|
Dan Curry Ranking
Informational Article
Sometime in the last ten years, “margin of
victory” became “verboten.” The college football BCS poobahs went so far as
to ban the element from their controversial “national championship”
formula. The critics claim “margin of victory” in no way can help
distinguish which teams are better than others. Ladies and gentlemen, the
critics are wrong.
Isn’t it fairly obvious that if Team A beats
Team Z by 20 points and Team B beats Team Z by 5 points, then Team A is
probably 15 points better than Team B? I use the word probable, because
maybe Team A had a good game or Team B had a bad game or Team Z was sleeping
through the Team A game. But the most probable result is Team A is 15
points better.
Long before I knew how an Excel spreadsheet
worked, I began building rating systems using this basic principle. Over
the years I accumulated large notebooks with game results for every Division
I college basketball team. Three years ago, I discovered a way to have my
computer do the number-crunching. And I settled upon a formula which I
believe can predict future results as good as anything out there.
The system works like this: Each Division I
college basketball team has a rating between 0 and 100. Adding up all 326
teams gives you 16300.00, which averages to 50.00. The formula is designed
like a rubber band. The farther a team gets from the middle, the harder it
is for that team to get farther away and the easier it is for that team to
fall back to the middle.
Say Team 60 (rated 60.00) is playing Team 40
(rated 40.00). The system expects Team 60 to win by 20. If Team 60 wins by
20, no change happens because the teams are placed properly. If Team 60
wins by more than 20, Team 60’s rating will increase and Team 40’s rating
will decrease the same amount of points proportionate to the difference
between the expected margin of victory and the actual margin of victory. If
Team 60 wins by less than 20 or loses, Team 60’s rating will decrease and
Team 40’s rating will increase the same amount of points proportionate to
the difference between the expected margin of victory and the actual margin
of victory.
Over the years I’ve learned that there is at
least one team every year that skews the ratings by running up the score. I
ended that problem by installing a cap of just over 7 points beyond expected
margin of victory. For instance, if Team 60 wins by 28 or more, they get
the maximum point increase. Notice that if Team 40 wins, they are already
20 points beyond expected margin of victory. So Team 40 would get the
maximum point increase.
What is the maximum point increase? It depends
on the relative strength or weakness of your opponent. Team 60’s maximum
point increase will be less than Team 40’s. Team 60 is playing a relatively
weaker opponent, while Team 40 is playing a relatively stronger opponent.
This is the rubber band effect.
We’ve covered what happens if either team wins
by the expected margin of victory or more, but what about when the favorite
wins but by less than the expected margin of victory? The formula is
applied at 25% of its value, giving credence to the fact that the favored
team did win.
Home court advantage is figured in. Teams carry
their ratings from game to game and year to year. Injuries and graduation
can and do hamper the system in the short term, but I have been amazed time
and time again at the ability of the system to adjust.
See for yourself. CHN is publishing my game
predictions and a weekly ranking of all 326 college basketball teams.
Enjoy!
Dan Curry
gauchodan@charter.net
RETURN TO THE CURRY RATING
HOME
|
