systems, with their own criteria, essentially creates 6 ballots. But because of the small sample size, the highest and the lowest computer results are thrown out.
So what does it mean? It means that 2/3rd of the computer systems have 1/3rd of the power in the BCS system. It also means that the most favorable and least favorable computer results are non-factors.
So, once again, let's look at our two team examples.
Alabama is getting 5, 1, 4, 5, 3, and 3 in the computers. A 5 and the 1 are getting thrown out, leaving them with 4, 5,3 and 3. Just like the human polls, 1st place is worth 25 points, 25th place is worth 1 point. This gives Alabama 89 points. This is also divided against the maximum amount of points possible. (4 first place votes at 25 points = 4*25=100). So Alabama gets .890 points.
Oregon is getting 6, 2, 6, 6, 7, and 7 in the computers. The 2 and a 7 are thrown out, leaving them with 6, 6, 6, and 7. That's 79 points. That's going to be a 0.79 points.
All three polls are then averaged out to get the final BCS score.
Alabama: (1.000+0.9976+0.890)/3 = .9625
Oregon: (0.9512+0.9485+0.790)/3 = .8966
So what's the bottom line?
There isn't any great conspiracy. There isn't any great BCS bias. Just because Oregon isn't sitting at #2 doesn't mean it's anything to worry about. The computers don't give any preferential treatment to any specific team or conference to tilt the favor one way or another. The simple fact is Oregon has not had a strong schedule yet. But if they win out, the computers will swing as it will be wins against good competition. They'll probably pick up a few additional human votes along the way, but the impact of those are going to be minimal.
Are the computers "stupid"? It's a reasonable debate, you could even have a good debate that they aren't stupid, but they have too much influence. But they aren't everything. 2/3rds of the factor is behind Oregon. The important pieces are in play. If Oregon only worries about what they are doing, they'll be fine.
The BCS is getting it right for where the season is at, and how it has gone. But really, wouldn't this all be so much simpler if Bill Snyder didn't cancel the K-State game this year? Shouldn't the point that Bill Snyder canceled his scheduled game against Oregon be emphasized more? I think so.