jlh@hou2e.UUCP (J.HEATWOLE) (12/18/84)
Several different football ranking and prediction algorithms
have been posted to the net this year. The following algorithm
is somewhat akin to the New York Times ranking in that it uses only
final scores of games and takes into account quality of competition.
It differs in that it counts the final score rather than merely the
win and loss data for each game.
The "Chuck Poll", named after its founder, uses an iterative process
based on the final score of each game to produce a numerical rating
for each team. Essentially, this rating equals ( average pts. per
game - average opponents points per game + average rating of opponents).
Each team starts with 100 points. As an example, the rating would reflect
that Oklahoma soundly defeated Pitt while BYU barely scraped by.
Chuck has kept track of the college scene this year and produced a
final regular season Chuck Poll that may shed some glimmer of light
on the "Who is number 1?" question. I have kept track of the NFL games
this year and extended the algorithm to predict final scores of each
week's games. Following is a discussion of both the pro and college
football final regular season rankings.
College
-------
Note: Diff + Opp rating do not add up to rating exactly because
some adjustment is done to reduce the effect of blowouts and "running
up the score".
Opp
Rank Team W L T Diff Rating Rating
1 Florida 9 1 1 15.55 113.24 128.74
2 Nebraska 9 2 0 23.09 104.15 128.26
3 Fla_St 7 3 1 13.82 109.47 122.55
4 BYU 12 0 0 22.17 100.42 122.03
5 Ohio_St 9 2 0 17.64 105.07 121.89
6 Boston_C 9 2 0 12.36 109.51 121.67
7 Oklahoma 9 1 1 13.91 106.22 121.28
8 Auburn 8 4 0 8.33 113.27 120.70
9 Wash 10 1 0 17.82 102.97 120.66
10 Maryland 8 3 0 9.00 110.09 119.73
11 Okla_St 9 2 0 14.27 105.59 118.97
12 LSU 8 2 1 9.73 109.98 118.95
13 Ga_Tech 6 4 1 8.64 109.20 118.56
14 Miami_Fl 8 4 0 7.50 110.38 118.29
15 South_C 10 1 0 10.91 105.60 118.10
16 Tenn 7 3 1 4.73 113.87 117.55
17 Clemson 7 4 0 11.91 108.22 117.32
18 Georgia 7 4 0 3.00 115.08 116.89
19 Arkansas 7 3 1 10.45 107.47 116.89
20 Illinois 7 4 0 10.36 106.71 116.72
21 Virginia 7 2 2 8.09 106.93 116.62
A few things are of interest. As in the NYT ranking, Florida is number
one. Nebraska is breathing down their throats, about six points better than
the rest of the pack. BYU at number 4 seems to be about .75 point better
than number 7 OU. While BYU had the weakest schedule in the top 30, OU
didn't exactly have it too tough either. Georgia at number 18 had the
toughest schedule in the nation (Clemson, Ga Tech, Mississippi, Vanderbilt,
Memphis State, Southern Mississippi, Alabama, Auburn, Florida,
Kentucky, and South Carolina.)
My big gripe is that the AP and UPI (Chuck calls these the "political polls")
continue to insist number 1 is between OU and BYU. Florida has played
the tougher schedule and has a record identical to OU's. They have done
this despite changing coaches in mid-season. I am no fan of Florida and
especially not of Charley Pell, but they seem to be deserving of the top
spot more than either BYU or OU.
Here are the 2 worst (Div 1) teams of 1984 (ten points behind the rest):
158 Davidson 0 6 0 -26.00 88.97 61.05
157 Columbia 0 9 0 -18.33 79.19 61.97
NFL
---
Based on the algorithm described above, I have devised a method
for predicting each week's games. By using the difference in ratings,
adding points for a home field advantage, and using a similar algorithm
to give a rating for how "hot or cold" a given team is, I have written
code to predict the final score of each NFL game. On the season, I am
at 100-67 (.599) vs. the spread. (Note: I need four weeks of data to
make a prediction, so only games in weeks 5-16 were predicted.)
Here are the final regular season rankings for the NFL.
Opp
Rank Team W L T Rating Diff Rating
1 San Francisco. 15 1 0 97.3 14.1 111.3
2 Miami......... 14 2 0 97.3 12.9 110.2
3 Seattle....... 12 4 0 100.7 7.4 108.1
4 Denver........ 13 3 0 100.7 7.0 107.7
5 LA Raiders.... 11 5 0 101.3 5.6 106.9
6 Washington.... 11 5 0 99.1 7.3 106.3
7 * St. Louis... 9 7 0 100.4 5.5 105.9
8 * Green Bay... 8 8 0 100.0 4.9 104.9
9 Chicago....... 10 6 0 100.0 4.4 104.5
10 Pittsburgh.... 9 7 0 98.5 4.8 103.3
11 LA Rams....... 10 6 0 100.2 2.1 102.3
15 NY Giants..... 9 7 0 101.0 -0.8 100.3
* - did not make playoffs
Diff = Ave. Victory Margin
While there is much more parity in the pros than in Div. 1 College,
it is still apparent that the 49ers and Dolphins were the happy
owners of the two easiest schedules in the NFL.
Without further delay, here are the playoff predictions using this
algorithm. Remember, these games are all picks as though they were
going to be played this weekend, with each team at its current hot/cold
level. These figures will change as more data become available.
NFC AFC
--- ---
Rams 27 Giants 17 Raiders 26 Seattle 23
SF 34 Bears 23 Miami 31 Raiders 27
Wash 27 Rams 24 Denver 26 Pitt 24
SF 35 Wash 27 Miami 28 Denver 23
Super Bowl
----------
SF 31 Miami 27
Hope all of this was of some interest.
Jeff Heatwole, Bellcore
..!hou2e!jlh