RATIONAL for CSL RANKINGS
In 2001, CSL_Ratings will rate NCAA Division IA, IAA, II, III, and NAIA college teams. The following paragraphs describe the assumptions and rationale for the CSL_Ratings.
The primary assumption in the CSL_Ratings is that winning isn't the most important thing, winning is the only thing. Margin of victory is immaterial in the CSL_Ratings.
The second assumption is that wins against quality opponents should be rewarded more than wins against lesser opponents, and that losses to quality opponents should not hurt as much as losses to lesser opponents.
The third assumption is that all teams within a division are initially equal, but not all divisions are equal. Since inter-division games are played, a means of weighting the results of those games is necessary. As such, a win by a better division team over a lesser division team will not count as much a win over a same (or better) division team. The amount of reduction will be determined by the results of the interdivision matchups. For example, if D1A has a 75% winning percentage against D1AA teams, a D1A win over a D1AA opponent is worth 75% of a win over another D1A opponent. Losses by higher division teams to lower division teams will be treated the same as any other loss. Similarly, lower division losses to higher division teams will not count as heavily against the lower division team.
A fourth assumption is that all games should count. The NCAA allows all teams to play up to 11 non-exempt games. The exempt games (pre-season "classics", road games in Hawaii, conference championships, and bowl games) allow some teams to play more than 11 games. Since each game carries the risk of a loss, each game should also carry the reward of a win. As such, the CSL_Ratings sum wins and losses rather than dealing in percentages.
The CSL_Ratings are computed based on three components.
The first component is a team's own performance in terms of wins and losses. This component is quantified simply by subtracting a team's losses from its wins.
The second component is a measure of a team's schedule strength. I am not trying to establish which team played the toughest schedule, per se. I am trying to rate teams based on the quality of the teams they beat and lose to. Losing to good teams does not mean that you are a good team, so pure schedule strength calculation is meaningless. Each win is a positive, and each loss is a negative. Specifically, Team A gets one positive point for each win accumulated by the teams that Team A beat. Team A also gets one negative point for each loss accumulated by the teams that beat Team A. Team A is neither rewarded for losing to a really good team nor penalized for beating a really bad team. However, Team A is not penalized severely for losing to a very good team but does take a serious hit it loses to a bad team. Similarly, Team A gets very little reward for beating a bad team, but much reward for beating a good team.
The third component is much like the second, but looks at a team's opponents' schedule strength. I get 1 point for each win by teams beaten by the teams I beat and I lose 1 point for each loss of teams that beat teams that beat me.
After the three components are calculated, each is normalized to ensure each is given equal weighting. The three components are then summed to give each team its rating. This is the number that appears in the CSL_Ratings. As such, the rating of any team is simply a relative strength indication, and has absolutely no relevance to projected point differentials between the two teams. It is not recommended that the CSL_Ratings be used to handicap games based on the relative rating value of the teams involved. The CSL_Ratings are reflective of a teams past performance and should not be considered predictive. Having said that, past versions of the CSL_Ratings have correctly predicted the winner in over 70% of games.
The CSL_Ratings are based solely on a team's performance in the current year. As such, early season ratings are largely meaningless. I use previous year results and current year schedule as a filter early in the season, just to disguise the fact that the early season ratings were worthless. The previous year record component is phased out over the first month of the season. Many of the computer-aided systems will not publish until the season is half over. Perhaps I should follow their lead, but I will publish the filtered results just to emphasize the lunacy of early season polls. When the journalists admit they know nothing during the first five weeks of the season and withhold their first polls until October, I will do likewise.
The CSL_Ratings will be posted for all divisions as soon as possible within the constraints of my time. I will strive to post NCAA D1A ratings no later than each Sunday Morning (those games in Hawaii really make it tough to post on Saturday night). The D1A ratings will include results from all games involving D1A teams played to date and all other divisions at least thru the previous week.
Ratings for D1AA will be posted as soon as possible following the D1A ratings. D1AA ratings will include all results from games involving D1AA and D1A played to date and all other divisions at least thru the previous week.
Ratings for NCAA Division II will be posted as soon as possible following the D1AA ratings, and so on.
While the ratings in a higher division may be affected by completion of ratings in a lower division, the effect will be minimal, especially later in the season.
As mentioned above, I will strive to post promptly each week, but Hunting season does occur during football season, and therefore, I may be tardy on occasion. I apologize for any inconvenience this may cause.
Since I do not have the luxury of unlimited time and resources, I will rely on the following sources.
For division affiliation, I rely on the list posted on the Peter Wolfe’s web site.
I will get game results from Peter Wolfe's web site at http://www.bol.ucla.edu/~prwolfe/cfootball/scores.html.
I extend a big thank you to the folks who maintain these pages.
Any comments on my assumptions are welcome. Please e-mail me at the address listed below.
Email: c_loest@yahoo.com
