What is all of this Nonsense?
2002 Update
After 5 years away from Utah and one year of law school, I am finally getting that stupid matster's project done! I have taken the base football model used here and tweaked it to compare the results with the BCS models. If anyone is interested, a draft of the paper to go with the project is
here . It contains a much better description of the base model and all of its variations along with some fairly technical mathematical explanations and a discussion of various other issues involving this and other models of its kind. Happy reading.As a result of the paper, I am quite confident about the stability of the model. After two years of really tweaking it, I have finally concluded just how to make it work the best. But...
Another result of this project is that I have developed an alternative maodel that I expect to produce results that are as accurate and just as objective as the old model and slightly more in line with consensus, but takes less time to prepare and uses more easily available computer resources. Mostly, though, it's because I like the theory behind this new model and it was a lot of fun to program. This new model is based on an iterative Elo system that incorporates margin of victory and allowance for home field advantage. These last two additions are worked in as a result of the paper. Because this system is intended to be retrodictive, I don't plan to make results available until the schedules start to become connected. (I wrote a simple algorithm to work out this characteristic, that's also pretty fun to run. Call me a geek.)
Another addition to the rankings this year will be the inclusion of some new lists. I used to list NFL and NBA teams, but the least squares system never produced any useful rankings. The Elo system seems to be much more responsive, especially due to its retrodictive nature. I plan to reintroduce rankings for professional sports starting with a current rankning for Major League Baseball that is effective 8/25/02. This will be followed up with high school rankings for Washington, Oregon and Utah as far as available data will allow. I chose these states because I have a personal interest in each of them. Oregon information may not be very easily collected, but including high school rankings is just a trial anyway.
I am not abandoning the old model. It is still valid, and I will produce end of year results just to compare with the Elo model for my personal interest, but I don't plan to publish those results. I may bring it back after I finish the J.D.
2001 Update
Just a little update to the intro below. I have had a ball in the four years since I started regularly maintaining this model. I have recently started law school at
Nortwestern School of Law of Lewis & Clark College in Portland, Oregon. The school was chosen strictly because of its location, but like my experience from Weber, it's so far turned out to be a great choice. But the key phrase here is law school, which means no time to sleep, eat or breathe. So excuse me if I'm ever a little late getting data out.I have tweaked the model a few times over the last few years and feel pretty good about the result. I thought about what kind of things would affect a college athlete's perception of the coming game and decided that if they are playing a team that's nationally ranked or from a substantial conference, there's bound to be some intimidation. So I have started to include ESPN/USA Today and AP rankings at the time each game is played and a variable for each conference. I had feared that the ranking variables would overpower the rest of the model, but it turns out that their impact, although positive and statistically significant, is still small.
It turns out there's a much bigger impact made by the various conferences. After the model picked Penn to lose to Connecticut in the West Regional Final during the 1999 Tournament, I knew I had to do something. Obviously Penn looked great because they play in the Ivy league, but it wasn't a fair comparison. Generally, using the conference variable works well.
During the 2001 football and 2001-2002 basketball seasons I'll be running four separate models weekly and comparing the results to see which one works the best. So, sometimes the variables used will be different from week to week, but you won't be able to tell from the results.
Thanks to
David Wilson for keeping such a comprehensive College football web site. Since he listed this site, I figured I needed to decide if my model was predictive or retrodictive. I believe it's predictive because using the numbers right you can come really close to predicting the final score, even exactly right in way too many cases to not feel eery. (And I don't include all the necessary numbers to do this anywhere on this site, so don't try. The rankings are good enough a predictor.) This question pushed me into trying to include the ability to produce a preseason ranking. I got it done but there was no time to get an update up before this week (Law School again). We'll see how it goes. And this list which obviously humbles my meager offering also prompted me to change the format of the list to a complete combined ranking instead of individual division top 25s. While I'm tweaking, though, don't be surprised to see some Division II or III teams in the top spot.Since the email news source I have used for four years went bad a few months ago, I have had to find another source. A great big thanks to
Peter Wolfe who keeps the best, most reliable, most complete, most user friendly and most up to date list of college football scores on the net.I must be doing something right because I'm hitting between 70% and 80% consistently, and finished in the top five in the hunt for the coveted Ostertag Plaque yet again. (It's finally going to be my year.)
Original 1998 Introduction
There are so many numbers guys out there who think they know the secret, why would we want another one? The only real reason is for the fun of it. Since they can't all be right, and no one is going to be all right, I might as well throw my hat into the ring.
Let me say right now that my first interest is NCAA Division I basketball and everything else in this site branches off from there. So there's no confusion about why you can't find what you're looking for. If it's not here, or there's no useful information, remember that I am just starting this and I probably haven't gotten to it yet. Please be patient.
Further, as a
Weber State Wildcat fan, I used to feel that the major polls only considered already recognized schools and the smaller schools were just ignored. (How else do you explain Weber being the only 20-win Division I team not invited to post-season play in 1996?) My rankings are intended to consider every team equally without regard to popularity, television appearances, location, endowment, big name coaches, etc. What I have noticed in the short time that I have worked on my base model is that my results tend to agree well enough with those of the major polls to rule out significant favoritism. It's a hard pill to swallow.The techniques used here are similar to those used by
Kenneth Massey. The results, therefore are very similar. Ken's method is a little more complicated than mine. We both begin with the assumption that each team should score a certain number of points (offense) and stop an opponent from scoring a different number of points (defense). (I was naïve enough to think I was the only one to figure this out until I read Ken's explanation.) The combination of offense, defense, home court advantage, and statistical error results in an expected score for every game. Feel free to correct me if I'm wrong, but Ken seems to have incorporated a way of allowing for Rick Majerus playing his third string against High Point. I'm still working on that, but don't expect any changes to the system before the 1999 NBA playoffs.The indices contained in this site are the result of techniques developed during years of studying statistics at the
University of Utah, and applying them to various sporting programs. They are meant only for entertainment use. The author claims no further authority on the subject of sports statistics than has been gained through years of being a fan. (I may not be able to play anything well, but I'm really good at watching!) Any use of the information contained herin for commercial purposes without the express written consent of the author is strictly forbidden. The author is not responsible for any liability resulting from unauthorized use of any information contained within this web site.Finally, excuse the simplicity of the web design for this page. You've got to remember I'm just a numbers guy.