This page provides detailed information for the paper "Can You Beat the Odds(makers)?" that appears in the April 2005 issue of the MAA journal entitled Math Horizons. That paper describes four different rating systems, with a fifth one being eliminated from the article due to space considerations. The fifth one is referred to as the "Virtual Game Model" in the pages below, and it was the least successful system, in regards to "beating the spread", so I guess it is no great loss that it was left out of that article. However, it is an actual, deterministic simulation of the game to be played, not just an arithmetic evaluation of the performance of the teams over the current season, so it is still amazing to me that said simulation model does such a reasonable job in emulating a football game (and the final result).

The pages representing each specific year below will list every bowl game played at the end of that year, where both teams were categorized as Division 1-A, and lists the point spread and how the five systems fared both straight up and against the spread - if they would have selected a team that is different from the one that was favored.

The pages below use W in columns 1-5 to indicate the system correctly chose the victor and L when it did not. In columns 1+ to 5+, non-blank entries indicate that the system would have chosen the underdog to win the game. In those columns, Y indicates that the favored team did not cover the spread given (and so 'yes' the system would have collected if a bet had been placed on that game), N for when it did not, P is a "push", i.e. when the actual point spread was the same as listed, and T is when the rating/ranking strategy called the game a toss-up. (T is also used in the leftmost column, and columns 1-5, when the actual game ended in a tie.)

 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Summary Page - results for each year, and cumulative totals across all 21 years. (NOTE: the summary, and a few of the year by year pages above have been modified since the paper came out. These changes don't really affect the overall conclusions, but updated a couple of games where the prediction was incorrectly listed, etc.)