A primer to the new BCS formula ...
Posted: Thursday July 15, 2004 2:12PM; Updated: Thursday July 15, 2004 2:38PM
Bowl Championship Series officials on Thursday unveiled what they're hailing as a "simpler and more precise" ratings system. So simple and precise, in fact, that I'm sure the following chart probably makes complete sense the first time you read it:
That's what I thought.
But give the BCS credit for one thing. In the past, officials would have responded to a crisis like last season's USC/LSU/Oklahoma controversy by adding yet another component to their convoluted formula. This time, they went for a complete makeover, essentially scrapping the model they used for the BCS' first six seasons and devising a new one. Here is a primer to the new formula using the 2003 season as an example:
In the old formula, there were four equally weighted components, each comprising 25 percent of a team's total score: losses, schedule strength, computer ratings and average ranking in the AP and Coaches' polls. There was also a bonus for "quality wins" tacked on to the final total. The reason Oklahoma finished the regular season ranked No. 1 in the BCS rankings -- despite USC and LSU jumping ahead of them in the AP and Coaches' polls -- was because the Sooners led the other two teams in three other categories: schedule strength, computer ratings and quality wins.
In the new formula, however, schedule strength, losses and quality wins go out the window. We're left with three, equally weighted components: The AP poll, the Coaches' poll and the computer ratings. The two human polls, therefore, go from representing a combined 25 percent of the old formula to 66 percent of the new formula, making it awfully hard for the No. 1 team in the polls to get left out of the BCS title game, which happened to USC last year.
Furthermore, the BCS no longer will use the team's actual ranking in each poll (No. 1, No. 2, etc.), but instead the percentage of possible points it received from the voters. For instance, if USC had received first-place votes from all 65 AP voters -- each worth 25 points -- it would have attained the maximum possible 1,625 points. The Trojans' actual points were 1,580, or 97.2 percent of the possible total. In the Coaches' poll, they received 1,516 of a possible 1,575 points, or 96.3 percent.
For the third component, the computer ratings, the BCS is using six ratings systems (Sagarin, Anderson and Hester, Billingsley, Colley, Massey and Wolfe) but dropping each team's highest and lowest score, so it's actually four. It then assigns 25 points for a first-place ranking, 24 for second, etc. After dropping their highest and lowest score, the Trojans are left with four third-place finishes, for 23 points each. That makes for 92 points out of a possible 100, or 92 percent.
USC's BCS total is then calculated by averaging the three percentages -- AP (97.2 percent), Coaches' (96.3) and computers (92) -- for a final score of 96 percent, or, as the BCS is going to list it, .960.
So now let's revisit that earlier chart, which applies the new formula to last season. Across the top are the three components, AP, Coaches' and computers, with their maximum possible points. Then the top three teams are listed with their placement in each category. The highest and lowest computer rating for each team is crossed through.
As you can see, LSU and USC, not Oklahoma, would have met in the Sugar Bowl. However, it's clear that under this new system the difference between No. 2 and No. 3 could be a matter of mere percentage points (in this case .014), to the point where one individual voter's ballot in the AP or Coaches' poll could theoretically dictate the national championship.
Stewart Mandel covers college sports for SI.com.