![[The Sports Report Letterhead]](tsr_letr.gif)
The Sports Report Ranking System(s)
Actually, there are several different versions of TSRRS.
Glad you asked.
The first version was inspired in the late 1970s by something called the Stegeman system, when the Atlanta Journal-Constitution began running rankings generated by taking wins and losses of AP- and UPI-ranked teams and adding in strength-of-schedule based on the AP and UPI rankings of the opponent. TSRRS v0.1 took that one step further by adding scores and also doing it for all NCAA Division I teams. Even after all these years, Brigham Young and Tennessee-Chattanooga still stand out. Incidentally, Oklahoma was #1.
The next version dumped the AP and UPI rankings and was self-contained. All teams started at the same level, and points were added or lost based on the next game. Very procedural. But it worked. In 1983, TSRRS v0.2 was used for lacrosse rankings, and the system made the ridiculous assertion that Syracuse was the best team in the country. In retrospect, it was right, of course, but who knew?
Then in 1984, the pop statistician Bill James began to be published on a national level, and his Baseball Abstract changed the way a lot of people think about statistics. And TSRRS adapted too, employing Bill James' "Pythagorean Theorem" to predict a winning percentage based on points for and points allowed. This was added to schedule strength, iterations occurred, and a ranking - by percentage - was born: TSRRS v1.0.
There are several problems with the Pythagorean Theorem version of point differentiation, the least of which is that the factor is not always 2. So TSRRS now employs a standard least-squares method of correlating the factor to winning percentage. It's still around 2 for college football, but some years it's a little higher, other years a little lower.
The version you see on the web is the records-only version of TSRRS v59.x. The TSRRS 50-series rankings care about what you did against which team. It still uses the same model, but it looks at games individually. The results are often stunning. When South Carolina and Oklahoma show up near the top the second week of the season, it's a little disconcerting - whether justified or not.
We do - now, although they're more for fun. Robert Morris went undefeated in 2000, yet for extremely good reasons was considered not to be a particularly good football team. Still, if TSR doesn't vote them #1, who will?
That depends on what you're looking for.
The predictions on the web are average at best. They do not take the 3 T's into account, nor do they take into account TSRRS performance trends. Taking these into account gets you into the 60-62% range.
Yet there is one class of game that happens about 20 times a year and has consistently predicted at 68-72% for college football over the past 10 years.
No.
No, that is not true - TSR was turned down in the early 1980s.
Yes, in the late 1980s. But Herman Matthews does a better job than TSR ever could.
Overtures have been made, but we have declined to discuss it.
Rankings and Statistics in General
Updated 3-October-2007.
Copyright © 2007 by The Sports Report. All rights reserved.
Return to The Sports Report Home Page.