If I Had a Computer Poll
Introducting the Wilson Dominance Index.
by James Leroy Wilson
December 12, 2001
If I Had a Computer Poll_James Leroy Wilson-Introducting the Wilson Dominance Index.
(Authors note: the following formula I propose here is, to the best of my knowledge, original with me. I do not know the methodology or criteria of the various computer rankings, and as far as I know none has been circulated widely. Also note that I wrote a few weeks ago about the Bowl Championship Series. Check the archives. This column is about my ideal computer ranking system if we were to have the BCS, not support for the BCS.)
The question for college football fans these days is, "What were those computers thinking?" Obviously, they weren't thinking anything. They just computed based on data fed into them, and along with other data (schedule strength, number of losses, "quality victories") used to compute the Bowl Championship Series (BCS) standings, helped determine that Nebraska is the #2 team in the nation, behind Miami (Fla). As a result, the Cornhuskers,#4 in both the coaches and the sportswriters polls, will face the Hurricanes in the Rose Bowl. Miami (Fla), fortunately, is also #1 in those polls. So the computers did at least one thing right, we suppose.
I do not know how the designers of the computer polls (there were actually eight computers) decided to objectify the quality of the teams. I am not a computer programmer myself with ready access to data, but I think I can a construct superior "objective" system to determine the top two teams in the nation, and this system could be combined with the sportswriters poll, which tends to favor strength of schedule, and the coaches' poll, which looks at number of losses.
The principles behind the scheme include: Dominance over the entire schedule; Holding teams accountable for their non-conference schedule, and to dominant performances in the face of weak schedules; Not rewarding performance in "extra" games such as pre-season classics or conference championships, but only on the set 11-game schedule all Division I-A teams play. This makes judging strength of schedule fair to all, including those who do not have opportunity to play extra games.
The principle of measuring dominance does not depend on margin of victory, yard differential, or any other common statistic. Rather, it depends three things:
1.Winning the game;
2.Maintaining the "final" lead;
3.Maintaining a "controlling final" lead.
The goal is to decide in each game, not only who won, but whether or not one team was actually "better," and for how long. This doesn't mean the sum of the length of times throughout the game that one team happened to hold the lead, but rather the point in which the victor took the lead and never relinquished it. If a team takes a 7-0 lead very early in the game, and the other team never took the lead or tied after that even though the final score turned out to be 49-48, I believe that should count for something, because at no point in that game did the opponent do anything to indicate it was a better team.
But of course, such games may still be close and rely on, say, poor officiating calls to preserve victory for the team that maintained a close lead for most of the game. To control for that possibility, the superior team should not just maintain the lead but hold a controlling lead, by which I mean a lead of two or more scores. The most "one score" can get in football is eight points, so by two scores I mean nine.
In football, as in hockey and soccer, a score can happen any time, on any play, no matter how fluky. The advantage of a controlling lead - two scores - is obvious: the team behind can make such a play yet still be losing. The key for the winning team is to take such a lead and build on it.
A commanding lead gives the team ahead more options, and reduces them for the team behind, especially as the clock winds down. Yet, in the interest of sportsmanship, rest for the first-string, or other reasons, it is legitimate, as the clock winds down and the lead becomes insurmountable, that the team ahead may sacrifice yards and points as it plays inferior players on its roster. It is wrong to punish this action, providing the controlling lead is maintained until the end of the game. That is why we shouldn't reward going ahead by three (17 points), four (25) or more scores.
So a team that not only wins the game, but also holds the winning go-ahead lead for a long time, and held that by at least two scores for a long time, should be awarded with a convincing victory that removes doubt about its superiority.
To quantify the ability to win, hold leads, and hold commanding leads, we must look at minutes in the lead, not margin of victory. Therefore I propose the Wilson Dominance Index (WDI). It would award points to teams in the following ways: For each victory, award 60 points (representing minutes). In each victory, add additional points (minutes) for how long (counting from the final gun backward) the victor held the final lead. If the victory was achieved only at the final gun, award zero; if the go-ahead score was in the first minute of the game, award 59. In each victory by nine or more points, add additional points according to how many minutes such a lead was maintained (throwing out last-second scores by the losing team in the closing seconds, which could not have affected victory).
For example, say a team takes a 13-10 eight minutes left in the third quarter, and takes a 20-10 lead with seven minutes left in the fourth, and wins the game by that margin. It would be awarded sixty points for the win, 22 minutes for holding the winning lead for that long, and another seven for holding a controlling lead for that long. The victory would count for 89 points.
But the index isn't done; what counts for victory should also count equally for defeat. The WDI takes into account performance in losses, determining how close or convincing such losses were. Therefore:
For each loss, the sixty points "lost" are only those that could have been gained. Therefore, a losing team neither gains nor loses anything just by losing. Inability to take leads would be subtracted equally as ability to hold them rewarded. For each loss, each minute from the final gun backward, during which the team was behind, will be deducted. Inability to make games close will be punished. Each minute a team was unable to penetrate a commanding two-score lead from the final gun back will be deducted. In the above example, the additional points (29) the winner achieved would be deducted from the loser.
In another example; say a team falls behind 7-0 with thirteen minutes left in the first quarter and 14-0 with ten left, would fall behind 35-3 in the second, come back to a margin of 42-30 in the third, yet that 12-point margin is as close as it got as it loses 62-36. In that case, it loses nothing for losing, but sacrifices 58 for being behind that long, and another 50 losing by a controlling, two-score lead for that long. That means losing 108 points.
To lose a game at the final gun or in overtime neither gains or loses a team anything, but does award victors in such games sixty. Drubbing teams early and often would gain teams points, being drubbed would lose a lot.
But the WDI goes one step further; it takes strength of schedule into consideration. It does this by simple logic; the weaker the schedule, the more winning points, leading points, and controlling points should be amassed. The more difficult the schedule, the harder these are to come by. So, after totaling the "raw WDI" of the total points of victory, leading and controlling points minus "behind" points and "controlling behind" points, the final WDI would add the average points of the raw WDI scores of their opponents.
Victories in "extra" games - 12th and 13th games, will not count because not every team gets the opportunity to play them. Neither will they be judged in schedule strength. Losses in those games, however, will count, with 60 points deducted for each regardless of the margin. We don't want a team that loses the Kickoff Classic and the SEC Championship game yet won the eleven games in between to play in the national championship game if that can be avoided.
This system balances won-loss record with schedule strength, while measuring superiority without rewarding "rolling up the score." They keys to finding the top two teams in the nation are all here. If superior, dominate by building a controlling lead and winning. If better or equal, just win. If inferior, do everything to keep the game as close as possible, because closeness might result in victory. The WDI is designed to recognize and award championship-caliber football every week, no matter the opponent, while respecting sportsmanlike play. If we are to have the BCS, the WDI should be at least some part of the equation.
Teachings of a Three Year Old... Turned Tyke,
by Hal Evan Caplan.
A father learns from the wisdom of his toddler.