KCFR Q&A: What is the intent of these ratings?

     KCFR intends to predict and rank college football as accurately as
possible using cumulative game statistics.  The Power Ratings display a
team’s current relative strength and is primarily PREDICTIVE.  The Power
Ratings employ mostly the Net Yards Difference to compute game results.
The Won-Lost Ratings are BCS style and rates the team's record against
their schedule.  The Won-Lost Rating is RETRODICTIVE.  The KCFR Ranking
method combines Power Rating and Won-Lost Rating, Power Rating having
the heaviest impact early in the season, and Won-Lost Rating the most
impact at the end of the season.  

     Strength-of-Schedule is the most important consideration when
ranking college football teams.  It is this aspect which elicits the
most 'bias' among rating systems.  Won-Lost is sufficient to rank 700
teams at once using a direct non-biased formula.  However, this method
does not relate well to the Top 25 or Top 10.  In Division 1-A, the
'mid-major' conferences receive too much credit when rating so many
teams from an equal starting point.  This method is approved by many,
including the BCS, in the interest of 'fairness'.  In the real world
of college football the teams do not start out equal.  The bigger
programs have bigger budgets, better facilities, higher-paid coaches,
often more scholarships, and the best athletes.  The 119 teams in
Division 1-A do not start out as equals.  If you want to wait until
the middle of the season to rank the teams after a sufficient number
of games, good for you.  But wins and losses from the same point of
origin is not accurate enough.
     KCFR handicaps the conferences prior to the season, that is, each
conference is assigned a pool rating close to its overall strength.
The pool ratings are confirmed by studying real head-to-head results,
using won-lost records, scoring margins, and yardage margins.  These
are reviewed every year during the off-season to help ensure accuracy.
If this is construed to be a 'bias' then so be it.  All rating systems
are defined by a particular bias when one decides what data should be
input.  Show me a perfect rating system and I'll bet you it takes into
account every conceivable piece of data and information available.
KCFR hopes to rate teams and conferences accurately based on actual 
on-the-field results, not theoretical sciences.  The fairness issue is
one that is highly overthought and constrains the effectiveness of a
computer rating system.  

     Actual game results (primarily Net Yards Margin and no scores)
are converted to a realistic point scale approx -25 to +25, and this then
is added to the opponent's power rating to measure its game performance.
Each game performance is measured separately throughout the schedule
with each most recent game given heavier weight than the previous.  The
opponent's rating is initially the pool rating, becoming its power rating
after the first game, and is continually updated for each game result as
the schedule is played.  All of these results are played through the
computer schedule many times until a consistent result is achieved.
The new Power Rating is then posted and used for predicting the next game.
     At the start of the season, pre-season Power Ratings are established
to make predictions.  These ratings are produced by checking changes in
the rosters, removing their statistics from the team, and adjusting plus
or minus from the norm against the team's final power rating.  Pre-season
ratings have strong weight in the early-season Power Ratings and then are
gradually dissolved from the formula by the team's fifth game and have no
effect whatsoever on the ratings after that.  This helps the predictive
purpose of the Power Ratings early in the season.  There are no talent
or coaching evaluations, roster depth or match-up data, or any inside
information input into the Power Ratings.  (I already have a full-time
paying occupation, and this is not it.)  

      Since the BCS has eliminated scoring margin from their computers,
wins and losses are their only input.  So, KCFR also has a Won-Lost
Rating, BCS style, to compare with the BCS's computers.  It also has a
portion of the KCFR Team Ranking, building to a 2-1 margin over the
Power Rating by the end of the season.  The Won-Lost Rating computes
the team's winning percentage against the strength-of-schedule equal
for all games.  Basically, it is +25 for a win (-25 for a loss) added
to the opponent's Power Rating.  This too is filtered through the
computer schedule until a constant number is attained.  The Won-Lost
Rating is primarily retrodictive.  

     Besides the current Power Rating, each team is assigned a program
strength rating and a home adjustment rating, and both are unchanged
throughout the season.  The program rating assigns points (usually 2 to
6) indicating the historical strength of the team based on attendance,
winning tradition and home field records.  The home adjustment rating
further tweaks the advantage up or down based on the team's record
against the pointspread for the last four seasons.  First, the teams'
current Power Ratings are compared, then both teams' program strength
(which instantly produces an inherent home advantage).  Then only the
home team's home adjustment is added in.  The total difference is the
KCFR game pointspread.  On the KCFR 1-A prediction page, any game which
is figured nine or more points different from the Sheridan Line in USA
Today will be designated as a Best Bet, with the favored team vs the
line marked with '#'.  

     KCFR's methods are in constant review to attain the best accuracy
possible.  If it's good enough for the BCS it's good enough for me.
Changes are not made from knee-jerk reaction or unsatisfactory results.
KCFR has evolved from 30+ years of studying college football trends and
has always looked for a better way to get accurate results.  During the
off-season, cumulative statistics are compared and reviewed to reveal
the latest trends relative to conference strengths and turnover of team
rosters.  It is a lot of research and hard work by studying a lot of
data.  It does not involve a singular ingenuous equation.  I am but a
humble user of basic algebra and spreadsheets, not a math professor.

     In the game of college football, the only really important stat is
the win.  In rating college football, the important thing is accuracy.
And to achieve accuracy, the more data input, the better chance for an
accurate readout.  Any statistics major knows that.  When the BCS 
committee decreed that their computers do without scoring margin input,
I looked for an alternative source and just wanted to be different.
The more I looked at the Net Yards Difference from game statistics, the
more I realized that it displayed a truer picture of game performance.
Net Yards is more indicative of one team's dominance over the other,
on the field of play, at the line-of-scrimmage.  Final scores are 
often skewed by turnovers and special team plays, both highly unpre-
dictable.  The Net Yards Difference is usually the widest margin of
statistical info and provides the greatest sampling range in which
to produce a more exact measurement.  The average margin of victory
is around 18 points, but the average margin of net yardage is about
132 yards.  While it is true that sometimes the winning team does not
finish with the most yardage, it is a very telling statistic about
the competitive story of the game.  Now that we have the Internet,
cumulative game statistics are available from every Division 1-A and
1-AA game soon after completion.  The reliance upon final scores and
won-lost record for accurate ratings is, in my opinion, not enough.

     Yes.  But hey, if you have read this far, then you are just as
big of a nut about college football as I am.  I believe it enhances
my viewing and following of college football.  And besides, I now
have a ridiculously cheap hobby which I really love, and it is also
an honor to be among a select body of ratings providers for such a
great sport.  I hope to someday be associated with the best of them.
     KCFR is published strictly for entertainment purposes, in the
public domain of the World Wide Web, forever free of charge and
without advertising.  KCFR does not endorse or engage in any form of
wagering or gambling, and will not be held liable by any viewer who
uses this information, misinformation, or disinformation resulting in
monetary losses or personal embarrassment, so there!  Any reprinting,
retransmission or rebroadcast of any or all parts of KCFR without the
expressed written or oral consent of KCFR is absolutely positively
okay fine by me.  September is my favorite month, so let's get going!
Hope that all of you enjoy the 2007 football season, as well as this
webpage, and thanks for clicking on.