|
An Interview with
Todd Beck
PerformanZ Ratings
NFL and NCAA Football
2000
YL: You have a prediction tracker on your site where you
monitor NCAA Football and NFL ratings. It's one of my favorite places to visit
because I like to see how my ratings are doing. What caused you to undertake the
project and will it continue?
Todd: The prediction tracker was just the natural next step after
automating my own system. The first step was making a program that would make
all my own game predictions and then score them so I wouldn't have to do it all
by hand. And once I had my own system automated it then became a rather simple
task of including others as well.
The BCS was starting its second year and had expanded to the
current 8 computer systems. So I originally planned to just follow those 8 and
my own to see how I stacked up against them. But as it turned out only half of
them gave any predictive information. The New York Times didn't even list all
the teams. So to try to make the table bigger I looked for other systems that
make predictions. After two seasons the list has expanded to 25 and it looks to
keep growing next season.
I think the prediction tracker serves several purposes. It combines
many systems on to one page, making it easy to compare the systems. Basically it
is a one stop location to get the predictions from 25 different systems on one
page. The results page then allows people to see which systems are the most
trustworthy. It also serves as an unbiased monitor to keep us all honest about
how good we are. The Harmon Forecast in USA Today boasts 72 to 78% accuracy but
in the two years I've tracked them they haven't even hit that 72% mark.
And finally, being a statistician, the prediction tracker gives me
data to play with and examine in my own personal study of the underlying
mathematical models that are being used to generate the different ratings. So as
long as the computer and time resources are still available to me I plan to
continue tracking computer ratings. Hopefully adding a little something new each
year.
YL: What is the future of computer ratings? Do you think they will ever
replace the human opinion of the polls?
Todd: So far I do not feel that the media are being very favorable to
computer ratings. All the stories I see are about how much more complicated they
make things and the people behind them considered geeks and nerds. So I think
the media coverage of computer ratings will ultimately decide how accepted the
ratings can become. And as of right now I don't know of any popular sports
figure that has ever came out and given computer ratings a strong endorsement.
And without a big name endorsement I don't see computer ratings making much more
of an impact than their current limited use. I don't think human polls will ever
be replaced and I don't think that they should be. The question should not
really be if one is better than the other because they both have strengths and
weaknesses that the other doesn't have. The only thing that needs to be done
about human polls is to make sure the person voting in the poll is actually
qualified to give an informed opinion. And a computer system should not be
included in a computer poll unless it has a sound theory behind it and has a
proven track record of superior performance.
YL: And your site provides that track record. You may be able to sell
that information later as you archive all the results. So, where did your own
interest in rating the teams come from?
Todd: My interest in computer ratings was just a natural blending of my
interests. I have been playing and watching sports all my life. And sports
statistics had a lot to do with me getting a degree in statistics in college.
I loved sports and I loved numbers.
I think there is one question that motivated me to start my own
system and look at others. That was the problem of intransitivity in football.
Team A beats Team B, Team B beats Team C and Team C beats Team A. Seeing that
you cannot have a perfect system I wondered if there existed an absolute best
system. And if it did exist was how good would it be, 60% accurate, 70%, 80%? So
my prediction tracker ultimately allows me to keep researching that question.
And from what I have seen so far I do not believe there exists one method that
is always better than the others.
YL: In calculating your ratings, what do you consider to be the most
important factors?
Todd: My system uses several factors to come up with a
rating but currently I do not give any of them any more weight than the others.
I use win/loss, points, offensive and defensive yardage and turnovers. The
theory is that the more phases of the game a team is good at the better it's
chance of winning. It is like a coach wanting to win in offense, defense and
special teams. The importance of adjusting for schedule strength certainly has
to be mentioned as well.
YL: Is your system predictive
or retrodictive?
Todd: I don't think my system can properly be labeled as
either one. In its raw form all my system does is come up with a relative
ranking based on past performance. My system is more like looking at the current
win/loss standings in a newspaper. The more wins, or in my case, more units, a
team has the closer they are to the top. I can turn this into predictions simply
by transforming the scale my units are on to a scale equivalent to points per
game. I don't call that retrodictive or predictive because I am not
intentionally trying to optimize how well it predicts past or future results.
YL: What is your profession?
Todd: I work as a statistician in the Rush Institute for
Healthy Aging at Rush-Presbyterian-St. Luke's Medical Center in Chicago. We do
research relating to health of the elderly. Alzheimer's Disease is our major
area of interest. A lot of my work involves the decline of cognitive function in
the elderly.
YL: That's a field I find particularly interesting. Aging
is a strange thing and I'm experiencing more and more of it. I can see how you
moved into football ratings with your expertise in statistics. What are some of
your other interests?
Todd: Up until this past summer I used to spend the
majority of my leisure time on the golf course. Then I came to the realization
that I wasn't very good and that I wasn't getting any better. So now I just
watch too much TV.
YL: Where is home? What else does the world need to know
about you?
Todd: I currently live in the Chicago area but I was born
and raised in Waxahachie, Texas. I got both my undergraduate degree in
mathematics and my graduate degree in statistics from Texas A&M. So I will
always be a big fan of the Aggies. And even though I am no longer in Texas I
will always be a fan of the Cowboys and the Rangers.
YL: Hooray! Go, Cowboys! (I know that will mean something
any year, now.)
We would like to thank Mr. Beck for
this interview. His web site can be found at http://tbeck.freshell.org/ Here are some other
places you will want to visit:
|