Every week I rate games as being zero-, two-, three-, four-, or five-star games. Five-star games (on paper) should be instant classics, while zero-star games either feature two bad teams or a huge mis-match.
I’ll try to explain the method behind how I pick which games should be the best to watch each week.
First, I pull the ratings put out by Brian Fremeau (FEI), Bill Connelly (S&P+), ESPN (FPI), and Ed Feng (the Power Rank). Each team now has four different values.
Second, I add the ratings from each team for each game to get a number that represents the overall combined quality of the two teams (higher numbers are better). For example, in the 2016 Orange Bowl featuring Florida St. and Michigan, FSU had ratings of 0.154 (FEI), 19.5 (S&P+), 21.5 (FPI), and 15.41 (PR) while Michigan had ratings of 0.257, 31.3, 27.8, and 18.58. So the “sum” values for this game are 0.411, 50.8, 49.3, and 33.99.
I then compare those sums to all possible sums based on average highest and lowest values from each rating system over the last three weeks. I then adjust those ranges to 90% of their totals to account for the fact that these extreme games are usually never played. Using FEI as an example, the highest average sum over the last three weeks before the Orange Bowl is 0.573 and the lowest is -0.461. I take that range and assign points to the top 12.75% (6 points), 25.5% (5 pts.), 38.25% (4 pts.), and 51% (3 pts.) of that range. The Orange Bowls FEI sum (0.411) falls into the top 25.5% of possible sums so it is awarded five points for that rating. With four rating systems, 24 points are possible for the sums of each team’s rating.
A great team playing a bad team will still have a high enough sum value to earn some points. So I also use difference values to represent the closeness of the match-up (smaller numbers are better) to weed out the big mis-matches.
The same process is repeated as above using the differences between each team’s rating with points being awarded to games with differences in the top 8.25% (6 points), 16.5% (5 pts.), 24.25%, (4 pts.), and 33% (3 pts.) of all possible match-ups based on a three-week rolling average. Again using the FEI values for the Orange Bowl, FSU vs. Michigan’s FEI difference of 0.103 puts it in the top 24.25% of possible match-ups, giving it four points. Again with four systems, 24 points are possible.
We now have a sum point total and a difference point total. In order for a game to be given a two- through five-star rating, it must have both sum and difference points. The Orange Bowl has 23/24 sum points and 19/24 difference points. The difference points are then weighted at 80% of the sum points giving the game an overall score of 38.2.
The top 10% of possible overall scores are rated at five stars, the top 25% at four stars, the top 45% at three stars, and the top 70% at two stars. Therefore, the Orange Bowl’s 38.2 points puts it in the top 25% and gets it a four-star rating.