I will post the reports of NFL_1993, an MS-DOS application written
by Thomas Kerr. Hope you find it interesting.
(P.S. I don't know where to get the program by ftp. I think the
filename to search for is nfl93.zip....tell me if you find it.)
Here is a quote from the .doc file about the methodology and issues
behind the computer predictions.
[T]he program is very simple. It processes game scores and
maintains a simple, two parameter database on each team. The data
can be used to predict the outcome of next week's games. Based on
statistics, the program should predict 70% of the games correctly
between now and infinity. Typically, over the last nine seasons,
the predictions have been pretty consistent in picking 9 to 10 winners
per week out of 14 games. For three years, I tried some pretty
sophisticated, numerical analysis techniques to improve the accuracy.
I tracked team performance, in terms of offensive and defensive
ratings, as a function of whether the teams were on the road or at
home. I tried Dow-Jones type time series analysis - moving averages
and weighted moving averages based on the last two weeks, three weeks,
four weeks ... nine weeks, etc. By the time I was done, I had
"successfully" lowered my prediction rate to low sixties. The more
complicated I made the model or the more parameters I added for
model sophistication, the worse I did. Finally, I gave up and went
back to the two parameter model -- how does each team perform, offen-
sively and defensively, relative to the average NFL team performance
of 21 points scored per game.
Over the course of a season, NFL teams average 21 points scored per
game. And typically, a team that puts 28 points on the board
per week and limits the other guy to 14 points per week generally
comes out ahead of the team that puts 14 points on the board while
giving up 28 points. In simple terms, that's the line taken by
When you stop and think about all the sophisticated statistics the
commentators quote, they don't really do any better. Average points
scored on the road in the fourth quarter - percentage of third down
conversions - turnover ratio - number of real blonde cheerleaders
between the 20-40 yardline markers. Statistically, the best team
seems to win about 70% of the time - the rest of the time, it's a
***shoot. And there's always one week per season when almost every
favored team loses.
Greg Franklin "Shopping on stones, wokking on coals,