The ARGH pre-season power ratings are computed using the following factors, with the strength of each factor varying each year based on what has happened in the three previous years:
Previous Year's Power Rating. This is the single most powerful predictor of the current year's power rating, both because much of the team responsible for the previous year's results will be back and because the previous year's results are an indication in and of themselves of the talent level of the team. This factor is up slightly for 2003.
Second Previous Year's Power Rating. This is used because much of the team will still be around after two years, and because it helps to create a historical context with which to judge a team. In other words, if a good team has an off year, this factor will tend to predict that team to bounce back. This factor is up slightly for 2003.
Number of returning starters. The number of projected starters is a fairly powerful indicator of whether a team will get better or worse. This factor is up significantly for 2003. Figures are from Athlon, The Sporting News, Goldsheet, Phil Steele, and Lindy.
Returning Coach. A team changing coaches tends to get a little worse during the year immediately after the change, presumably due to a period of adjustment to the new coach's system. This factor is up slightly for 2003.
Returning Starting Quarterback. The quarterback is far and away the most important position in terms of experience, and a returning starting quarterback is worth much more than a returning starter at any other position. This factor is down slightly for 2003.
Previous year's strength of schedule. This factor is supposed to work under the theory that teams tend to play up or down to the level of their competition. This factor is up slightly in 2003.
Recruiting. Analysis has consistently shown a small but reasonably significant positive correlation between recruiting and a team's preformance in the next season. The recruiting factor is up significantly in 2003. The final factors come from Athlon, Goldsheet, Phil Steele, and Lindy.
In actual usage, this factor will probably serve to keep teams and conferences somewhere near the level they've played at in previous years. The model I'm using isn't going to vault a team with a #1 recruiting ranking over a team with a #2 recruiting ranking unless they were very, very close to begin with, but it will usually jump a team ranked in the top 50 recruiting classes over a team that isn't from four or five places down.
Factors that are NOT used to calculate the ratings are:
Previous years' power ratings beyond two years. There's no evidence that these are statistically significant to the current year's power rankings.
Turnover Margin. Phil Steele is big on this one, so I decided to test it this year. The results of the test were that the previous year's turnover margin are utterly statistically insignificant in any combination with the factors I actually use. Sorry, Phil -- I love your work, but I think you're finding something more akin to the bounceback provided by the second previous year's power rating.
Current year's strength of schedule. Almost every year some team is touted for the national championship or at least a great year because their schedule is so weak. See Kansas State, North Carolina State, Oklahoma State, etc... Every year that team loses, usually more than one game, and usually early. Having a weak schedule does not make a team strong, and having a tough schedule does not make a team weak. These power ratings are meant to measure how good a team is, not to measure a team's ability to beat patsies.
Individual player talent. I don't pretend to know how to judge anything like this, and even the people who do pretend to do so are usually subject to either their own biases or a lack of context. Talent usually manifests itself in a team's record for the previous year, or possibly in a team's recruiting, and both of those are already factored in mathematically.
Because I like them. Cal, my alma mater, is ranked dead last among Pac-10 teams in the 2003 preliminary preseason rankings.
Because I dislike them. I have zero love for the Golden Dome, but I was one of the few who put Notre Dame in the top 25 last year, and my prediction for them turned out to be pessimistic. My preliminary preseason ranking for Notre Dame this year was very low, but with their usual granting of extra years of eligibility to several returning starters, they climbed back into the top 25 in the final preseason rankings. (Geez, guys, can't you redshirt players like everyone else instead of going through this pretense? It makes doing predictions for you in February damn near impossible! ;)) However Stanford is still fairly low in the pre-season rankings this year, although unfortunately not as low as Cal. I got plenty of nasty mail from Stanford types who disliked my preliminary pick of the Trees for last in the Pac-10. After a 2-9 season, I have to wonder how crow tastes on the Farm...
Because you like them. These ratings are strictly mathematical, and an e-mail that says something like
"Your ratings suck! I'm a 1977 alumni of Wild Card U., and in my unbiased opinion, my Jokers are going to go all the way!! Go Jokers!!!!!!"
--myopicphan@wildcardu.edu
won't change the formulas or the ratings.
This page last modified July 4, 2003.
Click here for a description of how the ARGH Power Ratings work.
Click here to return to the ARGH home page